ArticlePDF Available

A socially neutral disease? Individual social class, household wealth and mortality from Spanish influenza in two socially contrasting parishes in Kristiania 1918-19

  • OsloMet - Oslo Metropolitan University

Abstract and Figures

The Spanish influenza pandemic of 1918-19 was one of the most devastating diseases in history, killing perhaps as many as 50-100 million people worldwide. Much of the literature since 1918 has favored the view that mortality from Spanish influenza was class neutral. This view has prevailed, even though several contemporary surveys showed that there indeed were clear differences between the classes in disease incidence and that case fatality rates from influenza and pneumonia also varied according to socioeconomic status. Furthermore, studies of more recent influenza epidemics have also shown that there can be clear class differentials in mortality in this type of illness--is there any reason to believe that Spanish influenza was different? This paper is the first study in which individual- and household-level data which are unique for the period are utilized to test the conservative hypothesis that Spanish influenza was a socially neutral disease with respect to mortality. Through the use of Cox regressions in an analysis of two socially contrasting parishes in the Norwegian capital city of Kristiania, it is shown that apartment size as an indicator of wealth of a household, in addition to social status of place of residence, were the only socioeconomic variables that had an independent and significant effect on mortality after controlling for age, sex and marital status.
Content may be subject to copyright.
Social Science & Medicine 62 (2006) 923–940
A socially neutral disease? Individual social class, household
wealth and mortality from Spanish influenza in two socially
contrasting parishes in Kristiania 1918–19
Svenn-Erik Mamelund
Department of Economics, University of Oslo, P.O. Box 1095 Blindern, 0317 Oslo, Norway
Available online 8 August 2005
The Spanish influenza pandemic of 1918–19 was one of the most devastating diseases in history, killing perhaps as
many as 50–100 million people worldwide. Much of the literature since 1918 has favored the view that mortality from
Spanish influenza was class neutral. This view has prevailed, even though several contemporary surveys showed that
there indeed were clear differences between the classes in disease incidence and that case fatality rates from influenza
and pneumonia also varied according to socioeconomic status. Furthermore, studies of more recent influenza epidemics
have also shown that there can be clear class differentials in mortality in this type of illness—is there any reason to
believe that Spanish influenza was different? This paper is the first study in which individual- and household-level data
which are unique for the period are utilized to test the conservative hypothesis that Spanish influenza was a socially
neutral disease with respect to mortality. Through the use of Cox regressions in an analysis of two socially contrasting
parishes in the Norwegian capital city of Kristiania, it is shown that apartment size as an indicator of wealth of a
household, in addition to social status of place of residence, were the only socioeconomic variables that had an
independent and significant effect on mortality after controlling for age, sex and marital status.
r2005 Elsevier Ltd. All rights reserved.
Keywords: Spanish influenza mortality 1918–19; Socioeconomic factors; Contextual effects; Norway; Event history analysis; Cox
proportional hazards
Socioeconomic differences between classes as a factor
in the incidence of disease, mortality and survival have
been documented for different time periods in all
countries for which data exists (Feinstein, 1993;Kunst
& Mackenbach, 1994). These kinds of differences have
been documented for a number of causes of illness and
death, including cardiovascular diseases (Vallin, Mesle
& Valkonen, 2001), several types of cancer (Kravdal,
2003), chronic obstructive pulmonary disease (Prescott
& Vestbo, 1999), and mortality associated with excessive
alcohol consumption (Ma
¨, Valkonen, & Martelin,
1997). There is also ample evidence that there were
distinct differences between the social classes with
respect to the main causes of death in historical
populations, particularly for tuberculosis and cholera
fatalities (e.g. Gjestland & Moen, 1988;Hansen, 1985).
This paper addresses the general question whether a
social gradient is also prevalent in mortality from
influenza in annual epidemics. In particular, it examines
the role of socioeconomic status, both of individuals,
households and neighborhoods, in explaining the
0277-9536/$ - see front matter r2005 Elsevier Ltd. All rights reserved.
Tel.: +47 22 85 51 26; fax:+47 22 85 50 35.
E-mail address:
variance in mortality associated with the Spanish
influenza pandemic which may have killed 50–100
million worldwide in 1918–19 (Johnson & Mueller,
2002). Indeed, socioeconomic differentials have also
been reported for influenza and pneumonia combined,
not only in the risk of dying (Kitagawa & Hauser, 1968,
1973;Regidor, Calle, Navarro, & Domı
´nguez, 2003;
Singh & Siahpush, 2001), but also in the actual risk of
contracting influenza (Dutton, 1988;Glezen, Paredes, &
Taber, 1980). Moreover, the socioeconomic differences
in mortality from influenza and pneumonia found in
these studies, which include the United States in 1950,
1960 and the period 1979–89, and Spain in 1996–97,
were among the greatest compared to any other cause of
death. However, only a very few studies have been
specifically directed towards explaining the social profile
of these two causes of death. When examining the data
from these relatively recent influenza epidemics, can
there be any reason to doubt that the lower socio-
economic classes had a higher mortality from influenza
and pneumonia than the higher classes during the
pandemic in 1918?
House-to-house surveys conducted in the United
States and Norway in 1918–19 showed that there were
marked social differences in both the incidence and
lethality from Spanish influenza (Britten, 1932;Collins,
1931;Hanssen, 1923;Sydenstricker, 1931;Vaughan,
1921), while a similar study of four cities in England
found no clear relationship between incidence and/or
case fatality rates and social status (Great Britain
Ministry of Health, 1920). Debate has continued in the
literature ever since 1918 over whether social status
played any role in mortality from the 1918–19 pandemic;
most studies contend that it was socially neutral.
Proponents of the ‘‘socially neutral’’ view claim that
Spanish influenza struck blindly and randomly because
the pandemic introduced a new virus that few, if any,
had the immunity to fight. They argue that Spanish
influenza differed to annual epidemics in which a large
part of the population has acquired immunity from
exposure to previous epidemics (Brainerd & Siegler,
2003;Crosby, 2003;Rice, 1988;Stevenson, 1921;
Tomkins, 1992;van Hartesveldt, 1992;Winter &
Robert, 1997). A second argument commonly used to
support this view is the fact that in 1918, the largest
relative increase in death rates all over the world was
among people between the age of 20 and 40 years as
opposed to the very young and the elderly as is normally
seen during annual influenza epidemics. However, many
previous studies did not carry out a careful statistical
investigation of the association between mortality and
socioeconomic status themselves. They tended instead to
rely upon anecdotal evidence from physicians of the
time, using this to support the argument that Spanish
influenza was a ‘‘classless’’ disease because the odds of
survivability seemed to favor the most robust and
previously healthy of those aged 20–40 years. The
argument was made that even kings and presidents were
laid low by influenza. A possible explanation for this
view having prevailed in the literature may be that too
little distinction has been made between the risk of being
infected by influenza on the one hand (‘‘everybody gets
it’’), and the risk of actually dying from influenza or
pneumonia. In the first instance, the risk of contracting
the disease may be only moderately associated with
socioeconomic status, while several studies have shown
that there was a strong connection between mortality
from the disease and socioeconomic status. On the other
hand, some scholars argue that like tuberculosis and
cholera, Spanish influenza claimed higher death rates
amongst the destitute and most poorly situated than
among the wealthy and privileged (Hersch, 1920, 1932;
Johnson, 2001;Mamelund, 2003b;McCracken &
Curson, 2003;Sydenstricker, 1931;Zylberman, 2003).
Supporters of this view admit that the virus itself may
have had certain attack properties that were indepen-
dent of social class. However, it would appear that there
indeed were clear social differences in a person’s chance
of surviving the disease.
Most of the previous studies on socioeconomic status
and mortality from the 1918–19 pandemic have been
univariate and descriptive in type. The contributions
from them therefore fail to demonstrate the independent
effects on Spanish influenza mortality of age, sex,
socioeconomic class, crowding, ethnicity, spatial diffu-
sion, climate, and other geographical variables. Recent
exceptions to this pattern in the nature of the studies are
the cross-sectional studies of the United States (Brainerd
& Siegler, 2003) and England and Wales (Johnson,
2001), who, respectively, found no and only weak
indications of class differentials in mortality. Studies of
Sydney, Australia (McCracken & Curson, 2003), and
Norway (Mamelund, 2003b) on the other hand could
report a significant social gradient in mortality. How-
ever, because the associations in these studies were not
estimated using individual-level data, no definitive
conclusion could be drawn as to whether or not there
were any causal links between these variables.
