Content uploaded by Svenn-Erik Mamelund
Author content
All content in this area was uploaded by Svenn-Erik Mamelund on Jan 26, 2016
Content may be subject to copyright.
Cet article est disponible en ligne à l’adresse :
http://www.cairn.info/article.php?ID_REVUE=POPE&ID_NUMPUBLIE=POPE_402&ID_ARTICLE=POPE_402_0229
Can the Spanish Influenza Pandemic of 1918 Explain the Baby Boom of 1920
in Neutral Norway ?
par Svenn-Erik MAMELUND
| Institut National d’Etudes Démographiques | Population
2004/2 - Volume 59
ISSN 1634-2941 | pages 229 à 260
Pour citer cet article :
— Mamelund .-E., Can the Spanish Influenza Pandemic of 1918 Explain the Baby Boom of 1920 in Neutral Norway ?,
Population 2004/2, Volume 59, p. 229-260.
Distribution électronique Cairn pour Institut National d’Etudes Démographiques.
© Institut National d’Etudes Démographiques. Tous droits réservés pour tous pays.
La reproduction ou représentation de cet article, notamment par photocopie, n'est autorisée que dans les limites des
conditions générales d'utilisation du site ou, le cas échéant, des conditions générales de la licence souscrite par votre
établissement. Toute autre reproduction ou représentation, en tout ou partie, sous quelque forme et de quelque manière
que ce soit, est interdite sauf accord préalable et écrit de l'éditeur, en dehors des cas prévus par la législation en vigueur
en France. Il est précisé que son stockage dans une base de données est également interdit.
Population-E
2004, 59(2), 229-260
Can the Spanish Influenza Pandemic
of 1918 Explain the Baby Boom
of 1920 in Neutral Norway?
Svenn-Erik M
AMELUND
*
The baby boom of 1920 has received less attention than that
following the Second World War, and was for long thought to be
a simple catching up of the marriages and births that the war had
prevented, as part of a gradual return to normalcy. Yet this baby
boom occurred with similar intensity in non-belligerent countries.
Other lines of enquiry thus need to be explored. In this article
dealing with the large number of extra births registered in 1920
in Norway, a country that was neutral during the war,
Svenn-Erik
M
AMELUND
examines the hypothesis, sometimes proposed but
never verified, concerning an influence of the Spanish flu epide-
mic of 1918 on the decline in fertility in 1919 and its strong up-
swing in 1920. This particularly virulent epidemic is known to
have touched nearly a quarter of the world’s population and to
have caused between five and ten times as many deaths as the war.
Because influenza was a notifiable disease in Norway, the author
is able to examine on a monthly and regional basis the influence
on fertility of the morbidity and mortality attributable to the flu.
He concludes that the flu pandemic was indeed the main cause of
the Norwegian baby boom of 1920 and suggests that it probably
had a similar influence in other countries.
The European countries that were massively affected by the First
World War, whether by sending soldiers to the front, fighting the war on
their own soil, or both, experienced a sharp decline in fertility during the
period 1914-1918 (Chesnais, 1992). Figure 1 shows a steep decline of fer-
tility in Italy, France and Germany during the war. The fertility curves for
other warring nations, for example Austria-Hungary, England and Wales,
Scotland, Ireland, and Bulgaria, not shown, were very similar. The low
fertility in the belligerent countries during the war is probably explained
by the separation of soldiers from their wives and by the war having hin-
dered young adults from marrying. In addition, it is reasonable to assume
* Department of Economics, University of Oslo, Norway.
230 S.-E. M
AMELUND
that those who were not enrolled in the armed forces postponed marriages
and births from fear of the war and its consequences. The war thus created
a huge potential for a compensating later resurgence in marriages and
births as well as a “moral obligation to replace the deceased”. Indeed,
birth rates in most belligerent countries began to rise once again in 1919
(for France in 1917), so that by 1920 they had equalled or surpassed the
pre-war level.
The fertility patterns of nations that remained neutral, Norway,
Sweden, the Netherlands, Denmark, Switzerland and Spain, were not
affected in the same way (Chesnais, 1992). Figure 2 shows that the birth
rates in Norway, Sweden, and the Netherlands declined relatively
smoothly and continuously throughout the war until 1919. (The pattern of
these countries is also reasonably representative of Denmark, Switzerland,
and Spain.) This decline, however, is typical of the transition from higher
to lower birth rates that have marked all societies throughout Europe.
Hence, no substantial decline in birth rates linked to the war seems to have
occurred in neutral countries. Nonetheless, like the belligerent countries,
neutral countries faced a baby boom in 1920 (Figure 2).
France CBR
Germany CBR
Italy CDR
Germany CDR
France CDR
Italy CBR
1905 1910 1915 1920 1925 1930
Year
Deaths/births per 1,000 population
Ined 163 04
5
10
15
20
25
30
35
40
Figure 1.– Crude death rates (CDR) and crude birth rates (CBR)
for three belligerent countries, Italy, Germany, and France, 1905-1930
(per 1,000)
Source: Chesnais, 1992.
T
HE
S
PANISH
I
NFLUENZA
P
ANDEMIC OF
1918 231
Surprisingly, there are few studies that have analysed the baby boom
in Europe after the First World War in any depth, in contrast to the more
famous baby boom following the Second World War. Perhaps a natural
explanation for the 1920 baby boom, especially for the warring nations,
was the gradual return to normal patterns of life after the war (Henry,
1966; Winter, 1977). Yet despite this, the birth rates in 1919 seem to be far
lower than pre-war figures. Could there have been other factors at that
time to cause people to postpone the start of a family? There is also a clear
discrepancy between birth rates in 1919 and those of 1920 that perhaps
goes beyond the most immediate explanation, namely the First World War.
This article directs attention to a hitherto little appreciated factor, the
Spanish influenza pandemic. It spread around the globe in three waves in
1918. The first bout of influenza appeared during the months from March
to May, but at this point in time the disease was not very contagious and it
claimed few lives. In mid-June, however, the influenza returned, spread
fast, and reached pandemic proportions. Many came down with the flu dur-
ing the summer wave from July through September, but still relatively few
died from it. The Spanish influenza made a third appearance during the
autumn from October through December. The influenza virus had mutated
since the relatively mild summer outbreak and was now highly lethal to
1905 1910 1915 1920 1925 1930
Year
Deaths/births per 1,000 population
Ined 164 04
5
10
15
20
25
30
35
40
CBR
CDR
Netherlands
Sweden
Norway
Netherlands
Sweden
Norway
Figure 2.– Crude death rates (CDR) and crude birth rates (CBR)
for three neutral countries, Norway, Sweden, and the Netherlands, 1905-1930
(per 1,000)
Source: Chesnais, 1992.
232 S.-E. M
AMELUND
those who were infected. However, those with a previous infection were
partially protected against later waves due to the acquisition of immunity.
Spanish influenza affected at least 500 million persons or over one
fourth of the world’s population at that time (Laidlaw, 1935). According to
the most recent revision and update the global death toll was between 50
and 100 million (Johnson and Mueller, 2002). Hence the number of vic-
tims of Spanish influenza exceeded the number of casualties from the First
World War by five to ten times. People of all ages experienced an increase
in mortality, but those who suffered the most were people in the fertile
ages of 20 to 40 years who normally have little to fear from influenza. In
addition to influenza, pneumonia was the most important cause of death
during the pandemic. Figures 1 and 2 show for six European countries the
great impact of the Spanish influenza on the crude death rates in 1918.
Table 1 shows that the death rates from Spanish influenza were relatively
homogeneous across Europe, ranging from 3.3 to 7.3 deaths per 1,000, and
with no notable difference between neutral and belligerent countries.
Exceptions are the relatively high death rates in neutral Spain and in bel-
ligerent Italy, Hungary, and Portugal.
Höijer (1959) was one of the first to suggest that the Spanish influ-
enza of 1918 may have caused the baby boom in Europe in 1920. He spe-
cifically showed for neutral Sweden that there was a stunning decline in
the monthly conception rates (birth rates set back nine months) when the
death rates peaked during the autumn of 1918, resulting in low birth rates
T
ABLE
1.– E
STIMATED
NUMBER
OF
DEATHS
AND
DEATH
RATE
(
PER
1,000)
FROM
THE
SPANISH
INFLUENZA
,
SELECTED
EUROPEAN
COUNTRIES
Country Death toll Death rates
Neutral
Norway 14,676 5.7
Sweden 34,374 5.9
Denmark 12,374 4.1
Switzerland 23,277 6.1
Netherlands 48,042 7.1
Spain 257,082 12.3
Belligerent
France 240,000 7.3
Italy 390,000 10.7
Germany 225,330 3.8
Hungary ~100,000 12.7
Austria 20,458 3.3
England and Wales ~200,000 5.8
Scotland 27,650-33,771 5.7-6.9
Ireland 18,367 4.3
Portugal 59,000 9.8
Finland 18,000 5.8
Source
: Johnson and Mueller, 2002, Table 4, p. 113.
T
HE
S
PANISH
I
NFLUENZA
P
ANDEMIC OF
1918 233
during the summer of 1919 and consequently high birth rates in 1920.
However, the author did not undertake a detailed analysis of the monthly
conception and death rates, and did not discuss the relationship between
these factors in any greater detail. Höijer’s work did show, however, the
importance of analysing this phenomenon on a monthly basis.