This paper is the first to apply Cox proportional
hazard models combined with ‘‘state of the art’’ data for
the period on the level of individuals, households, and
parish, to contest the conservative assumption that
Spanish influenza was a socially neutral disease with
respect to mortality. No earlier study on the subject has
ever included data on different levels of aggregation in
the same model to search for the causes of variation in
Spanish influenza mortality. The paper uses mortality
and census data from two intentionally selected socially
contrasting parishes in the Norwegian capital of
Kristiania (renamed Oslo in 1924), namely Frogner
and Grønland-Wexels. The nominal censuses for 1918
and 1919 used here allow a very close follow-up of
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940924
individuals from the start of the pandemic in the early
spring of 1918 through to the end of it in the winter of
1919. Finally, registration of deaths and the carrying out
of the censuses were on the whole undisturbed by the
First World War, as Norway was a neutral country.
Such data are seldom available for the belligerent
countries. Existing data from such countries are usually
unreliable because of gaps in registration and because it
is difficult to separate deaths from the pandemic from
direct or indirect deaths caused by the war.
There are two reasons for limiting the analysis to the
two parishes of Frogner and Grønland-Wexels, which
are comparable in size and tally together 41,000
individuals or 16 per cent of a total population in
Kristiania of 260,000 on 1 February 1918 (Wexels was
merged with Grønland in 1919). First, significant
differences in all-cause mortality as well as cause-specific
mortality have already been shown to exist between the
traditionally poor, high-mortality parishes to the east of
the city and the wealthy, low-mortality parishes to the
west since the 1880s. Hence, Grønland-Wexels (east) and
Frogner (west) both constitute typical examples worthy
of study (e.g. Arctander, 1928;Barstad, 1997;Gjestland
& Moen, 1988;Rognerud & Stensvold, 1998). The
Kristiania of 1918 was a divided city, with large
east–west differences with respect to income, education,
and employment. The east–west differences could also
be seen in the stature and weight of individuals as a
proxy of disease and nutritional history, standard of
housing, sanitation, hygiene, household crowding,
crime, the number of social security recipients, and
child welfare cases (e.g. Arctander, 1928;Geirsvold,
1917;Kjeldstadli, 1990;Kristiania Statistiske kontor,
1920;Schiøtz, 1920;Statistisk sentralbyra
˚, 1955).
The extreme social polarization between the east and
the west sides of the city began when the industrial elite
moved to their summer homes on the west side of the
city and started to live in them year round in the 1860s.
This was probably a reaction to increasing inner-city
pollution and the desire to live further away from the
shabby and overcrowded industrial working class
suburbs (Myhre, 1990). In 1918, most of the wealthy
bourgeois and middle class in Kristiania therefore lived
in the western parishes, while the relatively poor
working class constituted the majority in the eastern
parishes of the city. However, Grønland-Wexels had a
substantial middle-class, and Frogner was actually
relatively heterogeneous with a substantial working
class population of maid servants, private chauffeurs
and porters who lived in the homes of the wealthy (see
Table 2). In 1901, the only year for which income data
are available by parish relatively close to the study
period, the average income per person after deductions
for Frogner residents was more than six times that of
residents in Grønland (Kristiania Statistiske kontor,
1900, p. 143). The two well-defined and well-known
socially contrasting parishes of Frogner and Grønland-
Wexels are intentionally selected to ensure that there is
sufficient variance in socioeconomic status within the
parameters of the study to be able to document
socioeconomic differences in Spanish influenza mortal-
ity, if these exist, at the level of individuals, households
and/or the parish. A second reason for selecting only
two parishes for this study is simply because digitization
of the census data for the entire city of Kristiania with its
19 parishes would have been too costly and time-
Study population
The data on the study population for Frogner and
Grønland-Wexels are taken from the two nominal
censuses of Kristiania made on the night of 31 January
to 1 February 1918 and 1919 (Oslo City Archive). The
total number of observations is close to 47,000 (Table 1).
Only de jure residents are considered, but the study
population also includes individuals who were perma-
nently resident or working in hostels or institutions.
Examples of institutions included in the study are a
prison, a Red Cross nursing home, a boarding school for
deaf pupils, and some rest homes for the elderly. For
details on reliability of census data, completeness of
record linkage, and assumptions on exposure time for
people who moved frequently, in particular maid
servants, see a previous version of this paper (Mame-
lund, 2004).
Dependent variable
In the analysis, the exposure to mortality risk starts on
1 February 1918. Right censoring is caused by deaths
from causes other than those associated with Spanish
influenza, when an individual moves out of the parishes
in question, and at the cutoff date of 1 February 1919.
The nominal data on deaths are from the unpublished
report Anmeldte døde i Oslo 1918–1921 (Oslo City
Archive). Of the 608 all-cause deaths in Frogner and
Grønland-Wexels during the intercensal year of
1918–19, 250 deaths are here linked with the Spanish
influenza pandemic, of which 81 and 169 deaths
occurred among de jure residents in Frogner and
Grønland-Wexels, respectively (see distribution of the
deaths by independent variables in Table 4). Fatal cases
of the disease usually occurred when influenza was
followed by bacterial complications such as bronchop-
neumonia and lobar pneumonia, or by viral or
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 925
combined viral and bacterial pneumonia. Of the 250
deaths, two-thirds were caused by influenza and
pneumonia while the rest were deaths caused by other
frequently reported complications (stated as secondary
cause of death on the death certificates) following
influenza (stated as primary cause of death on the death
certificates). Examples of such complications included
for example emphysema, pleuritis, lung embolus, acute
diarrhea, tetanus, nephritis, and cardiac failures (e.g.
myocarditis or pericarditis), or other diseases that were
symptomatically difficult to distinguish from influenza
and/or pneumonia and that might be labeled ‘‘acute
catarrhs in the respiratory organs’’, acute bronchitis and
diphtheria, bronchial asthma, or chronic bronchitis. It
should be pointed out that the data used here are not
controlled for the fact that some of the 250 deaths
considered also would have occurred during a normal
epidemic influenza season (see Mamelund, 1998). The
information given on the death certificates, including
cause and date of death, name, age, sex, and occupation,
was very reliable and generally in accordance with
similar information given for persons identified in the
1918 census (Mamelund, 2004).
Independent variables
The data for the independent variables included in the
analysis are taken from the 1918 and 1919 censuses for
Frogner and Grønland-Wexels described above. De-
scriptive statistics appear in Table 2.
Individual-level variables
Age for each individual is defined as the person’s exact
age on 1 February 1919. In the analysis, 13 age
categories are considered, and have been chosen to
reflect the W-shaped age pattern of Spanish influenza
death rates observed all over the world (Great Britain
Ministry of Health, 1920). The categorical age groups
were chosen after a model with a continuous age
distribution of death risks was inspected. A relatively
large age interval, 5–24 years, was selected as a reference
for the age effects in the analysis. This was done for two
reasons: first, it was selected because the reference
category should not be biased because of too few deaths
included, and second, because the mortality for any
given 5-year age group within the 5–24 year interval did
not differ significantly from one another.
Five marital status groups are included in the analysis
to control for the assumed protective and selective effect
of marriage on mortality: never married, married,
widow/widower, separated, and divorced. Three indivi-
dual-level, occupation-based social classes are also
defined. The bourgeois include capitalists, estate owners,
shipping, large-scale retail, whole salesmen, chief execu-
tives, chief editors, clergy, high-ranking military officers,
professors, doctors, dentists, attorneys, architects, lead-
ing authors, actors, and artists, pharmacists, Supreme
Court judges, engineers, ambassadors, consuls, senior
government officials, Members of Parliament, and
directors of banking, finance, insurance and so on.
Examples of professions considered as belonging to the
middle class are teachers, nurses, clerical officers, police
inspectors and constables, customs officers, office-clerks
in the postal services, telegraph messengers, librarians,
port authorities, vergers, sorters, poor-relief assistants,
the self-employed in small-scale retail, craft, and
industry, those with a craft master certificate, and
finally, pensioners. The working class includes foremen
and workers in the cottage industries, industrial
factories, quarrying industries, shipbuilding, sawmills
and construction work, transportation, cleaning and
janitorial services, cashiers and shop assistants, seamen,
fishermen, porters, and household staff and servants.
Included in this category are also some of the poorest
individuals in society, namely private and national
social security recipients, a handful of prisoners, certain
foster children, the disabled, the mentally ill and the
physically handicapped. However, this heterogenic
Table 1
The study population in the parishes of Frogner and Grønland-Wexels combined from 1 February 1918 to 1 February 1919
Individuals followed
From To Cases Per cent
02.01.1918 02.01.1919 34,127 72.7
02.01.1918 or intercensal date of birth Date of death 608 1.3
02.01.1918 Date of moving out of parish 6087 13.0
Date of moving into parish 02.01.1919 5591 11.9
Date of intercensal birth 02.01.1919 559 1.1
Number of observations 46,972 100.0
Source:Oslo City Archive, Censuses of 1918 and 1919 for the parishes of Frogner and Grønland-Wexels, and Anmeldte døde i Oslo
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940926
group constitutes only a little more than one per cent of
the working class.
All individuals are assigned their own scores for social
class except for two special cases. Children under 18 years
of age who have not yet entered the labor force and who
do not have employment (including students and 20–22
year old conscripts with no stated occupation) are
assumed to belong to the same occupational social class
as the head of the household, typically their father.
Secondly, housewives are placed in the same social class
as their husbands. A retired person falls in the social class
of his or her former employment, if stated.
Table 2
Distribution for the independent variables
Independent variables Both parishes Frogner Grønland-Wexels
Count Per cent Count Per cent Count Per cent
Individual-level variabels
0–1 1314 2.8 486 2.2 828 3.3
2–4 1868 4.0 690 3.2 1178 4.7
5–24 16,679 35.5 7308 33.5 9371 37.3
25–29 5681 12.1 2885 13.2 2796 11.1
30–34 4091 8.7 2113 9.6 1978 7.9
35–39 3207 6.8 1691 7.7 1516 6.0
40–44 3119 6.6 1576 7.2 1543 6.1
45–49 2432 5.2 1213 5.5 1219 4.9
50–54 2192 4.7 1062 4.9 1130 4.5
55–59 1800 3.8 799 3.7 1001 4.0
60–69 2772 5.9 1184 5.4 1587 6.3
70–79 1406 3.0 646 3.0 760 3.0
80+ 411 0.9 191 0.9 220 0.9
Male 20,977 44.7 7874 36.0 13,103 52.1
Female 25,995 55.3 13,971 64.0 12,024 47.9
Marital status
Never married 30,221 64.3 14,722 67.4 15,499 61.7
Married 13,495 28.7 5781 26.5 7714 30.7
Widow/widower 2772 5.9 1158 5.3 1614 6.4
Separated 336 0.7 54 0.2 282 1.1
Divorced 148 0.3 130 0.6 18 0.1
Social class
Bourgeois 8937 19.0 7939 36.4 998 4.0
Middle class 10,595 22.6 6035 27.6 4560 18.1
Working class 25,749 54.8 6966 31.9 18,783 74.8
Occupation not stated 1691 3.6 905 4.1 786 3.1
Household-level variable
Size of apartment*
One room 9101 19.4 3328 15.2 5773 23.0
Two rooms 7205 15.3 672 3.1 6533 26.0
Three rooms 10,154 21.6 1529 7.0 8625 34.3
Four rooms 4654 10.0 2204 10.1 2450 9.8
Five rooms 4504 9.6 3334 15.3 1170 4.7
Six rooms 4875 10.4 4558 20.9 317 1.2
Seven rooms 2792 6.0 2663 12.2 129 0.5
Eight rooms+ 3687 7.9 3557 16.2 130 0.5
Number of observations 46,972 100.0 21,845 100.0 25,127 100.0
All variables are dummy variables which take the value 0 or 1.
*1 room ¼1 room, no kitchen; 2 rooms ¼1 room and kitchen; 3 rooms ¼2 rooms and kitchen, etc.
Source: Oslo City Archive, Censuses of 1918 and 1919 for the parishes of Frogner and Grønland-Wexels.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 927
The three occupation-based social classes are assumed
to encompass differentials in income and education
which in their turn may have differential effects on
health and mortality. Although a relatively large
heterogeneity may appear within each class with respect
to both income and/or education (e.g. the academic vs.
industrialist bourgeois elite), it may be seen that social
class as defined here is a reasonably good proxy for these
two indices of wealth. The income of an average white-
collar clerical officer before the First World War, for
example, was double the average income of a factory
worker, while a principal officer, here considered part of
the bourgeois, earned a salary at least five times that of a
factory worker (calculations from Kristiania Statistiske
kontor, 1910, pp. 114–15). Although there was some
redistribution of income during the First World War,
the overall result was greater social differences between
the newly rich and old money on the one hand, and the
poor on the other hand. The top five per cent and the
bottom third had, respectively, 60 and 5 per cent of the
total income in 1918–19, while comparable figures in
1909–13 were 35–40 and 10 per cent (calculations from
Kristiania Statistiske kontor, 1910–1919).
Household-level variable
One covariate at the household level is defined. This is
the size of the apartment in which each household lived
in terms of number of rooms. Approximately 95 per cent
of all apartments in Kristiania in 1918 were rented
(Kjeldstadli, 1990). Apartment size appears to be
perfectly correlated with rent: the larger the apartment,
the higher the rent (see Tables 3 and 4). Further, it may
thus be assumed that size of apartments is a proxy,
though an imperfect one, for household income.
Although one had to be among the wealthiest to afford
the rent of an apartment with six, seven or eight rooms
and more (see Tables 3 and 4), the possibility cannot be
ruled out that some of the wealthiest persons may have
preferred to live in smaller apartments, for example of
four or five rooms. Nevertheless, size of apartment is
probably a more direct and crisper proxy for income
than the occupation-based social classes defined above.
Furthermore, the size of each apartment/room in square
meters is not known. For example, a sample survey from
1914 of the two- to three-room apartments (one to two
rooms including kitchen) indicated that apartments were
23 per cent larger in Frogner than in Grønland
(Kristiania Statistiske kontor, 1915). Size of an apart-
ment will therefore not give information on the expected
higher rent and income of people residing in Frogner
compared to Grønland-Wexels. The same survey found
no difference between Frogner and Grønland-Wexels in
the proportion of the apartments considered to be
unhealthy, namely dark, damp and shabby. This was
partly explained by the fact that Frogner had a much
higher proportion of apartments located in basements
than Grønland-Wexels (8.6 vs. 0.8 per cent).
There are several reasons why income and education
(which is assumed to be picked by social class and size of
apartments) may be important determinants of mortal-
ity. First, as most food in Kristiania had to be bought in
stores, nutritional status is dependent on the level of
income. Undernourishment does not increase the
individual’s susceptibility to viral infections such as
influenza (Scrimshaw, Taylor, & Gordon, 1959). On the
other hand, malnutrition associated with a low intake of
nitrogen results in a definite impairment of immune
response and a corresponding increase in susceptibility
to bacterial diseases (Fox, Hall, & Elveback, 1970).
Consequently, whether its victims were undernourished
or not played no role as to where the Spanish influenza
struck. However, bacterial complications following
Spanish influenza, for instance pneumonia, are believed
to have taken a greater toll among those who were
malnourished. Food expenses increased markedly dur-
ing the First World War because of increasing shortages
and rationing. Some groups, in particular those depen-
dent on public assistance—the disabled, widows, aban-
doned wives with children, the old and the sick—may
have experienced additional problems associated with
malnutrition. However, the daily calorie intake of the
working class and among low-paid clerical officers in
Kristiania appeared not to decline during the years
1914–17, because food that was less expensive, more
abundant in supply and equally nutritious replaced the
more expensive food stuffs that were in short supply
(SSB, 1917, 1918a, 1919, 1920). Nevertheless, after a
number of food articles were rationed in the beginning
of 1918, the calorie intake for craftsmen hard at work
(e.g. carpenters, warehousemen) may have been at
subsistence level and possibly only marginal for main-
taining body functions and the ability to work. Second,
the presumably higher nutritional standard of the more
affluent may also have bolstered their immune systems,
better enabling them to fight disease, for instance
tuberculosis. The risk of a fatal outcome was greater
for Spanish influenza patients suffering from active lung
tuberculosis or for Spanish influenza patients who had
reduced lung capacity after having suffered a non-
tubercular lung disease (e.g. chronic bronchitis, bron-
chial asthma, emphysema, and cystic fibrosis) (Noymer
& Garenne, 2000). It was also reported that Spanish
influenza activated latent tuberculosis, which in turn
may have led to higher mortality. Generally speaking,
those with impaired or damaged cardiovascular (e.g.