Other studies that have considered the hypothesis that Spanish influ-
enza in 1918 caused the baby boom of 1920 have also been univariate and
descriptive (Pool, 1973; Rice, 1983; Underwood, 1983; Mills, 1986;
Johnson, 2002). Although previous studies have acknowledged that the in-
crease in influenza morbidity and mortality may have affected fertility
negatively, none was able to determine whether morbidity had an inde-
pendent effect on fertility net of the effect of mortality between 1918 and
1920. Nor could previous studies determine whether morbidity had a spe-
cific effect on fertility during the three outbreaks of influenza in 1918. In
contrast to previous studies, this article attempts to separate the effects of
morbidity and mortality on fertility across time and space using cross-
sectional data. The reason why this separation is important is that the time
it takes for a couple to resume sexual relations after an interruption due to
disease—and thus the time it takes to achieve a conception—would be
shorter than for a person who has lost a spouse to influenza.
The main hypothesis that is tested in this article is that the 1918 flu
caused the baby boom of 1920 in Norway. In order to do so, we use multi-
variate models to analyse monthly morbidity, mortality and fertility data
for 37 rural and urban areas in 20 Norwegian counties. A study of Norway
should be valuable for three reasons. First, the country was neutral
throughout the war, and this is assumed to control for the influence of the
war on fertility and nuptiality. Indeed, as the number of military casualties
was insignificant in Norway (Mamelund, 2003b), the war did not cause a
“lost generation” and thus a “demand” to replace the dead and to sustain
the population by giving birth. In discussing the case of Great Britain,
Johnson (2002, p. 230) states that “quantifying [the decline and subse-
quent increase in births], particularly the portion due to influenza and that
due to the effects of war, is virtually impossible”. Johnson explains this by
“the fact that both events fell so heavily on the same segments of the po-
pulation, the young adults. Further, the disruption of population by war
rendered the basic population data unreliable” (p. 231). The second asset
of Norway is that registration of population data, including vital statistics,
continued normally and was on the whole not affected by the war. The
third asset is that influenza was a reportable disease in Norway in 1918 in
contrast to most other countries. Neutral Norway is thus one of very few
countries for which it is possible to carry out a multivariate study to esti-
mate the effect of influenza morbidity on fertility net of the effect of
war
and
influenza
mortality over the years 1918-1920.
The present article first presents the framework of the demographic
developments during a mortality-fertility crisis and discusses the underly-
234 S.-E. M
AMELUND
ing reasons why fertility was affected by the Spanish influenza, with spe-
cial reference to the Norwegian case. After a description of the data and
the indicators used in the study, we then proceed to describe the course of
fertility and nuptiality from 1914 to 1918 in Norway. The effects on ferti-
lity and mortality that the outbreak of Spanish influenza may have had are
thoroughly examined. Finally, we present the results of a multivariate
analysis using least squares regression models.
I. The time profile of a combined
mortality-fertility crisis
1. The general framework
The general framework that describes the demographic changes that
commonly occur during and after a combined mortality-fertility crisis is
based on work by Juhasz (1971), Preston (1978), Menken et al. (1981),
Watkins and Menken (1985), Palloni (1988), Wrigley and Schofi eld
(1989), Livi Bacci (2000), and in particular Lee (1989, 1990). There is fair
agreement in the literature that four
phases
can be distinguished in con-
nection with epidemic crisis mortality:
Phase 1: A sudden increase in morbidity and mortality is accompa-
nied by a drop in coital frequency and conceptions below the normal level.
Phase 2: Coital frequency and conceptions drop further and reach the
lowest relative levels when morbidity and mortality peak. The increase in
the number of marriages dissolved by death leads to a further decline in
coital frequency and conceptions.
Phase 3: A drop in morbidity and mortality either at or below pre-
crisis level (due to decrease in the number of non-immune individuals and
to negative selection of the frail) is followed by a surplus of conceptions
compared to the normal level as couples compensate for the deficits in
Phases 1 and 2 and “replace” those children who have died.
Phase 4: Fertility rebounds a year or two after the crisis to levels
higher than the normal pre-crisis level (because of effects of compensation
and replacement).
2. The Spanish influenza
and fertility between 1918 and 1920 in Norway
Negative perceptions about the future may lead to voluntary post-
ponements of births (Lee, 1989; Menken et al., 1981). It is conceivable
that the fear induced by the pandemic itself (people could die within a
T
HE
S
PANISH
I
NFLUENZA
P
ANDEMIC OF
1918 235
three-day period) as well as a fall in social integration (the closings of
schools, churches, theatres, and the banning of public meetings so as to
prevent the spread of the disease) affected the desire for children nega-
tively when pandemic mortality peaked during the autumn of 1918.
There was a widespread fear of death, and anxiety was intensified
when it was realized that physicians were helpless. There were no vac-
cines or effective antiviral drugs. Two folk memory quotations describe
the everyday atmosphere of fear. One woman recalled that “Aftenposten [a
major newspaper of the capital] was full of death notices. I read them all
every day, and it was awful”. Another woman remembered that “everyone
was afraid of everyone else, the contagion was everywhere” (Mamelund,
1998). The terror and fear caused by the disease had a profound psycho-
logical impact on people and exerted a powerful influence on the choices
they made. It seems reasonable therefore to believe that even people who
were not struck down by the flu would have decided to wait before they
had a baby.
There are two reasons to assume that a couple would reduce sexual
intercourse if one of the spouses became infected (Phase 2). First, the
physical condition of the patient (high fever, headache, etc.) inhibited
sexual activity for 2 to 4 weeks (Mamelund, 1998). Second, in order to re-
duce the risk of infecting one another, couples not previously infected may
have decided to abstain from sexual intercourse even though able and will-
ing (Mills, 1986). This assumption is grounded in the belief that the
Norwegian health authorities’ massive public information campaign
through notices in newspapers and posters in public places was effective
and had a preventive effect. People were urged to cover the mouth when
coughing, to go to bed as soon as possible upon the onset of illness and to
remain isolated from the rest of the family until they were free of symp-
toms. In light of this, it seems reasonable to assume that couples in which
both spouses had survived would have decided to try for a pregnancy in
1919 (Phase 3), thereby leading to a baby boom in 1920 (Phase 4).
According to Norwegian marriage law at the time, a widow was not
permitted to remarry before she had spent at least a year of mourning.
Although the same law did not apply to men, it was nevertheless also cus-
tomary for widowers to have a year of mourning. Assuming therefore that
the widowed did not remarry in 1918 or 1919, that there was little sex out-
side of marriage, and that the widows were not pregnant before their hus-
bands died, neither widows nor widowers would have been able to achieve
a conception in 1918 or in 1919 to compensate for the postponed concep-
tions of 1918. If this mechanism works as described, it would also impede
the rebound potential of fertility from being fully released until the first
half of 1920.
Pregnant women, especially those in the last trimester, were ex-
tremely vulnerable to spontaneous abortion, stillbirth and maternal death
if Spanish influenza was followed by pneumonia (Harris, 1919; Bourne,
236 S.-E. M
AMELUND
1922). Two contemporary investigations report that anywhere from a fifth
to half of pregnant women with pneumonia died (Bland, 1919; Harris,
1919). This certainly had a negative effect on the birth rate in 1918. Infant
and child mortality on the other hand may have affected fertility positively
(see Preston, 1978). Parents who lost infants and young children to
Spanish influenza in 1918 may have wished to have another baby to
achieve a certain family size, thereby increasing the number of concep-
tions in 1919. This may have contributed to the baby boom in 1920.
The probability of conception may have dropped in 1918 (Phases 1
and 2) due to stress imposed by the Spanish influenza. Support for this
hypothesis is found in the literature. Biraben (1973) for example, reports
that males, but not females, have become temporarily sterile in connection
with influenza epidemics. If men also experienced temporary sterility
during the Spanish influenza, this would inhibit the making-up in 1919
(Phase 3) of the conceptions postponed in 1918 (Phase 2). Changes in fe-
male fecundity associated with Spanish influenza may have stimulated
conceptions, but the biological factors go in both directions. On the one
hand, women who failed to conceive in Phase 2 were not pregnant in
Phase 3. They were therefore at risk of another pregnancy after a short
while, even during the peak of the crisis (Juhasz, 1971). In addition,
women who failed to conceive would not have been breastfeeding. This is
expected to give extra “power” to the making-up process. The net effect of
high (male) morbidity rates (which is assumed to give relatively high
temporary male sterility) and high spontaneous abortion and stillbirth
rates for women in Phases 1 and 2, is assumed to be negative, thereby in-
hibiting the potential rebound of conceptions in Phase 3.
II. Available data and indicators used in the study
1. Conceptions and the crude conception rate
The regional monthly live births, stillbirths and population at risk
that are used to calculate the decline in the crude conception rate asso-
ciated with Spanish influenza come from the annual population reports
Folkemengdens bevegelse
published by Statistics Norway (
Statistisk sen-
tralbyrå
, cited henceforth as SSB). From 1919 onwards live births and
stillbirths were compiled by place of residence of the parents (
de jure
po-
pulation), while the practice in previous years was to compile the data for
the present population of parents (
de facto
population). Unfortunately,
stillbirths are probably underreported, but there have apparently not been
large regional differences in their underreporting over time. The seasonal
patterns of foetal mortality before the fifth month of pregnancy are not
known. However, the seasonality of foetal mortality in the 5-6th month of
T
HE
S
PANISH
I
NFLUENZA
P
ANDEMIC OF
1918 237
pregnancy is available, but only for the whole nation. The crude concep-
tion rate is therefore defined as conceptions ending either in a stillbirth (7-
9 months later) or a live birth (9 months later) per 1,000 people per month
(Leridon, 1977; Wrigley and Schofield, 1989).