rheumatic heart disease) and/or respiratory systems are
the most prone to succumbing to pneumonic complica-
tions following influenza. Third, the affluent and the
highly educated classes probably had better chances of
taking time off from work to convalesce when ill than
the poor, as they may have had more saved capital to
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940928
Table 3
Influenza and pneumonia mortality (SMR) and indices of deprivation and wealth for 19 parishes in Kristiania 1918–24
Parish SMR
Adult poor
relief recipients
1919–1923 per
1000 adults 1
December 1920
Child welfare
1918–1924 per
1000 children 1
December 1920
Number of
adults fined
1920–1924 per
1000 adults 1
December 1920
Male adults in
custody 1919,
1921, and 1923
per 1000 adult
men 1
December 1920
Percentage of
pupils aged
7–15 that were
March 1920
number of
persons per
Average monthly rent (in
1918 NOK) for a flat with
kitchen and
Proportion of households 1
February 1918 with
One room Six rooms Maids Bathroom Electricity
City centre
Vor Frelser 96.6 3.3 49 68.7 42.1 12.2 1.4 17.8 128.1 23.2 21.0 72.5
Johannes 92.8 2.9 74 55.6 44.2 14.8 1.5 17.7 112.3 14.6 12.5 75.9
Trefoldighet 116.0 3.3 41 43.1 36.1 12.2 1.5 18.4 106.2 15.2 9.0 70.9
Jacob 90.1 5.5 65 65.9 38.7 12.2 2.1 18.1 98.5 5.4 1.7 80.1
Frogner 74.9** 0.4 9 7.8 9.6 2.9 1.0 20.0 138.3 48.8 67.2 97.0
Uranienborg 81.2* 1.1 20 16.9 12.6 7.0 1.1 19.9 127.4 39.3 42.0 95.0
Fagerborg 82.2* 1.2 16 15.2 18.3 7.9 1.5 16.3 114.4 26.8 30.9 89.7
Gamle Aker 102.3 2.3 25 23.5 16.4 10.8 1.5 19.2 118.5 16.4 21.2 92.3
Markus 111.1 1.1 16 16.9 13.9 10.9 1.5 21.4 102.9 16.2 22.0 95.8
Sagene 108.5 11.5 49 66.1 49.2 13.6 3.1 19.3 56.8 2.3 2.9 78.9
Lilleborg 109.2 9.3 39 52.7 40.3 15.0 2.8 18.1 100.0 3.1 0.7 81.0
Paulus 88.4 6.2 32 54.9 27.7 12.4 2.4 19.5 84.5 2.5 0.1 76.0
Hauges 88.1 4.2 39 52.7 29.7 2.0 18.9 99.6 4.6 1.0 76.4
Petrus 132.9*** 12.9 58 96.2 54.3 13.6 2.9 17.2 60.0 3.0 0.1 62.0
Mathæus 88.2 6.9 29 50.7 52.2 2.3 19.6 85.8 3.2 1.9 71.0
Grønland-Wexels 124.4** 8.0 54 105.6 62.6 14.4 2.5 17.7 84.7 3.1 0.9 62.7
Kampen 100.5 8.2 40 94.8 50.5 14.5 2.8 18.0 89.4 2.6 1.7 77.6
Oslo 109.2 4.3 44 53.6 30.7 12.4 1.9 20.9 91.8 7.3 1.5 75.5
Vaalerengen 116.7 5.1 39 86.4 46.0 14.0 2.5 17.6 86.5 3.2 0.8 79.3
Correlations with
1.00 0.51** 0.53** 0.65** 0.72*** 0.68** 0.53** 0.21 0.59** 0.53** 0.52** 0.62**
*po0:10, **po0:05, ***po0:01.
Source: Parish-specific deaths from influenza and pneumonia by age and sex, parish-specific population by age and sex, and standard population by age and sex for the city as a
whole. These figures are used to calculate the standardized mortality ratios, and are taken from Kristiania sundhetskommision (1919,1920) and Kristiania Statistiske kontor (1920),
respectively. The data on crowding, proportion of households with maids, bathroom and electricity are from Kristiania Statistiske kontor (1920). The data on weight are from
Schiøtz, (1920), while the data on crime, poor-relief recipients, and child welfare cases are from Arctander (1928).
The survey was carried out in March 1920, and included 30,000 children in public schools located all over the city as well as private schools which were essentially located in the
western parishes of Frogner, Uranienborg and Fagerborg. The figures for underweight youth are corrected for a somewhat different distribution of age and sex across each school
(by Schiøtz, (1920)). It should be noted that the schools included in the survey do not recruit pupils exclusively according to the parish borders.
Exclusive kitchen.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 929
live on. A high proportion of the wealthy probably also
had the resources to invest in and benefit from private
health insurance. The wealthy would therefore have
been able to take to their beds when ill, and would
probably have convalesced long enough to avoid
pneumonia and other bacterial complications carrying
Table 4
Results of Cox proportional hazard models for Spanish influenza mortality (N¼250 deaths) in the parishes of Frogner and Grønland-
Wexels combined in 1918
Independent variables Deaths Model 1 Model 2 Model 3
t-stat e
t-stat e
Individual-level variables
0–1 28 11.58*** 9.70 11.55*** 9.65 11.43*** 9.61
2–4 13 2.95*** 3.34 2.95*** 3.33 2.93*** 3.31
5–24 (ref) 31 1.00 — 1.00 — 1.00 —
25–29 28 2.57*** 3.68 2.53*** 3.60 2.59*** 3.68
30–34 25 3.00*** 4.00 2.89*** 3.84 2.99*** 3.96
35–39 15 2.21** 2.42 2.12** 2.29 2.19** 2.39
40–44 13 1.88* 1.82 1.80* 1.69 1.85* 1.78
45–49 11 2.01** 1.90 1.91* 1.75 1.97* 1.84
50–54 8 1.64 1.20 1.59 1.12 1.64 1.19
55–59 8 1.96** 1.63 1.85* 1.48 1.88* 1.53
60–69 21 3.30*** 3.87 3.11*** 3.67 3.18*** 3.73
70–79 28 9.00*** 7.40 8.25*** 7.09 8.43*** 7.16
80+ 16 19.65*** 8.58 16.50*** 8.02 16.49*** 8.03
Female (ref) 119 1.00 — 1.00 —
Male 131 1.48*** 2.79 1.41*** 2.63 1.36** 2.33
Marital status
Never married (ref) 134 1.00 — 1.00 — 1.00 —
Married 82 1.01 0.07 1.08 0.41 1.04 0.24
Widow/widower 32 0.95 0.19 0.97 0.10 0.94 0.24
Separated 1 0.51 0.66 0.52 0.65 0.48 0.73
Divorced 1 1.48 0.39 1.53 0.42 1.69 0.52
Social class
Working class (ref) 152 1.00 — 1.00 — 1.00 —
Middle class 43 0.69** 2.17 0.75 1.62 0.81 1.18
Bourgeois 38 0.61*** 2.69 0.64** 2.06 0.75 1.26
Occupation not stated 17 1.16 0.56 1.21 0.71 1.28 0.89
Household-level variable
Size of apartment (average
monthly rent in 1918 NOK in
One room (11.7 NOK) (ref) 73 1.00 1.00
Two rooms (18.5 NOK) 49 0.70* 1.87 0.67** 2.09
Three rooms (30.1 NOK) 53 0.61*** 2.67 0.59*** 2.89
Four rooms (42.4 NOK) 15 0.42*** 3.04 0.44*** 2.85
Five rooms (62.3 NOK) 18 0.55** 2.19 0.63* 1.66
Six rooms (80.9 NOK) 11 0.36*** 3.04 0.45** 2.31
Seven rooms (101.0 NOK) 18 1.00 0.01 1.22 0.67
Eight rooms+ (8 rooms;
135.5 NOK)
13 0.47** 2.16 0.58 1.48
Frogner (ref) 81 1.00
Grønland-Wexels 169 1.49** 2.05
*po0:10, **po0:05, ***po0:01.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940930
higher fatality risks than influenza. Persons with higher
education were probably also more likely to retain and
follow up the instructions from municipal health
authorities than those of less education.
Parish-level variable
The study only includes a dummy variable (0/1) to
control for place of residence. It is assumed that
residence in Grønland-Wexels may have a significant
and conducive effect on individual mortality relative to
those residing in Frogner because of unmeasured
characteristics of material deprivation, including dilapi-
dated housing, poor sanitation, and pollution in Grøn-
land-Wexels. This effect may remain even when
controlling for individual social class and household-
Method of analysis
The analysis consists of two parts. The first part is
descriptive, and examines the correlation between
mortality from Spanish influenza and indices of
deprivation and wealth in the city of Kristiania as a
whole, with special focus on the contrast between
Grønland-Wexels and Frogner. In the second part of
the analysis, the effects of the covariates described above
upon those surviving Spanish influenza are estimated
using Cox proportional hazards models. The hazard rate
for the individual iwith ncovariates, X¼ðX1;X2;
...:; XnÞ, is modeled as
where tis time elapsed from 1 February 1918 ðt¼0Þto
death from Spanish influenza and where the baseline
hazard h0ðtÞis a hazard function for an individual who
scores zero on all ncovariates. Kaplan–Meier estimators
were calculated for all covariates included in the models.
No serious deviations from the proportionality assump-
tion were found over the time period studied.
Descriptive analysis of the whole capital city of Kristiania
The first cases of influenza associated with Spanish
influenza in Norway were reported in the first week of
April 1918. However, the first scattered cases of
influenza in Kristiania, which later proved to be the
smoldering of a pandemic wave, occurred on 15 June
1918 (Mamelund, 1998). It was not until the first half of
July that the reported disease incidences of influenza
began to skyrocket, taking the dimensions of a
pandemic wave. Fig. 1 clearly shows a second outbreak
of Spanish influenza in 1918, whereby peaks in the crude
death rate and the influenza death rate occurred both in
mid-July and at the end of October. Furthermore, when
comparing the weekly crude death rates in 1918 with the
average monthly crude death rates of the non-pandemic
years of 1915–17 (which may be considered a norm for
mortality levels), it may be seen that the excess in all-
cause mortality is explained by an increase in influenza
mortality. Note also that the peak in mortality during
the summer wave and the fall wave occurred 1–2 weeks
after the respective peaks in disease incidence. Unfortu-
nately, weekly disease and mortality figures for 1919 are
not available.