2. Influenza and
pneumonia cases and the morbidity rate
The monthly reported influenza and pneumonia cases by region,
taken from the annual health reports
Sundhetstilstanden og medisinal-
forholdene
, published by
Det civile medisinalvesen
(cited henceforth as
DCM), constitute the raw data input for calculating the increase in the in-
fluenza and pneumonia morbidity associated with Spanish influenza. The
population at risk is taken from the annual population report
Folke-
mengdens bevegelse
published by SSB.
In contrast to most other countries, influenza was a reportable
disease in Norway in 1918. After some time, however, it was realized that
reported
cases of influenza and pneumonia fell short of
actual
morbidity
(DCM, 1922). The disease probably struck 1.2 million Norwegians
whereas only a third of the cases were reported (Mamelund, 1998). The
underreporting is explained by the short supply of doctors, and in some
cases may be attributed to the long distances to doctors and hospitals,
especially in the rural parts of Finnmark County in the high north.
Furthermore, in many families doctors were only called for the bread-
winner or the person who had to claim sick benefits when reporting unfit
for work to a health insurance fund (Kristiania Sundhetskommision,
1919). More people employed in the secondary and tertiary sectors and
living in cities had health insurance than people employed in the primary
sector living in rural areas. It was the impression of the medical profession
that underreporting was generally lower in cities than in rural areas
(Mamelund, 1998).
With respect to underreporting over time, the physicians claimed that
it was lowest during the autumn wave of 1918. This was explained by the
increase in lethality of the Spanish influenza from the summer to the au-
tumn, making reporting of a case more likely in the latter period. For
Norway’s second largest city, Bergen, this assumption is confirmed. When
comparing the figures from an “influenza census” of the Bergen population
with official statistics, it is estimated that a third of all cases were reported
during the summer and almost 100% during the autumn (Hanssen,
1923)
(1)
. The “influenza censuses” are the only source that gives a reliable
picture of morbidity. Similar “influenza censuses” carried out for several
(1)
The census covered 10,000 persons or one-ninth of the population and was carried out
by nurses who asked heads of household whether any of the members were sick or died from in-
fluenza during the different waves of 1918.
238 S.-E. M
AMELUND
cities in England and in the United States generally confirm the finding
from Bergen (Great Britain Ministry of Health, 1920; Vaughan, 1921;
Collins, 1931; Sydenstricker, 1931; Britten, 1932).
The influenza and pneumonia morbidity rate is defined as the number
of
cases
per 1,000 in the present (
de facto
) population
reported
to district
physicians or to doctors in hospitals per month. The morbidity rate may be
higher than 1,000 as a person might visit a physician because of influenza
or pneumonia more than once a month. Unfortunately, the data needed to
estimate
duration
of the disease for the 37 units of analysis do not exist.
Obviously, the morbidity rate alone cannot satisfactorily explain the as-
sumed decline in the conception rate. A sick leave of 1-2 weeks for mild
cases as compared to a sick leave of 3-4 weeks for severe cases would
clearly have different effects on the conception rates.
3. Deaths and the crude death rate
Unfortunately, monthly numbers of influenza and pneumonia deaths
do not exist for the units of analysis used in the article. However, monthly
numbers of deaths from all causes are available from the annual popula-
tion report
Folkemengdens bevegelse
published by SSB. Mamelund (1998)
has shown that the calculation of excess mortality for all causes of death,
compared to the previous years, gives a good approximation of the morta-
lity associated with the Spanish infl uenza. Therefore, excess in the
monthly number of deaths from all causes in each of the 37 regions in
1918 is used to estimate the increase in mortality associated with the
Spanish influenza. Although doctors were overworked and failed to report
complete morbidity fi gures, every death was probably reported
(Mamelund, 1998). The crude death rate in 1918 is defined here as the
number of deaths per month for 1,000 persons in the present population,
while in 1919 the crude death rate is the number of deaths per month per
1,000 people with permanent residence.
III. Combined fertility-mortality profiles
during the 1914-1918 war and during the successive
waves of influenza in Norway from 1918 to 1919
To analyse the effect of Spanish influenza on fertility, we will con-
front the observed seasonal fluctuations of fertility and mortality
(2)
during
the period 1918-1920 with the reference seasonal levels of the same demo-
graphic indicators calculated for the period 1915-1917. Let us first justify
our choice of the 1915 to 1917 period as a reference.
(2)
Seasonal fluctuations in total mortality, incidence of and mortality from influenza and
pneumonia, spontaneous abortions, stillbirths, maternal deaths, live births, and marriages dis-
solved by death.
T
HE
S
PANISH
I
NFLUENZA
P
ANDEMIC OF
1918 239
1. 1915-1917 as reference years
The period 1915-1917 is chosen because the general level of fertility
should not be very different from that of the years immediately following.
Moreover, Cassel (2001) has shown for Sweden, which shares climate and
cultural traditions with Norway, that the First World War did not affect the
seasonality of conceptions. Both pattern and levels of conceptions, there-
fore, provide an appropriate standard of comparison with the time of the
influenza pandemic.
In the period 1915-1917 there were no influenza
pandemics
. How-
ever, annual influenza
epidemics
were reported as usual in the months
from October to March
.
The morbidity and mortality of the three influenza
epidemics in the period 1915-1917 are representative of previous epi-
demics in that a large part of the population had immunity to fight the
viruses, and that influenza and pneumonia mortality, which was generally
low, was highest for the very young and the elderly. It seems therefore rea-
sonable to use the years 1915-1917 as a baseline for comparing the pan-
demic of 1918-1919.
Some of the same arguments for selecting 1915-1917 as norm for
fertility also apply for mortality. First, the death rates for the nation as a
whole did not increase during the war, as shown in Figure 2 (Mamelund,
2003b). Second, because the period under analysis is part of a long period
of mortality decline, the
level
of mortality would probably be exaggerated
if years prior to 1915-1917 had been included to define normal conditions.
2. The Norwegian fertility-
nuptiality-mortality regime from 1914 to 1919
Birth rates 1914-1918
There was no dramatic change in fertility in Norway between 1913
and 1918 (Table 2). Indeed, total fertility fell by 7% from 1914 to 1915,
perhaps because of a decline in conceptions in the months following the
outbreak of war in July-August 1914, but in 1917 the total fertility rate
rose slightly. The decline in fertility from 1914 to 1915 may have created a
potential for a compensating rebound, although at first glance it was not
realized in the following years. The relatively high and stable fertility in
the years 1916-1918 occurs, however, during a long period of more or less
constant decline in fertility from the 1890s to the mid 1930s (Brunborg
and Mamelund, 1994). It is therefore likely that most of the potential for
making up the conceptions postponed in 1914 was realized in the years
1915-1918.
240 S.E. M
AMELUND
T
ABLE
2.– L
IVE
BIRTHS
AND
BIRTH
RATES
BY
LEGITIMACY
(
PER
1,000),
AND
TOTAL
FERTILITY
RATE
, N
ORWAY
1913-1923
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
Live births 61,294 62,111 58,975 61,120 63,969 63,468 59,486 69,326 64,610 62,908 61,731
Legitimate 56,881 57,608 54,881 56,753 59,516 59,256 55,600 64,065 59,725 58,575 57,658
Illegitimate 4,413 4,503 4,094 4,367 4,453 4,212 3,886 5,261 4,885 4,333 4,073
Birth rate 25.0 25.1 23.6 24.2 25.0 24.6 22.9 26.3 24.2 23.3 22.8
Legitimate 23.2 23.3 22.0 22.5 23.3 23.0 21.4 24.3 22.4 21.7 21.3
Illegitimate 1.8 1.8 1.6 1.7 1.7 1.6 1.5 2.0 1.8 1.6 1.5
Total fertility 3.6 3.6 3.4 3.4 3.5 3.4 3.2 3.6 3.3 3.2 3.1
Source
: Brunborg and Mamelund, 1994; SSB, 1995.
T
ABLE
3.– N
UMBER
OF
MARRIAGES
AND
MARRIAGE
RATE
(
PER
1,000
UNMARRIED
PERSONS
), N
ORWAY
1913-1923
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
Marriages 15,262 15,773 15,940 17,312 18,086 20,019 15,379 18,460 18,063 17,185 16,963
Rate 9.2 9.4 9.4 10.2 10.5 11.5 8.8 10.4 10.0 9.5 9.3
Note
: The unmarried population is estimated as the total population excluding the married and the separated.
Source
: Mamelund, Brunborg and Noack, 1997.
T
HE
S
PANISH
I
NFLUENZA
PANDEMIC OF 1918 241
Marriage rates 1914-1918
As may be seen from Table 3, there was a boom in marriages during
the war. SSB (1926) explains the increase in marriages by the economic
boom, by relatively large birth cohorts reaching marriageable age, and by
low rates of emigration (emigration rates were around 4.0 per 1,000 popu-
lation in the years 1913-1914, lowest in 1918 with 0.5 per 1,000, and
around 2.0 per 1,000 in the period 1920-1922). This finding seems to sup-
port the hypothesis that the war did not lead people to postpone their mar-
riages.