Table 3 shows that there were clear east–west
differences in Spanish influenza mortality in Kristiania.
In Grønland-Wexels and Petrus, mortality was signifi-
cantly higher than the average for the whole city, while
the mortality in Uranienborg, Fagerborg, and in
particular Frogner, was significantly lower than the
average for the whole city. Mortality in Grønland-
Wexels was 50 per cent higher than in Frogner. The
east–west differences may also be seen for several indices
of deprivation and wealth (Table 3). As for mortality,
Frogner is ranked last on all indices for deprivation, but
first on all measures for wealth, while the opposite is
generally true for Grønland-Wexels. The differences in
the indicators between the two parishes are all statisti-
cally significant at a very low level. The bivariate
correlations between the SMRs and the socioeconomic
characteristics for the 19 parishes are moderately strong,
but follow the expected directions. All are significant at
the 0.05 level, except for monthly average rent for a flat
with one room and kitchen. It should be remembered
that the associations between the SMRs and the indices
of deprivation measured in the years immediately
following 1918 must be considered with caution because
the deprivation may be a consequence of the influenza
pandemic and not the other way around. However, the
regional differences in these indices of poverty are
considered to be relatively stable, and are probably also
good proxies for several years prior to 1918 (Arctander,
Of particular interest is the finding that the negative
correlation between the SMRs and average rent of
apartments increases with size (results only shown for
apartments of, respectively, one and six rooms, see Table
3). This indicates that mortality differences between
those who only could afford to rent small apartments in
the two parishes are smaller than for those who could
afford to rent large apartments. Furthermore, it is also
an intriguing finding that the difference in the average
rent of an apartment of the same size between Frogner
and Grønland-Wexels increases in a linear fashion with
size (not shown); it was found that apartments with one
room including kitchen or six rooms including kitchen
were, respectively, 1.1 and 1.7 times more expensive in
Frogner than in Grønland-Wexels. This result may
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 931
indicate that people were willing to pay more for
apartments, especially the larger ones located in a highly
privileged parish with a good reputation (e.g. less crime)
than for accommodation located in a socially deprived
area with a bad reputation. However, the differences in
the rent between the two parishes may be explained in
part by the fact that apartments, at least those with two
to three rooms, were larger in Frogner than in Grønland
(Kristiania Statistiske kontor, 1915).
Multivariate analysis of Grønland-Wexels and Frogner
According to model 1, the mortality from Spanish
influenza was, respectively, 39 and 31 per cent lower
among individuals belonging to the bourgeois and the
middle class than among individuals in the working
class, net of the effect of age, sex, and marital status
(Table 4). The effect is highly significant in statistical
terms and seems to contest the commonly held assump-
tion that mortality from Spanish influenza was class
neutral. The mortality among the bourgeois was 11.6 per
cent lower than among the middle class, but the
difference was far from being statistically significant at
the 0.10 level.
The mortality premium of individuals belonging to
the two upper classes relative to the lowest class remains
statistically significant only for the bourgeois when
apartment size, a proxy for household income, is
included in the analysis in model 2. Anyway, as might
be expected, the mortality in the two upper classes was
still much lower than in the working class, respectively,
by 25 per cent in the middle class and 36 per cent in the
bourgeois (Table 4). Mortality was gradually falling for
individuals living in apartments with up to six rooms
compared to those residing in one-room apartments.
The nearly perfect linear decline is broken only by those
residing in five-room apartments. The drop in mortality
from one apartment size category to the next for
apartments with two to six rooms in model 2 is not
statistically significant at the 0.10 level, but seems
nevertheless to coincide with theory. However, those
who lived in six-room apartments had significantly
lower mortality than those living in apartments with up
to two rooms. Of all apartment size categories, only
those residing in apartments with seven rooms did not
demonstrate significantly different mortality from those
living in one-room apartments.
When a control for residence in either Frogner
or Grønland-Wexels is included in model 3 while
1 3 5 7 9 11131517192123252729313335373941434547495153
Deaths per 1 000
Influenza cases per 1 000
Weekly crude death rates 1918
Weekly death rates of influenza and pneumonia 1918
Monthly crude death rates 1915-17
Weekly reported incidence of influenza 1918
Fig. 1. Weekly crude death rates, incidence rates and death rates from influenza in 1918, and monthly average crude death rates for the
years 1915–17 in Kristiania. Sources:Kristiania Sundhetskommision, 1919;Mamelund, 2003a.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940932
individual-level and household-level variables are simul-
taneously accounted for, it is found that the negative
effect on mortality of belonging to the bourgeoisie is no
longer statistically significant (Table 4). Nevertheless,
the mortality in the two most advantageous classes is
still, respectively, 19 (middle class) and 25 per cent
(bourgeois) lower compared to the working class. This
finding thus remains inconsistent with the conservative
and dominant class neutrality hypothesis. However, the
gradually dropping mortality by apartment size remains
statistically significant, the exceptions being the two
largest apartment size categories.
In model 3, it was estimated that Spanish influenza
mortality was 49 per cent higher in the deprived parish
of Grønland-Wexels than in the privileged parish of
Frogner, all other factors being the same. (Note that the
difference in mortality between the parishes in the
multivariate analysis is comparable to that found in the
descriptive analysis.) This would suggest that there are
unobserved and unaccounted factors at the parish level
that affected Spanish influenza mortality above and
beyond the characteristics of individuals and house-
holds. Alternatively, the parish of residence picks up
individual- or household-level variations due to omitted
or poorly specified variables already included in the
Surprisingly, there are no significant effects of marital
status on mortality. However, when estimating model 3
in Table 4 for all-cause mortality (608 deaths) rather
than for Spanish influenza mortality (250 deaths), it
appeared that married persons had 20 per cent lower
mortality than the reference group of never married
(significant at 0.10 level). The W-shaped effects of age
and sex on mortality are shown as expected in all models
net of the effect of other variables. There was a higher
risk of mortality for infants, young adults and the
elderly, and the risk of mortality was higher for males
than females. Only infants and individuals 70 years and
older had significantly and substantially higher mortality
than individuals 30–34 years of age. There was a
continuous decline in mortality from one 5 year age
group to the next for ages 30–34 to 50–54, but the
mortality risk of the 50–54 year age group was not
statistically different from those in the high-risk age
group of 30–34 years. Although half of the genetic
material of the Spanish influenza virus has been
discovered in the last couple of years, it is still not
understood why the death rate in young adults from
Spanish influenza was relatively large compared to a
normal epidemic influenza season (Oxford et al., 2005).
In model 3, it was found that the mortality among
men from Spanish influenza was 35 per cent higher than
the mortality among women when all other factors were
the same. There are usually little or no discernable
male–female influenza mortality differences during a
normal epidemic influenza season, but in 1918–19, the
influenza mortality for males in the age group 20–40
increased far more than for females in many countries,
as was observed for example in the United States,
Norway, and New Zealand (Crosby, 2003;Mamelund,
1998;Rice, 1988). Noymer and Garenne (2000) have
convincingly argued that the differences between the
sexes in mortality rates from Spanish influenza might be
due to the higher disease incidence of tuberculosis in
men than in women. In this paper, interactions of age
and sex are not included in the models because such
interactions are not thought to affect—or add to the
understanding—of the variables of focus in this study,
namely social class and wealth. Moreover, the models
with only main effects presented here may already be at
the limit of being methodologically sound because of the
low number of events (deaths) available, and a further
separation of the independent variables would probably
have made the estimated coefficients unreliable and
difficult to interpret.
Effects of individual social class and household wealth on
One of the main findings in this analysis is that
mortality from Spanish influenza was 19–25 per cent
lower among the two upper classes compared to the
working class when all other factors were the same
(model 3). This relationship between mortality and
social class was not statistically significant, but never-
theless in accordance with the hypothesis. Furthermore,
when using apartment size as a proxy for household
income, it was found that size of apartment had a
negative effect on mortality. This relationship is
statistically significant and partly linear, all other factors
being the same. For example, it was estimated that those
living in apartments with two, three and four rooms had,
respectively, 34, 41 and 56 per cent lower mortality than
those residing in a one-room apartment.