The increase in the number of marriages concluded from 1917 to
1918 and the decline in marriages from 1918 to 1919 are explained by
more restrictive marital legislation that went into effect on 1 January
1919, rather than by the armistice or Spanish influenza (SSB, 1926). The
most important change in the law was a rise in legal age of marriage, from
16 to 18 years for women and from 18 to 20 years for men. According to
SSB (1926), this resulted in an increase of 2,000 more marriages in
December 1918 compared to normal figures, and 2,000 fewer marriages
during the first half of 1919. If the figures in Table 3 are adjusted accord-
ing to this displacement, it will be seen that there was no increase in the
number of marriages from 1917 to 1918 and only a moderate decline by
640 marriages from 1918 to 1919. A re-analysis of monthly marriages for
the purpose of this article (not shown here) led to the same conclusions as
SSB. Hence, it seems safe to conclude, at least for the nation as a whole,
that the baby boom in 1920 was not caused by a marriage boom after the
armistice or by the withdrawal of Spanish influenza. Likewise, the result
indicates that few marriages were postponed because of the disease. More-
over, since there were no apparent changes in the number of marriages be-
cause of Spanish influenza, one would not expect any indirect negative
effects of marriage on the number of conceptions. As data on monthly
marriages are published only at the level of the whole nation, marriage
cannot be included in the multivariate analysis.
The 1918 spring wave of Spanish influenza
Approximately 6,000 people were off from work with infl uenza
symptoms in the first week of April 1918 (Mamelund, 1998), but these
cases were apparently not reported in the official statistics as influenza, as
can be seen in Figure 3. A suspiciously high number of influenza cases
were also reported in military camps, but the disease did not spread to the
general population. At this stage of the pandemic, mortality was low and
deaths were only reported among population groups where fatalities from
influenza are most common, that is the very young and the elderly. The
death rate from January to June 1918 was also lower than the “normal”
seasonal levels. This is rather surprising as this was a time of relatively
strict food rationing (Mamelund, 2003b). Nevertheless, the increased in-
242 S.-E. MAMELUND
fluenza activity in April may explain why the conception rate declined by
0.7 conceptions per 1,000 from March to April rather than increasing by
1.9 conceptions per 1,000 as it did on average in 1915-1917 (Figure 3).
The fall in the conception rate by 1.0 conception per 1,000 from April to
May might be expected, but it occurs at a rate of 2.5 conceptions per 1,000
fewer than normal.
On 28 May 1918, Norwegian newspapers printed a wire from the
Reuters news agency revealing that a seemingly new type of influenza was
spreading throughout Spain. Only two weeks later, on 15 June, the very
first scattered cases of influenza, which later proved to be part of a pan-
demic wave, occurred in the capital of Kristiania (renamed Oslo in 1924)
(Mamelund, 1998). It was not before the first week of July, however, that
the number of reported cases skyrocketed to epidemic proportions (see
Figure 3). One out of the four conceptions per 1,000 that were postponed
in total between April and December 1918 would have occurred in the
second quarter from April to June. It would appear that the decline in con-
ceptions in this period can be attributed to the scattered cases of influenza
in April and possibly also to a voluntary postponement of conceptions due
to negative perceptions about the near future.
The 1918 summer wave of Spanish influenza
The conception rate normally rises and reaches the highest level of
the year during the three summer months ending in August. This was not
true in 1918. Instead of a normal seasonal increase of 1.4 per 1,000 from
June to August, the conception rate decreased by 1.0 per 1,000. Of the de-
cline of 1.7 conceptions per 1,000 from August to September, 40% is ex-
plained by normal seasonal fluctuations. The remaining 60% is probably
explained by a decline in coital frequency in connection with the increase
in the morbidity rate and the death rate during the first pandemic wave of
Spanish influenza. Of the total decline in conceptions from April to
December 1918, 44% would have occurred during the third quarter from
July to September.
The third quarter of 1918 was not the quarter with the highest
increase from the 1915-1917 norm in mortality in 1918, but according to
the only “influenza census” in Norway, that of Bergen, as well as to
reports from physicians from other parts of the country, actual morbidity
was clearly highest in this quarter as the flu met a “virgin soil” population
(Hanssen, 1923; Mamelund, 1998). Upon examination of the morbidity
curve in Figure 3, however, it seems that reported morbidity was higher
during the autumn of 1918 than during the summer of 1918 (574 cases
per 1,000 in October versus 297 in July). As discussed above, this is due to
underreporting of cases during the summer wave because of the mild char-
acter of the virus at this stage, while there was no or little underreporting
during the highly virulent autumn wave. Actual morbidity for July 1918 is
THE SPANISH INFLUENZA PANDEMIC OF 1918 243
far higher than reported morbidity for October 1918, which is assumed to
be close to actual morbidity. The increase in the death rate during the au-
tumn wave was twelve times as high (8.6 deaths per 1,000) as the corre-
sponding increase in the death rate during the summer wave (0.7 deaths
per 1,000). To conclude, morbidity seems to have been more important
than mortality in explaining the decline in the conception rate during the
summer wave of relatively mild Spanish influenza.
The 1918 autumn wave of Spanish influenza
The average of 21.6 conceptions per 1,000 in the period from
September to November 1918 is nearly 4.0 conceptions per 1,000 lower
than normal, a fact that is probably due to the high morbidity and death
rates during the autumn wave (Phase 2). Almost 60%, or 2.8, of the in-
crease of 4.8 conceptions per 1,000 in the rate between November (21.7
per 1,000) and December (26.5 per 1,000) may be explained by normal
seasonal fluctuations (Figure 3). The other 40% are probably explained by
a strong decline in both the morbidity of influenza and mortality.
Monthly average
conception rate 1915-17
Observed monthly conception
rate from 1918 through 1920
Observed monthly death
rate from 1918 through 1920
Monthly average
death rate 1915-17
Observed monthly
morbidity rate from
1918 through 1920
Deficit in
conceptions
Surplus in
conceptions
Excess in
all-cause mortality
Deaths (all causes) and
conceptions per 1,000 per month
Influenza and
pneumonia per 1,000 per month
Ined 165 04
5
10
15
20
25
30
35
0
100
200
300
400
500
600
700
135791113579111357911 Months
Years
19181918 1919 1920
Spanish Influenza
End of WWI
Figure 3.– Monthly influenza and pneumonia morbidity, and differences
of the crude death rate and conception rate from the monthly average
for 1915-1917, Norway 1918-1920 (per 1,000)
Sources: DCM, 1922, 1923, 1924; and SSB, 1920a, 1920b, 1921a, 1921b, 1923a, 1923b.
244 S.-E. MAMELUND
Of the total decline in conceptions from April to December 1918, it
appears that 31% can be accounted for by the fourth quarter of 1918. The
reason why a larger proportion of the decline in conceptions occurred dur-
ing the third quarter (44%) than during the more lethal fourth quarter
(31%), must be the fact that actual influenza morbidity for men and
women in their most reproductive ages 20-40 was much lower in the
fourth quarter than during the third quarter of 1918, thanks to a gain of
relative immunity (Hanssen, 1923). Figure 3 indicates that the curve for
reported morbidity was highest in the last quarter of 1918, but this is again
probably due to the virus becoming more virulent, with more persons re-
porting themselves sick. To sum up, it seems that mortality was more im-
portant than morbidity in explaining the downturn in the conception rate
during the lethal autumn wave.
The pandemic subsides:
A rebound of conceptions in 1919
The increase in the conception rates in 1919 and during the first three
months of 1920 (Phase 3) from the 1915-1917 norm more than compen-
sated for the decline in the conception rates of the last nine months of
1918 from the 1915-1917 norm (Figure 3). Although the conception rate
increased from November 1918 to March 1919 (positive excess rates com-
pared to normal seasonal fluctuations in January-March 1919 only), births
declined by 4,000 (0.26 births per woman) to 60,000 from 1918 to 1919
(Table 2, Figure 3). Obviously, conceptions that might have been expected
to occur during the last nine months of 1918 were postponed to 1919 be-
cause of influenza (SSB, 1926 came to the same conclusion). Nearly
70,000 births were recorded in 1920 (Table 2). This was about 10,000
more, or 0.44 more births per woman, than in 1919. (In fact, the Spanish
influenza baby boom cohort of 1920 is the second largest birth cohort ever
registered in Norway, only exceeded by the Second World War baby boom
cohort of 1946).
Spontaneous abortions and stillbirths in 1918
We have argued that Spanish influenza seems to have had a negative
effect on fertility in 1919. Rising rates of spontaneous abortion and still-
birth due to the Spanish influenza probably had a negative effect on ferti-
lity in 1918. There was actually a significant increase from 1917 to 1918
in both the spontaneous abortion rate, defined as spontaneous abortions
per 1,000 women who were 5-6 months pregnant, and the stillbirth rate,
defined as stillbirths per 1,000 women who were 7-9 months pregnant.
Based on data from DCM (1918-1923) and SSB (1920a, 1920b, 1921a,
1921b, 1923a, and 1923b), it is estimated that 45 excess spontaneous abor-
tions or 4.0 per 1,000 women 5-6 months pregnant, and 335 excess still-
births or 22.1 per 1,000 women 7-9 months pregnant, were due to the
THE SPANISH INFLUENZA PANDEMIC OF 1918 245
Spanish influenza. However, these calculations may be understatements,
especially those for spontaneous abortions that may have been systemati-
cally underreported. Also, abortions that occurred between conception and
the fifth month of pregnancy are omitted for lack of official statistics.
Deaths of pregnant women in 1918
Rising death rates from Spanish influenza among pregnant women
are also thought to have negatively affected fertility levels in 1918. At the
peak of the pandemic in November, maternal mortality, here defined as
deaths of pregnant women, occurred at a rate that was five times as high as
that in the average November month in the previous three years
(see Mamelund, 2003a). Based on data from DCM (1918-1923), it is esti-
mated that 67 excess maternal deaths from all causes, or 18.3 deaths
per 1,000 pregnant women, were caused by the Spanish influenza. Based
on the findings of Harris (1919) and Bourne (1922), higher maternal mor-
tality rates should be expected for pregnancies in the last trimester. Unfor-
tunately, the official data on maternal mortality are not broken down by
month of pregnancy, and as such, this issue remains open to speculation.
Marriages dissolved by death in 1918
In a preliminary version of this article (Mamelund, 2003a), it was es-
timated that an excess of 3,528 marriages, or 7.7 per 1,000, were dissolved
by death in 1918 compared to the annual average in the period 1915-1917.
It was also found that newly-wed women in their most reproductive ages
ran the greatest risk of losing their husbands to Spanish influenza. Here it
is assumed that wives who were not pregnant at the point of becoming
widows in the latter half of 1918 would not give birth to a baby in 1919.
IV. Multivariate analysis
To better examine the relationship between the assumed increase in
influenza and pneumonia morbidity and in mortality from all causes in
1918 on the one hand, and a decline in conceptions in 1918 on the other
hand, in comparison with the 1915-1917 norm (models 1-4), we use ordi-
nary least squares regressions. An analysis of the possibility that the as-
sumed rebound in conceptions in 1919 can be explained by the decline in
conceptions in 1918 and by the increase in morbidity and in mortality dur-
ing the second half of 1918, again when compared to the norm (models 5-
7), is also performed using the same type of regression analysis.
246 S.E. MAMELUND
TABLE 4.– DEFINITIONS AND DESCRIPTIVE STATISTICS FOR THE DEPENDENT AND INDEPENDENT VARIABLES
IN
37 URBAN AND RURAL PARTS OF NORWEGIAN COUNTIES
Model Variable Dependent variables Min. Max. Mean St. dev.
1 1 Largest monthly difference from the norm of the conception rate between July
and September 1918 0.26 – 28.95 – 7.95 5.41
2 2 Difference from the norm in the month with the lowest conception rate
between July and September 1918 1.05 – 28.95 – 6.62 5.85
3 3 Largest monthly difference from the norm of the conception rate between
October and December 1918 – 2.04 – 29.63 – 8.43 5.12
4 4 Difference from the norm in the month with the lowest conception rate
between October and December 1918 – 0.70 – 29.63 – 7.46 4.99
5-7 5 Difference in the conception rate for 1919 from the average for 1915-1917 – 50.81 72.90 23.51 25.64
Model Variable Independent variables Min. Max. Mean St. dev.
1 1 Difference from the norm of the conception rate between April and June 1918 19.54 – 21.03 – 7.84 8.03
2 Difference from the norm of the death rate in the month with the largest diffe-
rence from the norm of the conception rate between July and September 1918 – 7.58 11.73 1.54 4.16
3 Difference from the norm of the morbidity rate in the month with the largest
difference from the norm of the conception rate between July and
September 1918 22.49 980.99 239.90 247.13
2 1 Difference from the norm of the conception rate between April and June 1918 19.54 – 21.03 – 7.84 8.03
4 Difference from the norm of the death rate in the month with the lowest con-
ception rate between July and September 1918 – 7.58 34.33 3.75 8.33
5 Difference from the norm of the morbidity rate in the month with the lowest
conception rate between July and September 1918 18.72 1145.96 265.01 284.27
3 6 Difference from the norm of the conception rate between July and
September 1918 10.13 – 34.44 – 14.74 9.70
7 Difference from the norm of the death rate in the month with the largest diffe-
rence from the norm of the conception rate between October and
December 1918 – 3.81 87.85 18.14 18.06
THE SPANISH INFLUENZA PANDEMIC OF 1918 247
Model Variable Independent variables Min. Max. Mean St. dev.
8 Difference from the norm of the morbidity rate in the month with the largest
difference from the norm of the conception rate between October and
December 1918 65.01 1980.69 403.59 407.02
4 6 Difference from the norm of the conception rate between July and
September 1918 10.13 – 34.44 – 14.74 9.70
9 Difference from the norm of the death rate in the month with the lowest con-
ception rate between October and December 1918 – 3.55 87.85 15.95 17.24
10 Difference from the norm of the morbidity rate in the month with the lowest
conception rate between October and December 1918 52.36 1504.74 351.97 330.52
5-7 11 Difference from the norm of the conception rate between April and
December 1918 – 2.52 – 73.93 – 39.34 14.50
12 Difference from the norm of the death rate between July and December 1918 33.49 126.32 58.62 18.04
13 Difference from the norm of the morbidity rate between July and
December 1918 898.83 4571.00 2120.95 1027.48
14 Difference from the norm of the stillbirth rate between July and
December 1918 – 5.53 3.95 0.43 1.59
Note: The norm consists of the average value of the variable for the corresponding month(s) of the period 1915-1917.
248 S.-E. MAMELUND
1. Dependent and independent variables
Four dependent variables (labeled 1 to 4) are used to test the hypo-
thesis of a decline in the conception rate from the 1915-17 norm associ-
ated with Spanish influenza in 1918, while variable 5 is used to test the
hypothesis of a make-up of conceptions in 1919. Variables 1 and 2
measure the decline in conceptions during the summer of 1918, and
variables 3 and 4 the decline during the autumn. The decline is measured
in two different ways. Variables 1 and 3 use the difference between the
conception rate in the month with the largest decline and that in the cor-
responding month during the normal years of 1915-1917. Variables 2 and
4 use that difference for the month with the lowest observed level of con-
ception in 1918. Dependent variable 5 is the difference in the conception
rates from the norm of 1915-17 during the period from January through
December 1919. The definitions and descriptive statistics for the five de-
pendent variables and the fourteen independent variables included in the
seven models can be found in Table 4.
The denominators in the monthly conception, morbidity, death and
stillbirth rates were adjusted for length of the month (see Cassel, 2001).
The dependent and independent variables were not standardized, since
some of the data needed to calculate standardized rates (age and sex distri-
bution) exist only for the census of 1920 (and not for the years 1915-1919).
It appears, however, that the regional differences in the age distribution of
those in the 20-40 year age bracket, who had the highest morbidity, morta-
lity and fertility rates, were too small to bias the computation. In order to
account for spatial differences in underreporting of influenza and pneumo-
nia cases, weights were applied to all observations used in the analysis. The
weights were computed by multiplying the number of doctors per km
2
, the
percentage of population with health insurance, the number of physicians
per 1,000 of present population, and the number of hospitals per 1,000 of
present population (with average 3.5 and standard deviation 6.5).
Table 5 presents correlation coefficients between the independent
variables in the models. The correlations between the independent varia-
bles included in most models are relatively low, and should therefore
cause no multicollinearity problem. The correlations between variables in-
dexing mortality and morbidity differences from the norm (variables 7 and
8 in model 3, and variables 9 and 10 in model 4) are high and may cause a
multicollinearity problem. If so, although the parameter estimates will be
correct, they will be unstable and have excessive standard errors. It may
also be impossible to assess the independent effect of each of the two cor-
related covariates on the dependent variable. The issue of multicollinearity
is taken up again when discussing the results from models 3 and 4.
THE SPANISH INFLUENZA PANDEMIC OF 1918 249
TABLE 5.– CORRELATIONS OF INDEPENDENT VARIABLES
Model 1 Model 2 Model 3 Model 4 Model 5
No. 1 2 3 No. 1 4 5 No. 6 7 8 No. 6 9 10 No. 11 12 13 14
1 1.0 1 1.0 6 1.0 6 1.0 11 1.0
2 – 0.2 1.0 4 – 0.1 1.0 7 – 0.4 1.0 9 – 0.3 1.0 12 – 0.3 1.0
3 – 0.3 0.3 1.0 5 – 0.1 0.4 1.0 8 – 0.3 0.7 1.0 10 – 0.2 0.8 1.0 13 – 0.2 0.4 1.0
14 0.0 0.0 0.1 1.0
Note: See Table 4 for the definition of the variables.
250 S.-E. MAMELUND
The 1918 summer wave of Spanish influenza
It appears necessary to include the assumption that a large previous
decline in the conception rate in a region in early 1918 affected the con-
ception rate in a negative direction in that region later in 1918. When con-
trolling for the decline in the conception rate in the second quarter of 1918
from the 1915-1917 norm, the increase over the 1915-1917 norm in mor-
bidity and mortality during the summer wave has a negative effect on the
conception rate in the same period (Table 6, models 1 and 2). However,
only the morbidity rate contributes significantly to explaining the decline
in the conception rate compared to the norm, albeit in a very weak manner.
An increase of one influenza and pneumonia case in a region leads to a
later decline of only 0.006–0.007 conceptions.
This result seems to confirm the finding in the descriptive analysis
that morbidity is more important than mortality in explaining the down-
turn in the conception rate during the summer of 1918. This is not surpris-
ing since the increase in morbidity was much larger than the increase in
mortality in this period. The decline in the conception rate in the second
quarter of 1918 compared to the 1915-1917 norm has a strong negative ef-
fect on the decline in the conception rate in the third quarter of 1918.
When there is a decline of one conception in a region in the second quarter
of 1918 compared to the norm, it leads to a decline of only 0.51–0.60 con-
ceptions in that region in the third quarter, again compared to the norm.
The result implies that the population living in regions in which large de-
clines in the conception rates were observed in the second quarter of 1918
experienced a smaller decline in the conception rate in the third quarter of
1918, and vice versa.