It is not surprising that the largest relative effects on
mortality of any of the covariates included in the models
are found for infants, young adults and persons older
than age 70. However, it is interesting to note that the
relative mortality risk for persons 30–34 years of age
(experienced a three times greater risk of mortality than
the reference category), a population group that
normally has little to fear from influenza, was much
greater than the largest relative mortality risk for any
apartment size category (households residing in apart-
ments with four and six rooms experienced 0.55–0.56
times lower risks than the reference category). Never-
theless, the results of this paper seem to contest the
commonly held view that the Spanish influenza pan-
demic was a socially neutral disease with respect to
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 933
mortality. It also challenges the premise that the social
mortality profile from Spanish influenza was any
different than that of a number of other causes of
death, including mortality in more recent influenza
epidemics (Kitagawa & Hauser, 1968, 1973;Regidor et
al., 2003;Singh & Siahpush, 2001). More specifically,
the socioeconomic differences in mortality reported in
this paper are consistent with several other modern and
historical studies of general as well as cause-specific
mortality in Kristiania by class, wealth and parish (e.g.
Barstad, 1997;Geirsvold, 1917;Gjestland & Moen,
1988;Hansen, 1985;Rognerud & Stensvold, 1998). The
findings are also consistent with the cross-sectional and
multivariate studies of Spanish influenza by McCracken
and Curson (2003),Mamelund (2003b) and Johnson
(2001) for, respectively, the city of Sydney, Australia, for
Norway, and for England and Wales; however, they do
not concur with the findings of a similar study by
Brainerd and Siegler (2003) for the United States.
Likewise, the results also concur with the contemporary
house-to-house influenza surveys conducted for cities
in the United States and Bergen, Norway, in 1918
(Britten, 1932;Collins, 1931;Hanssen, 1923;Syden-
stricker, 1931;Vaughan, 1921), although not with a
study of certain English cities (Great Britain Ministry of
Health, 1920).
How large are the class and wealth differentials in
Spanish influenza mortality found in this paper com-
pared to other studies? Unfortunately, there are no other
studies of Spanish influenza which have used individual-
level data and multivariate models to estimate the effects
of class or social factors on mortality. However, there
are a couple of studies of more recent influenza
epidemics who have used comparable data and methods.
The socioeconomic indices (direct measures of education
and income) used in these studies are not directly
comparable to the ones used here (social class as proxy
for individual-level income and education, and size of
apartment as proxy for income levels of households),
but a comparison of the size of the effects of these
measures on mortality is nevertheless worth doing. For
example, Kitagawa and Hauser (1968) in their study of
mortality from combined influenza and pneumonia in
the United States in 1960 have found that male and
female (white) mortality amongst those with less than 8
years of schooling was, respectively, 1.6 and 1.7 times
that of those with one or more years at college. Another
study by the same authors documents that the mortality
of men from low-income families in Chicago in 1950 was
twice that of men from high-income families (Kitagawa
& Hauser, 1973). Singh and Siahpush (2001) have found
similar negative effects of education and income on
mortality for the United States in the period between
1979 and 1989, but these effects were only statistically
significant for men; the combined influenza and pneu-
monia mortality of those with 8 or fewer years of
schooling was 1.4 times that of those with 16 or more
years of education, while the mortality of low-income
families was 2.7 times that of high-income families. In a
study of Madrid, Spain, in the time period 1996–97,
Regidor et al. (2003) demonstrated that a reduction in
education of one year caused a 3.7 per cent increase in
mortality among men and 3.4 per cent among women.
The average Spanish influenza mortality of those living
in four- to six-room apartments in Kristiania in 1918–19
was 50 per cent lower than that of those living in one-
room apartments, i.e. a relative effect not very different
from that for education in the United States 1960 and
1979–89, but much lower than the relative income effect
on mortality in Chicago 1950 and the United States
1979–89. When it is assumed that mortality risks in
model 3 in the present paper are a linear function of size
of apartments, a reduction of size of an apartment with
one room caused an increase in mortality by 7 per cent,
i.e. an effect twice as large as that found for education in
the Madrid study.
How important, in epidemiological terms, are the
socioeconomic differences in the combined influenza and
pneumonia mortality demonstrated in the present and
previous studies? In other words, how large are the
socioeconomic differences in influenza and pneumonia
mortality compared to corresponding differences for
other causes of death? Kitagawa and Hauser (1973)
showed that influenza and pneumonia were the two
causes of death that showed the greatest socioeconomic
differences in Chicago in 1950, surpassed only by
tuberculosis, while Singh and Siahpush (2001) found
that only mortality from stomach cancer and homicides
displayed larger socioeconomic differences than influen-
za and pneumonia combined. Estimations of socio-
economic mortality differences for other causes of death
in 1918, in particular tuberculosis, would probably be
distorted because many individuals who died in that
year would normally have died from other causes than
Spanish influenza (see Noymer & Garenne, 2000).
There are a number of interesting interactions that
were not modeled in the present paper due to the rather
limited number of events (deaths) available. Among
them is the interaction between age and sex. In the
models that were estimated, it was indirectly assumed
that households that resided in apartments of equal size
were equally wealthy across the two parishes studied,
and that the different classes were similarly homoge-
neous. However, in the descriptive analysis it was for
example found that for apartments of equal size,
households in Frogner paid more in monthly rent than
those in Grønland, and that this difference increased
with the size of the apartment. A control for such
differences would have been made if interactions
between parish and apartment size had been included
in the models; nevertheless the even more important
questions would have remained unanswered, namely
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940934
what is the household income and the percentage of
income spent on rent?
Explaining the contextual effects on mortality
Another important finding of the analysis is that the
population residing in one of the most deprived parishes
in the Norwegian capital in 1918, Grønland-Wexels,
experienced a 49 per cent higher mortality from Spanish
influenza than the population living in the chiefly
wealthiest parish of Frogner, even after controlling for
individual occupation-based social class and apartment
size. How may this contextual effect be explained?
The contextual effect may be of two types. The first is
group means for individual-level variables, e.g., average
income, wealth or the level of education of the
individuals residing in an area. The second type is
commonly referred to as being global, for example
geographical characteristics confined to an area. Global
effects include such factors as climate, pollution,
reputation of an area, quality of housing, the location
of a hospital or medical resources, in general, aspects
that cannot be summed up from individual-level
characteristics. In this paper, contextual effects are
picked up in a rather simple way through the use of a
dummy variable. Simply substituting the dummy vari-
able with one of the deprivation or wealth indices seen in
Table 3 would of course not give sufficient variation to
explain why location of residence should have an
independent effect on individual mortality. Ideally,
some global or group mean variable at a level lower
than the parish, e.g., a locally based definition of
neighborhoods, or administrative units like census
tracts, wards, blocks or streets, should therefore have
been included in the analysis. In principle, it would have
been possible to sum up occupation-based social class or
other social factors for a given unit from the individual-
level data used in the analysis, but this approach was not
pursued due to the limited number of events (deaths)
available. Instead, the analysis relies on official statistics
on socioeconomic characteristics and the local history
and sociology of the two parishes in question. Other
studies not related to Spanish influenza that have
analyzed a few well-defined and intentionally selected
socially contrasting neighborhoods have also been
extremely useful in assessing the role of individual and
contextual effects on individual mortality.
Effects of group means?
Concentrated poverty in Grønland-Wexels. There are
several multilevel studies which have found no or only
weak effects of area of residence on mortality when
controlling for individual-level socioeconomic indices,
while other studies have demonstrated a strong effect on
mortality of living in socially deprived areas (Pickett &
Pearl, 2001). A classic example of the latter is a study of
a federally designated ‘‘poverty area’’ in the city of
Oakland, California (Haan, Kaplam, & Camacho,
1987), while a more recent example is a study of
Renfrew/Paisley, the most deprived area in Scotland
(Davey Smith, Hart, Watt, Hole, & Hawthorne, 1998).
The Haan et al. (1987) study is particularly interesting
because the authors found no significant effect of
individual socioeconomic indices on mortality while
the poverty area dummy variable was highly significant.
The suggested explanation was that the residents in the
poverty area were exposed to higher crime rates, poorer
housing, lack of transportation, and higher levels of
environmental contaminants. The results of the Haan et
al. (1987) study, including analyses of specific causes of
death, have also been replicated for several other
metropolitan poverty areas in the United States (Waitz-
man & Smith, 1998).
The extreme social polarization between Grønland-
Wexels and Frogner had prevailed for more than half a
century by 1918. Thus, there seems to be good reasons
to believe that the dummy variable for parish picks up
some of these well-known but unaccounted and strongly
significant differences in wealth and poverty (see Table
3). Moreover, not only official deprivation statistics, but
also qualitative research shows that Grønland-Wexels
could have been defined a ‘‘poverty area’’ as in Haan et
al. (1987), with certain apartment blocks (‘‘gra
der’’) and streets demonstrating concentrated poverty
(Kjeldstadli, 1990, p. 71). In a quantitative and
qualitative study of the spatial differences in poverty
and mortality in Kristiania, Arctander (1928, p. 143)
concluded that living in these pockets of extreme
poverty for a long period of time had an independent
and deteriorating effect on physical and mental health
above and beyond the individual’s own social status.