The 1918 autumn wave of Spanish influenza
The results of the regression models for the effects of the increase in
morbidity and mortality during the autumn wave on the one hand, and the
decline in conceptions during the summer wave on the other hand are
striking, and surprising with respect to the decline in conceptions during
the autumn wave (Table 7, models 3 and 4). First, the size of the downturn
in the conception rate in a region during the summer wave has a negative
effect on the downturn in the conception rate in that region during the
autumn wave, as expected. However, the effect here is weaker than the
corresponding relationship found in models 1 and 2 (0.19–0.22 concep-
tions per 1,000 compared to 0.51–0.60), implying that, as a consequence
of a previous decline in a region, the potential for further decline in con-
ceptions in that region is decreasing. Second, but not surprisingly, the
effect of mortality is highly significant and strongly negative on the down-
turn in conceptions during the autumn wave. An increase in the death rate
by one death in a region during the autumn wave gives a reduction in the
conception rate in that region of 0.46–0.49 conceptions, all other factors
THE SPANISH INFLUENZA PANDEMIC OF 1918 251
TABLE 6.– RESULTS OF LINEAR REGRESSION ANALYSES OF THE LARGEST MONTHLY DIFFERENCE OF THE CONCEPTION RATE
OVER
THE NORM
(a)
(MODEL 1) AND OF THE DIFFERENCE OF THE LOWEST MONTHLY CONCEPTION RATE OVER THE NORM (MODEL 2)
BETWEEN JULY AND SEPTEMBER 1918
No.
(b)
Independent variables
Model 1 Model 2
Coefficient t-stat Coefficient t-stat
1 Difference from the norm of the conception rate between April and June 1918 – 0.5978 – 4.40*** – 0.5119 – 4.10***
2 or 4 Difference from the norm of the death rate in one of the months between July
and September 1918 – 0.2200 – 0.99 – 0.0275 – 0.17
3 or 5 Difference from the norm of the morbidity rate in one of the months between
July and September 1918 – 0.0063 – 1.79* – 0.0071 – 2.07**
Constant – 10.7598 – 6.09*** – 10.7332 – 5.95***
N (rural and urban parts of counties) 37 37
Adjusted R
2
0.3761 0.3507
(a)
The norm consists of the average value of the variable for the corresponding month(s) of the period 1915-1917.
(b)
The number for each variable corresponds to the number given for the variables in Table 4.
* p < 0.10, ** p < 0.05, *** p < 0.01.
TABLE 7.– RESULTS OF LINEAR REGRESSION ANALYSES OF THE LARGEST MONTHLY DIFFERENCE OF THE CONCEPTION RATE
OVER
THE NORM
(a)
(MODEL 3) AND OF THE DIFFERENCE OF THE LOWEST MONTHLY CONCEPTION RATE OVER THE NORM (MODEL 4)
BETWEEN OCTOBER AND DECEMBER 1918
No.
(b)
Independent variables
Model 3 Model 4
Coefficient t-stat Coefficient t-stat
6 Difference from the norm of the conception rate between July
and September 1918 – 0.2238 – 3.70*** – 0.1941 – 4.27***
7 or 9 Difference from the norm of the death rate in one of the months between
October and December 1918 – 0.4581 – 7.13*** – 0.4918 – 9.54***
8 or 10 Difference from the norm of the morbidity rate in one of the months between
October and December 1918 0.0095 5.11*** 0.0158 5.66***
Constant – 8.0227 – 5.80 *** – 8.9587 – 8.29***
N (rural and urban parts of counties) 37 37
Adjusted R
2
0.5860 0.7272
(a)
The norm consists of the average value of the variable for the corresponding month(s) of the period 1915-1917.
(b)
The number for each variable corresponds to the number given for the variables in Table 4.
* p < 0.10, ** p < 0.05, *** p < 0.01.
252 S.-E. MAMELUND
being equal. Third, and most surprisingly, the effect of morbidity goes in
the opposite direction to what was expected. According to models 3 and 4,
an increase in the morbidity rate by one influenza and pneumonia case
gives an increase in the conception rate of 0.0095–0.0158 conceptions.
This effect is not large, but it is highly significant.
In the section describing the independent variables, it was argued
that the relatively high correlations between variables 7 and 8 in model 3
and variables 9 and 10 in model 4 may cause a multicollinearity problem.
However, when model 3 is applied with and without, for instance, the in-
crease in the morbidity rate, but with controls for previous decline in con-
ceptions and the increase in the death rate, a substantial difference in the
explanatory power occurs (uncorrected R
2
increases by 30 percentage
points to 62% when including the morbidity rate). In addition, the sign of
the coefficients already included in the model does not change and the sta-
tistical levels of significance increase. It is worth noting that the two esti-
mated coefficients were significant at the 1% level when the morbidity rate
was not included in the model. The morbidity rate and the death rate to-
gether explain more than half of the variance in the conception rate—52%
in model 3 and 58% in model 4— and independently each has a specific
effect. More or less the same story goes for model 4. This reduces the pos-
sibility of a multicollinearity problem in models 3 and 4.
The pandemic subsides:
A rebound of conceptions in 1919
If one does not control for the increase above the norm of morbidity,
mortality, and stillbirths in the second half of 1918, a decline in the con-
ception rate by one conception in a region in the last nine months of 1918
leads to an increase in the conception rate in that region of two concep-
tions in 1919 (Table 8, model 5). The positive coefficient implies that the
larger the downturn in the conception rate was in a region in 1918, the
larger was the rebound in that region in 1919 and vice versa.
In model 6 (Table 8), however, two possible confounding factors are
added. One is morbidity, in order to control for the possibility that the
disease affected the fecundity of survivors negatively. The second is mor-
tality, to control for the fact that the chances of those widowed in 1918
realizing a pregnancy in 1919 were low in view of the required year of
mourning. In this case, the independent effect of a decrease from the norm
by one conception in the last nine months of 1918 leads to an increase
over the norm of 1.7 conceptions in 1919. This effect is lower than in
model 5 by 0.3 conceptions. At the same time, the explanatory power (R
2
)
of model 6 increases by 16 percentage points to 73%. Surprisingly, the in-
crease in morbidity and mortality in 1918 did not negatively affect the size
of the rebound in the conception rate in 1919 as expected. This does not
necessarily mean that the assumed effect was not present. More likely,
other mechanisms may have worked simultaneously, positively affecting
THE SPANISH INFLUENZA PANDEMIC OF 1918 253
TABLE 8.– RESULTS OF LINEAR REGRESSION ANALYSES FOR THE DIFFERENCE IN THE CONCEPTION RATE FOR 1919
FROM THE AVERAGE FOR 1915-1917
No.
(b)
Independent variables
Model 5 Model 6 Model 7
Coefficient t-stat Coefficient t-stat Coefficient t-stat
11
Difference from the norm
(a)
of the conception rate between April
and December 1918 1.9937 6.98*** 1.6863 6.07*** 1.4613 5.59***
12 Difference from the norm of the death rate between July
and December 1918 0.6868 2.12** 0.3749 1.21
13 Difference from the norm of the morbidity rate between July
and December 1918 0.0085 1.75* 0.0139 2.94***
14 Difference from the norm of the stillbirth rate between July
and December 1918 – 3.9104 – 2.97***
Constant 104.3500 7.87*** 35.1778 1.98* 32.3270 2.02*
N (rural and urban parts of counties) 37 37 37
Adjusted R
2
0.5699 0.7323 0.7834
(a)
The norm consists of the average value of the variable for the corresponding month(s) of the period 1915-1917.
(b)
The number for each variable corresponds to the number given for the variables in Table 4.
* p < 0.10, ** p < 0.05, *** p < 0.01.
254 S.-E. MAMELUND
the dependent variable, and thus cancelling out the expected negative
effect of morbidity and mortality on the assumed rebound of conceptions
in 1919.
One reason why the effect of the death rate in 1918 (as a proxy for
marital dissolution due to fatalities from Spanish influenza) on the re-
bound of conceptions of 1919 was positive may be that a larger proportion
of the conceptions than usual took place out of wedlock. Indeed, the op-
portunity for remarriage in 1919 was limited due to the year of mourning
imposed by the Norwegian marriage law. Data for the nation as a whole
support this hypothesis: illegitimate births per 1,000 people increased by
20% in 1920 over the normal period of 1915-1917, and the legitimate
births only by 8% (see Table 2). Remarriage rates for both widowers and
widows, however, declined from 1918 to 1919, and increased from 1919 to
1920, as expected (see Table 9). This suggests that the original assumption
of no sexual intercourse prior to remarriage was too strict. However, when
running model 6 with only legitimate conceptions in both the dependent
and the independent variables for the 37 areas (estimates not shown here),
the death rate in 1918 still has a positive effect on the rebound, casting
doubt on the explanation. Anyway, the result also begs the question
whether there were any illegitimate-legitimate differences in the downturn
in birth rates in 1919. When using 1915-1917 to exemplify a normal pe-
riod, no significant differences were found, however, as the decline was
8.0 and 6.0% in legitimate and illegitimate birth rates respectively. This
result is not surprising, as there is no reason to believe that the psycho-
logical or biological aspects of Spanish influenza would affect the fertility
of married persons differently from that of the unmarried.
TABLE 9.– REMARRIAGES AND REMARRIAGE RATES (PER 1,000) FOR WIDOWERS
AND
WIDOWS, NORWAY 1913-1923
Period
Remarriage
of widowers
Remarriage
per 1,000 widowers
Remarriage
of widows
Remarriage
per 1,000 widows
1913 1,343 30.2 593 6.1
1914 1,267 28.1 567 5.7
1915 1,236 27.0 549 5.5
1916 1,323 28.4 598 5.9
1917 1,389 29.4 655 6.4
1918 1,436 29.9 653 6.3
1919 1,164 23.9 504 4.8
1920 1,562 31.6 721 6.8
1921 1,449 29.2 682 6.4
1922 1,332 26.8 670 6.3
1923 1,286 25.8 646 6.0
Note: The widowed population at risk between the censuses of 1910, 1920 and 1930 is obtained by linear
interpolation.
Source: Regional population database (RD) of Statistics Norway, SSB, 1995.
THE SPANISH INFLUENZA PANDEMIC OF 1918 255
A second explanation of the positive effect of mortality in 1918 on
the rebound in conceptions in 1919 is that those married couples living in
regions where death rates and marriage dissolution rates due to Spanish
influenza deaths were high, who did not lose their spouse to the disease,
caught up relatively quickly with more conceptions than similar couples
living in regions with low death rates and low marriage dissolution rates.
The “underlying” explanation might be twofold. First, loss of a child due
to the flu may have intensified the desire to have another child to replace
the loss. Second, high community mortality in general may have fostered a
desire to “replace” the dead.
In model 7 (Table 8), stillbirth rates in the second half of 1918 are
included as a new independent variable to control for the assumed nega-
tive effect of high stillbirth rates in 1918 on the rebound of the conception
rates in 1919. The coefficient is negative as expected, yet significant: an
increase by one stillbirth in a region in 1918 reduces the rebound potential
in that region in 1919 by 3.9 conceptions. This is explained by an increase
in spontaneous abortions and stillbirths, making the women who experi-
enced this event temporarily infecund. The independent effect of a decline
in the conception rate in 1918 by one conception, all other factors remain-
ing the same, in model 7 is now an increase of 1.5 conceptions in 1919. In
total, the best model (model 7) explains nearly 80% of the variance in the
rebound stage of the pandemic. Note also that the effect of the death rates
in 1918 is no longer significant on the rebound in model 7, though the sign
is still positive. The reason for a decline of one conception in 1918 leading
to an increase of more than one conception made up in 1919 may be due to
two inseparable factors: first, but not necessarily the more important, joy
that the Spanish influenza and the First World War were over; second, a
desire to replace the dead, especially small children.
Conclusion
This article has tested the hypothesis that the 1918 Spanish influenza
pandemic, which infected 1.2 million Norwegians (45% of the population)
and killed 15,000 (5.7 deaths per 1,000 population), with most victims in
the fertile ages of 20-40, caused the baby boom of 1920 in Norway. The
descriptive analysis showed that the observed time profile of Spanish in-
fluenza fits well with the demographic changes that occur commonly dur-
ing and after a general mortality crisis. The course of the 1918 infl uenza
and the ensuing effects on fertility and nuptiality also correspond fairly
well to the experience reported for other countries (Höijer, 1959; Pool,
1973; Rice, 1983; Underwood, 1983; Mills, 1986; and Johnson, 2002).
However, previous studies have been univariate and descriptive and thus
could not investigate whether Spanish influenza morbidity had an inde-
pendent effect on fertility net of the effect of mortality, or whether mor-
256 S.-E. MAMELUND
bidity had a different effect on fertility than mortality had during the
summer and autumn waves of 1918.
The study applied multivariate analysis to shed light on these ques-
tions, and used detailed and reliable cross-sectional demographic data
available for Norway that have not been analysed before. The data on mor-
bidity are particularly valuable for the analysis. Such data are very diffi-
cult to obtain for other countries and when available, are often of poor
quality. Another advantage of Norway is that it remained neutral through-
out the First World War. Registration of population data including vital
statistics was therefore on the whole relatively undisturbed by the war.
Moreover, any assumed effect of the Spanish influenza on fertility and
nuptiality was not affected by the war.
There are two key hypotheses concerning why Spanish influenza is
believed to have caused a decline of conceptions in 1918. First, the disease
probably restrained the ability and willingness to have sexual intercourse
for some weeks during each of the waves. Second, those who lost their
spouses to the Spanish influenza were bound by the legal and moral cus-
toms of the time to desist from remarriage at least for a year. The regres-
sion analysis showed that the higher the increase in influenza morbidity
was in a region during the relatively mild summer outbreak in 1918, the
larger the decline in the conception rate was in that region. The increase in
mortality also had a negative effect on conceptions during the summer out-
break, but the effect was not statistically significant. Likewise, it was
found that the higher the increase in Spanish influenza mortality was in a
region during the highly lethal autumn outbreak in 1918, the larger the de-
cline in the conception rate was in that region. As mortality increased
much more during the autumn wave than during the summer wave, the ef-
fect of mortality on conceptions was also strongest during the former. It
was also found that mortality had a stronger effect than morbidity on con-
ceptions, probably because those who came down with influenza but did
not die had a shorter period of sexual abstinence than those who lost their
spouse. Surprisingly, the effect of the increase in morbidity on the concep-
tion rates during the autumn outbreak—net of the effect of mortality —
was positive and highly significant.
The regression analysis also showed that the size of the rebound of
conceptions in 1919 surpassed the compensating rebound potential
—created by the decline of conception in 1918— by 50%. This means that
a decline of one conception in a given region in 1918 resulted in 1.5 con-
ceptions made up in that region in 1919. The reason for this is probably
that, in addition to making-up of the previous decline in conceptions, ad-
ditional conceptions were realized to replace the dead, especially lost chil-
dren. These resulted in births during 1920. The conclusion of this article is
thus that the Spanish influenza pandemic of 1918 created the baby boom in
Norway in 1920.
THE SPANISH INFLUENZA PANDEMIC OF 1918 257
One important speculative issue that arises from the article, but
which is not addressed in the analysis, is why the baby boom of 1920 ap-
peared to be modest in the belligerent countries. The potential for a re-
bound was certainly large given the great fertility decline during 1914-
1918 that was associated with the First World War, to which was added a
possible decline in fertility associated with the onset of Spanish influenza
in 1918. The reason for this is perhaps that both the war and Spanish influ-
enza took their heaviest toll among males aged between 20 and 40 years,
thereby lowering the chances of marriage for spinsters and of remarriage
for widows. This in turn may explain why the potential making-up of
births seems to have been far from fully exploited.
Acknowledgements: Comments to this article from Nico Keilman, Øystein Kravdal,
Alberto Palloni, Susan De Vos and two anonymous reviewers are gratefully acknow-
ledged. Thanks are also due to Halvard Skiri, Kirsten Dybendal and Britt Elin Bråten
who assisted in finding sources of historical data in Statistics Norway. I am grateful
to the Center for Demography and Ecology at the University of Wisconsin-Madison
who provided excellent working conditions while I wrote a preliminary version of
this paper in the autumn of 2001. Last but not least, thanks to Jennifer Chisholm-
Høibråten for proofreading. The article is part of the research project Spanish influ-
enza and beyond: The case of Norway, financially supported by the Norwegian Re-
search Council and Department of Economics, University of Oslo.
REFERENCES
BIRABEN J.-N., 1973, “Aspects médicaux et biologiques de la démographie historique”, in Inter-
national Population Conference, Liège, IUSSP, 3, pp. 9-22.
BLAND P.B., 1919, “Influenza in its relation to pregnancy and labor”, American Journal of Ob-
stetrics, 79, pp. 184-197.
BOURNE A.W., 1922, “Influenza: Pregnancy, labour, the puerperium and diseases of women”,
in F.G. Crookshank (ed.), Influenza: Essays by Several Authors, London, William
Heinemann (Medical Books) Ltd, Ch. XIV.
BRITTEN R.H., 1932, “The incidence of epidemic influenza, 1918-19”, Public Health Reports, 47(6),
pp. 304-339.
BRUNBORG H., MAMELUND S.-E., 1994, Kohort- og periodefruktbarhet i Norge 1820-1993 (Co-
hort and Period Fertility in Norway 1820-1993), Rapporter (reports) 94/27, Oslo-
Kongsvinger, Statistisk sentralbyrå [English abstract].
CASSEL P.G., 2001, “Changing seasonality of births in Sweden 1900-1999”, Paper presented at
the 14th Nordic Demographic Symposium, Tjøme 3-5 May 2001, Norway.
CHESNAIS J.-C., 1992, The Demographic Transition. Stages, Patterns, and Economic Implica-
tions. A Longitudinal Study of Sixty-Seven Countries Covering the Period 1720-1984,
Oxford, Clarendon Press.
COLLINS S.D., 1931, “Age and sex incidence of influenza and pneumonia morbidity and mortality
in the epidemic of 1928-29 with comparative data for the epidemic of 1918-19”, Public
Health Reports, 46(33), pp. 1909-1937.
DCM (D
ET CIVILE MEDISINALVESEN), 1918-1924, Sundhetstilstanden og medisinalforholdene
1915-1920, Norwegian Official Statistics, VI.133, VI.186, VII.3, VII.58, VII.108,
VII.138.
HANSSEN O., 1923, Undersøkelser over influenzaens optræden specielt i Bergen 1918-1922,
Arbeider fra Den medicinske Avdeling av Haukeland sykehus, Skrifter utgit ved Klaus
Hanssens Fond. Nr. III, Bergen, A.S, John Griegs Boktrykkeri og N. Nilssen & søn.