Arctander (1928) also made a statistical comparison of
the deprivation status of different streets within each
parish and combined it with statements from various
local child welfare clerks and poor-relief assistants. A
deprivation score was calculated combining the disease
incidence of tuberculosis, proportion of adult poor-relief
recipients, child welfare cases, and crime rates, but
unfortunately, these scores were not published by
identifiable street names. Arctander (1928) found that
the most disadvantageous streets created a suburban
belt, including a large part of Grønland-Wexels, which
embedded the city center. This pre-industrial area was
originally built outside the city, but as the city sprawled,
it was left as a belt with slums encircling the city center
(Kjeldstadli, 1990, p. 62). No disadvantageous streets
were found in Frogner. Unfortunately, Arctander (1928)
did not classify the streets according to indices of wealth.
Concentrated affluence: is there a health culture in
Frogner?Lagasse, Humblet, Lenaerts, Godin, and
Moens (1990) have disputed the finding that an area
may impose low mortality on its resident individuals net
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 935
of the effect of individual socioeconomic factors,
arguing that this may be more the result of a ‘‘health
culture’’ among wealthy and highly educated people
residing in this area. Several studies have documented
that social differences exist in the ability to make self-
diagnosis and the ability to understand and to retain
general knowledge about health care, and that such
differences may result in class differences in disease
incidence and mortality (Feinstein, 1993;Townsend &
Davidson, 1982). Furthermore, while becoming aware of
health information is important, there may also be
differences between the socioeconomic classes in their
awareness of the importance of also following through
on the instructions issued by health authorities or
The health authorities in Kristiania basically limited
their activities to surveillance of cleanliness and sanita-
tion in the city throughout the Spanish influenza
pandemic, and to give precautionary health advice
(Kristiania Sundhetskommision, 1919). In mid-October
1918, the health authorities issued an advisory urging
people not to voluntarily expose themselves to infection,
especially those who were not infected during the
relatively mild summer wave and had thereby not gained
relative immunity. In addition, people were urged to
wash their hands and to refrain from the bad habit of
spitting, to cover their mouth when coughing or
sneezing, to go to bed as early as possible after the
onset of influenza symptoms, and to remain in bed until
they were free of fever. This information was printed in
the newspapers and on posters in public places, but
probably fewer of the less well educated and thus more
of the bourgeois and the middle class than working class
became aware of the importance of the messages. It may
therefore be possible that individuals living in Frogner
had relatively low mortality because they lived in an area
where a majority of the neighbors belonged to the
industrialist or cultural elite who are assumed to be
highly conscious of health matters. People living in
Frogner would therefore run a relatively low risk of
being infected by colleagues at work, neighbors or
family members because they would have taken to their
beds early after the onset of influenza symptoms, and
because they stayed in bed until free of symptoms
(unfortunately, morbidity data to test this hypothesis is
not available).
Global effects?
Location of working class employment. It has pre-
viously been reported that Spanish influenza in Norway
first started in congested industrial areas, spreading
most rapidly within the working class population
(Mamelund, 1998). Several of the factories that offered
employment to the working class in Kristiania were
located along the Aker River in the eastern parishes of
the city. Because a large part of the working class in
Grønland-Wexels probably both lived and worked in
the same community, and worked at factories and
industries where they encountered numerous workers
from neighboring parishes, the influenza may have
spread faster here than among the elite who lived in
Frogner. These people worked either downtown where
most of the offices of private firms where located, or in
public offices which were scattered all over the city.
Furthermore, most everyday interactions, for instance
grocery shopping, or a visit to some city office, took
place within the parish of residence, or at least,
respectively, within the eastern or western side of the
city (Kjeldstadli, 1990).
Location of hospitals and medical resources. There
were no effective vaccines or antiviral drugs in 1918 to
combat Spanish influenza. Doctors and nurses were
therefore more or less helpless, and patients who were
hospitalized and who received care from professional
health practitioners did not appear to have any lower
mortality than those who were nursed by their families
at home. Indeed, on the contrary, quiet nursing at home
may have been the key to survival. Hospitalization may
in fact have worsened the condition of a patient due to
transmission and spreading of bacteria such as pneu-
mococcus within the wards (Oxford, Sefton, Jackson,
Johnson, & Daniels, 1999). The argument that the
affluent could afford better (private) health care and
medicine than the poor may thus not be the most
relevant for explaining the differences in Spanish
influenza mortality.
Location of dilapidated housing. The quality of housing
may have played a role in the risk of developing and
dying from influenza. This study did not include
indicators of the socio-environmental risk factors
associated with dilapidated housing. Reference has
already been made of a survey from 1914 showing that
there was no difference between Frogner and Grønland-
Wexels in the proportion of the two- to three-room
apartments considered to be unhealthy (Kristiania
Statistiske kontor, 1915). Little is known about the
relative conditions of apartments with three or more
rooms. However, because of increasing shortage of
housing during the 1914–18 war (only 0.1 per cent of the
apartments were on average vacant), an increasing
number of families, probably more in Grønland-Wexels
than in Frogner, were forced to live in cold and damp
basements, draughty attic stories, garden pavilions and
hen houses, in conditions not normally permitted for
human habitation by the city government (SSB, 1955).
Low-income groups probably also lacked sufficient
heating, despite their access to discount stamps on coal
(SSB, 1918b). In addition to poor ventilation, hygiene
and sanitation, all of the conditions mentioned above
may be associated with respiratory symptoms, reduced
lung function, and lower socioeconomic status (Prescott
& Vestbo, 1999). Some of the effect of parish on
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940936
individual mortality may therefore be explained by the
fact that the dummy variable picks up unaccounted
spatial differences in the quality of housing.
Unspecified confounders?
In the analysis it was found that individual occupa-
tion-based social class had a weak and non-significant
effect on mortality net of the effect of a seemingly strong
effect of place of residence. A reason for this may also be
that the dummy variable for parish picks up unobserved
and unaccounted variation at the individual level. This is
because occupational social class may not have been
modeled with enough theoretical crispness with respect
to education and income. The dummy variable for
parish may also pick up unobserved and unaccounted
variations at the household level. The reason that
households residing in apartments of seven or more
rooms did not display significantly lower mortality than
households residing in one-room apartments may be
that size of apartment is only a proxy for household
income. In other words, if direct information on income
and education had been available and included in the
models, instead of proxies for these two indices of
wealth, the seemingly positive effect of residency in
Grønland-Wexels on mortality may have disappeared.
Omitted confounders?
A further reason that place of residence seems to play
an important role in explaining the variation in mortality
may be that it picks up individual-level variation due to
omitted confounders. Anthropometric measures as
proxies of nutritional history and information on pre-
existing diseases which are associated with a higher
susceptibility to influenza are examples of omitted
individual-level variables that probably could have added
further insight into the variation of mortality (although
some of these aspects may be accounted for by
occupation-based social class or size of apartment).
Unfortunately, such information is not available in the
current database. The relatively high bivariate correlation
found between the percentage of the pupils that were
underweight in the different parishes in Kristiania in 1920
and mortality from Spanish influenza in 1918–19 (see
Table 3), lends support to the hypothesis that poor
nutrition leads to higher mortality. Furthermore, Eche-
verri (2003) and Mamelund (2003b) have found signifi-
cant negative associations between the average height of
conscripts and influenza death rates in, respectively,
Spain and Norway in 1918–19 using cross-sectional data.
Geirsvold (1917) reported that mortality from tubercu-
losis was three times higher in Grønland (2.9 deaths per
1000) than in Frogner (0.8 deaths per 1000) during the
First World War. Arctander (1928) also reported similar
differences between the same parishes. Actual work
experience—including work load, type of work (man-
ual/non manual), working conditions, whether a worker
does shifts, works overtime, and how many hours are
worked per day—is a final example of omitted and
unaccounted information not included in this study
(although some of these aspects may be accounted for
by occupation-based social class). It may be reasonable to
believe that many years of heavy manual and physical
labor, at least 10 h a day possibly combined with poor
and overcrowded working conditions, might have da-
maged or impaired the general health of the working class
relative to that of the bourgeois and the middle class.
Debate has continued in the literature ever since 1918
over whether there were socioeconomic differences in
mortality from the 1918–19 pandemic, with massive
support for the view that it was socially neutral. Cox
proportional hazard models and unique data at the level
of individuals and households combined with a control
for place of residence are used for the very first time in
this paper to ascertain whether there were differences in
mortality from Spanish influenza with respect to
individual occupation-based social class and household
wealth in two socially contrasting parishes in the
Norwegian capital of Kristiania (renamed Oslo in
1924) in 1918.