258 S.-E. MAMELUND
HARRIS J.W., 1919, “Influenza occurring in pregnant women: a statistical study of 1350 cases”,
JAMA, 72, pp. 978-980.
H
ENRY L., 1966, “Pertubations de la nuptialité résultant de la guerre 1914-1918”, Population,
21(2), pp. 273-332.
HÖIJER E., 1959, Sveriges befolkningsutveckling genom tiderna, Stockholm.
JOHNSON N.P.A.S., 2002, Aspects of the Historical Geography of the 1918-19 Influenza Pande-
mic in Britain, PhD thesis, Cambridge, University of Cambridge.
JOHNSON N.P.A.S., MUELLER J., 2002, “Updating the accounts: Global mortality of the 1918-1920
‘Spanish influenza’ pandemic”, Bulletin of the History of Medicine, 76, pp. 105-115.
JUHASZ L., 1971, “Demografiske kriser”, Heimen, XW, pp. 397-417.
KRISTIANIA SUNDHETSKOMMISION, 1919, Beretning fra Kristiania sundhetskommision og Kris-
tiania kommunale sykehus for året 1918, Kristiania.
L
AIDLAW P.P., 1935, “Epidemic influenza: A virus disease”, Lancet, I, pp. 1118-1124.
L
EE R., 1989, “Short-term variation: vital rates, prices and weather”, in E.A. Wrigley, R.S.
Schofield (eds), The Population History of England 1541-1871, A Reconstruction, Cam-
bridge, Cambridge University Press, pp. 356-401.
LEE R., 1990, “The demographic response to economic crisis in historical and contemporary
populations”, Population Bulletin of the United Nations, 29, pp. 1-15.
LERIDON H., 1977, Human Fertility: The Basic Components, Chicago, University of Chicago
Press.
LIVI BACCI M., 2000, “Mortality crisis in a historical perspective: The European experience”, in
G.A. Cornia, R. Paniccià (eds), The Mortality Crisis in Transnational Economies,
Oxford, Oxford University Press.
MAMELUND S.-E., 1998, Spanskesyken i Norge 1918-1920: Diffusjon og demografiske kon-
sekvenser (unpublished Master-thesis), Hovedoppgave i Samfunnsgeografi høsten 1998,
Institutt for Sosiologi og Samfunnsgeografi, Oslo, Universitetet i Oslo.
MAMELUND S.-E., 2003a, “Can the Spanish influenza pandemic of 1918 explain the baby-boom
of 1920 in neutral Norway?”, Memorandum No. 01, Department of Economics, Univer-
sity of Oslo (see pdf format at www.oekonomi.uio.no/memo/).
M
AMELUND S.-E., 2003b, “Effects of the Spanish influenza pandemic of 1918-19 on later life
mortality of Norwegian cohorts born about 1900”, Memorandum No. 29, Department of
Economics, University of Oslo (see pdf format at www.oekonomi.uio.no/memo/).
MAMELUND S.-E., BORGAN J.K., 1996, Kohort- og periodedødeligheten i Norge 1846-1994
[Cohort and period mortality in Norway 1846-1994], Rapporter (reports) 96/9, Oslo-
Kongsvinger, Statistisk sentralbyrå (English abstract).
MAMELUND S.-E., BRUNBORG H., NOACK T., 1997, Skilsmisser i Norge 1886-1995 for kalenderår
og ekteskapskohorter [Divorce in Norway 1886-1995 by calendar year and marriage
cohort], Rapporter 97/19, Oslo-Kongsvinger, Statistisk sentralbyrå (English abstract).
MENKEN J., TRUSSELL J., WATKINS S., 1981, “The nutrition–fertility link: An evaluation of the
evidence”, Journal of Interdisciplinary History, 11(3), pp. 425-441.
MILLS I.D., 1986, “1918-1919 influenza pandemic: The Indian experience”, The Indian Eco-
nomic and Social History Review, 23, pp. 1-40.
PALLONI A., 1988, “On the role of crisis in historical perspective: An exchange”, Population and
Development Review, 14(1), pp. 145-164
POOL D.I., 1973, “The effects of the 1918 pandemic of influenza on the Maori population of
New Zealand”, Bulletin of the History of Medicine, 47(3), pp. 273-281.
PRESTON S., 1978, The Effects of Infant and Child Mortality on Fertility, New York, Academic
Press.
RICE G., 1983, “Maori mortality in the 1918 influenza epidemic”, New Zealand Population
Review, 9(1), pp. 44-61.
SSB (STATISTISK SENTRALBYRÅ), 1920a, 1920b, 1921a, 1921b, 1923a, 1923b, Folkemengdens
bevegelse 1915-1920, Norwegian Official Statistics VI.111, VI.163, VII.2, VII.31,
VII.72, VII.92. Kristiania.
SSB (STATISTISK SENTRALBYRÅ), 1926, Folkemengdens bevegelse 1911-1920. Hovedoversikt,
Norwegian Official Statistics VIII.6, Oslo.
SSB (STATISTISK SENTRALBYRÅ), 1995, Historisk Statistisk 1994, Norwegian Official Statistics
C.105, Oslo.
SYDENSTRICKER E., 1931, “The incidence of influenza among persons of different economic sta-
tus during the epidemic of 1918”, Public Health Reports, 46(4), pp. 154-170.
THE SPANISH INFLUENZA PANDEMIC OF 1918 259
UNDERWOOD J.H., 1983, “Effects of the 1918 influenza pandemic mortality experience on sub-
sequent fertility of the native population of Guam”, Micronesia, 19(1-2), pp. 1-9.
V
AUGHAN W.T., 1921, Influenza. An Epidemiological Study (Monograph Series, No. 1), Balti-
more, The American Journal of Hygiene.
WATKINS S.C., MENKEN J., 1985, “Famines in historical perspective”, Population and Develop-
ment Review, 11(4), pp. 647-675.
WINTER J.M., 1977, “Britain’s ‘Lost Generation’ of the First World War”, Population Studies,
31(3), pp. 449-466.
WRIGLEY E.A., SCHOFIELD R.S., 1989, The Population History of England 1541-1871. A Recon-
struction, Cambridge, Cambridge University Press.
260
MAMELUND Svenn-Erik.– Can the Spanish Influenza Pandemic of 1918 Explain the Baby
Boom of 1920 in Neutral Norway?
Two years after the First World War ended there was a surge in European birth rates, in-
cluding in Norway that had been a neutral country. This paper tests the hypothesis that it was in
fact the Spanish influenza that caused the Norwegian baby boom rather than the close of the war.
The paper uses multivariate regression analysis, while previous studies have been univariate and
largely descriptive. By using regional monthly data, the independent effect of the Spanish influ-
enza morbidity on fertility over the years 1918-1920, net of the effect of mortality, is estimated.
The fact that Norway was neutral was important in counter-balancing the influence of the war
on fertility and nuptiality. Furthermore, the Norwegian data utilized in the analysis are of supe-
rior quality in a European context in that registration of population data, including vital statis-
tics, continued normally in Norway undisturbed by the war.
MAMELUND Svenn-Erik.– La grippe espagnole de 1918 est-elle responsable du baby-
boom de 1920 en Norvège ? Le cas d’un pays neutre
Deux ans après la fin de la première guerre mondiale, les taux de natalité ont enregistré
une vive progression en Europe, y compris en Norvège, pays resté neutre lors du conflit. Cet
article tente de valider l’hypothèse selon laquelle la responsabilité du baby-boom en Norvège
revient plutôt à la grippe espagnole qu’à la fin de la guerre. Alors que les recherches antérieures
ont reposé sur des analyses univariées et essentiellement descriptives, cette étude s’appuie sur
des analyses multivariées. L’effet propre de la morbidité due à la grippe espagnole sur la fécon-
dité, en contrôlant l’effet de la mortalité, a été estimé pour la période 1918-1920 à partir de don-
nées mensuelles régionales. En raison de la neutralité de la Norvège, la guerre n’a pas interféré
avec l’effet de la grippe espagnole sur la fécondité et la nuptialité. De surcroît, les données dis-
ponibles sont parmi les plus fiables d’Europe, étant donné que l’enregistrement des données dé-
mographiques, y compris d’état civil, n’a pas été perturbé par le conflit.
M
AMELUND Svenn-Erik.– ¿La gripe española de 1918 es responsable del baby-boom de
1920 en Noruega? El caso de un país neutro
Dos años después del fin de la primera guerra mundial, las tasas de natalidad han
aumentado considerablemente en Europa, incluso en Noruega, país neutro durante la guerra.
Este articulo intenta verificar la hipótesis según la cual el baby-boom en Noruega fue provo-
cado mas bien por la gripe española que por el fin del conflicto. Contrariamente a los análisis
precedentes que se han basado en análisis univariados y esencialmente descriptivos, este es-
tudio se apoya en análisis multivariados. El efecto propio de la gripe española sobre la fecun-
didad, controlando el efecto de la mortalidad, ha sido estimado para el periodo 1918-1920 a
partir de datos regionales. Dada la neutralidad de Noruega, la guerra no ha interferido con el
efecto de la gripe sobre la nupcialidad y la mortalidad. Los datos utilizados figuran entre los
más fiables de Europa, ya que su registro no ha sido perturbado por el conflicto.
Svenn-Erik MAMELUND, Department of Economics, University of Oslo, P.O. Box 1095 Blindern,
0317 Oslo, Norway, Tel: + 47 22 85 51 26, Fax: + 47 22 85 50 35, e-mail: sem@econ.uio.no