The analysis showed that mortality for the two upper
classes was 19–25 per cent lower compared to the
working class (not significant) and that the mortality of
those living in apartments with four to six rooms were
on average 50 per cent lower than those residing in one-
room apartments (significant). The analysis also showed
that when individual social class and household-level
wealth is accounted for, simply living in a ‘‘poverty
area’’ has a significant effect in explaining the variance
in Spanish influenza mortality. Those living in one of the
most impoverished parishes in the city, Grønland-
Wexels, had 49 per cent higher mortality than those
residing in the Norwegian capital’s most privileged
parish of Frogner. The finding may have two explana-
tions: first, the parish of residence may pick up income,
wealth and educational differences imperfectly captured
by social class and size of apartments, or alternatively,
parish may pick up other omitted and unaccounted
social and/or epidemiological risk factors. Second, the
effect of parish may also be explained by unmeasured
characteristics of material deprivation, including dilapi-
dated housing, poor sanitation and pollution in the
‘‘impoverished area’’. Furthermore, because the wealthy
understood the importance of remaining in bed until free
of symptoms, and indeed, were also in the economic
position to do so, they had a greater chance of avoiding
pneumonia which often followed severe cases of
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 937
The findings in this paper challenge the conservative
view that Spanish influenza was an ‘‘egalitarian’’ or
classless disease whose victims were struck randomly in
terms of mortality from the disease. In addition to the
four most peculiar and well-documented features of the
Spanish influenza, its high death toll, with 50–100
million deaths worldwide, the relatively high overall
lethality, its proclivity to affect males, and the fact that
the highest relative increase in the death rates was
experienced by those between the age of 20 and 40,
another prominent, but perhaps non-peculiar feature of
the disease in Kristiania, and possibly also in other
locations, was the significantly lower mortality of the
wealthy residing in large apartments, and the higher
mortality in the impoverished parish of Grønland-
Wexels compared to the wealthy parish of Frogner.
The findings in this study are important for two
reasons. First, the socioeconomic differences in mortal-
ity from Spanish influenza in Kristiania are comparable
to later studies in which it was found that mortality from
influenza and pneumonia in annual epidemics demon-
strates some of the largest socioeconomic differences of
all causes of death. Second, the results from the present
study are of international relevance because unique
individual and household-level data have been used to
make the analysis. This type of data is difficult to obtain
for other countries; furthermore, because Norway was
neutral and as a whole not affected by the war, it was
therefore possible to estimate an effect of the Spanish
influenza on mortality net of the effect of the war.
However, in future studies it would be desirable to have
an even larger sample to work with to ensure not only
large variation in the independent variables, but also
more variation in the dependent variable, namely the
number of deaths.
The World Health Organization and leading experts
on influenza agree that another influenza pandemic is
inevitable and possibly imminent. A new pandemic may
arise if avian influenza not only develops ability to jump
directly from birds to humans, as has been reported to
have occurred in Southeast Asia several times since
1997, but also goes trough a genetic transformation such
that the virus that causes the disease can easily jump
from human to human. The worst case scenario is a
pandemic like Spanish influenza. Because relatively large
social differences in mortality from Spanish influenza as
well as later influenza epidemics have been documented,
it is tempting to predict significant socioeconomic
differences in mortality risks when the next influenza
pandemic comes.
I am most grateful to Sølvi Sogner, Nico Keilman,
Øystein Kravdal, Anders Barstad, and Hans Henrik
Bull for their helpful comments to the paper. My thanks
also go to the staff of Oslo City Archives, who were
helpful in localizing the data applied in this paper, and
for providing my research assistant Kirsti Hansen with
excellent working conditions while making copies of the
data. Furthermore, the staff at the Norwegian Historical
Data Center at the University of Tromsø and Bardufoss
rendered invaluable assistance, in particular Gunnar
Thorvaldsen, Randi Eriksen, and Marianne Jarnæs
Erikstad, who did a great job digitalizing the data. I
would also like to thank research assistant Erik Wold
Aunemo and my colleague Ka
˚re Bævre who carried out
the record linkages. The paper is part of the research
project Spanish influenza and beyond: The case of
Norway, funded under a research grant by the Norwe-
gian Research Council and Department of Economics,
University of Oslo, to whom I extend my thanks.
Arctander, S. (1928). Miljøforholdene i Oslo. En socialstatistisk
studie. Oslo: Særtrykk i det Statistiske Centralbyra
Barstad, A. (1997). Store byer, liten velferd? SØS (Social and
Economic Studies) 97. Statistisk sentralbyra
Brainerd, E., & Siegler, M. V. (2003). The Economic Effects of
the 1918 Influenza Epidemic, Discussion Paper Series at
Centre for Economic Policy Research, No. 3791. (Available
online at
Britten, R. H. (1932). The incidence of epidemic influenza,
1918–19. Public Health Reports,47(6), 304–339.
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), 1909–1937.
Crosby, A. W. (2003). America’s forgotten pandemic: The
influenza of 1918. Cambridge: Cambridge University Press.
Davey Smith, G., Hart, C., Watt, G., Hole, D., & Hawthorne,
V. (1998). Individual social class, area-based deprivation,
cardiovascular risk factors, and mortality: The Renfrew and
Paisley study. Journal of Epidemiology and Community
Health,52, 399–405.
Dutton, D. B. (1988). Worse than the disease: Pitfalls of medical
progress. Cambridge: Cambridge University Press.
Echeverri, B. (2003). Spanish influenza seen from Spain. In H.
Phillips, & D. Killingray (Eds.), The Spanish influenza
pandemic of 1918–19. New perspectives (pp. 173–190).
London: Routledge Social History of Medicine series.
Feinstein, J. S. (1993). The relationship between socioeconomic
status and health: A review of the literature. The Milbank
Quarterly,71(2), 279–322.
Fox, J. P., Hall, C. E., & Elveback, L. R. (1970). Epidemiol-
ogy—Man and diseases. London: Macmillan.
Geirsvold, M. (1917). Boligforholdenes inflydelse pa
˚sygdom og
sundhet. Tidsskr nor lægeforen,37(14), 620–628.
Gjestland, T., & Moen, E. (1988). ‘‘East is east, and west is west-
’’: levealderen i Oslo er fortsatt lavest øst for Akerselva: en
sammenliknende undersøkelse av dødeligheten i Oslo øst og
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940938
Oslo vest i periodene 1890–1940 og 1971–1980. NIBR-
rapport no. 21. Oslo: Norsk institutt for by- og regions-
Glezen, W. P., Paredes, A., & Taber, L. H. (1980). Influenza
in children. Relationship to other respiratory agents.
Journal of the American Medical Association,243(13),
Great Britain Ministry of Health. (1920). Report on the
1918–19 pandemic of influenza. Reports on public health
and medical subjects. No. 4. GBMH, London.
Haan, M., Kaplam, G. A., & Camacho, T. (1987). Poverty and
health. Prospective evidence from the Alameda county
study. American Journal of Epidemiology,125(6), 989–998.
Hansen, L. I. K. (1985). Koleraen i Christiania i 1853. Oslo:
Seksjon for medisinsk historie, Universitetet i Oslo.
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 Bok-
trykkeri og N. Nilssen & søn.
Hersch, L. (1920). 0L0ine
´devant la mort d’apre
´s les
statistiques de la ville de Paris. Revue d’Economie Politique,
34, 273–302.
Hersch, L. (1932). Paurete
´et mortalite
´selon les principales
causes de de
`s d’apre
`s les statistiques de la Ville de Paris.
Johnson, N. P. A. S. (2001). Aspects of the historical geography
of the 1918–19 influenza pandemic in Britain. Ph.D. thesis.
University of Cambridge, 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, 105–115.
Kitagawa, E. M., & Hauser, P. M. (1968). Education
differentials in mortality by cause of death: United States,
1960. Demography,5(1), 318–353.
Kitagawa, E. M., & Hauser, P. M. (1973). Differential mortality
in the United States: A study in socioeconomic epidemiology.
Cambridge, MA: Harvard University Press.
Kjeldstadli, K. (1990). Oslo bys historie. den delte byen. fra 1900
til 1948. bind 4. Oslo: J.W. Cappelen Forlag A S.
Kravdal, Ø. (2003). Children, family and cancer survival in
Norway. International Journal of Cancer,105(2), 261–266.
Kristiania Statistiske kontor. (1900, 1910, 1920). Statistisk
aarbok for Kristiania by 1901, 1909, 1918. Kristiania.
Kristiania Statistiske kontor. (1910–1919). Statistisk aarbok for
Kristiania by 1909–1918. Kristiania.
Kristiania Statistiske kontor. (1915). Beboelsesforholdene i
smaaleiligheter i Kristiania 1913/1914. Kristiania.
Kristiania sundhetskommision. (1919, 1920). Beretning fra
Kristiania sundhetskommision og Kristiania kommunale
sykekasse for a
˚ret 1918, 1919. Kristiania.
Kunst, A. E., & Mackenbach, J. P. (1994). International
variation in the size of mortality differences associated with
occupational status. International Journal of Epidemiology,
23, 742–750.
Lagasse, R., Humblet, P. C., Lenaerts, A., Godin, I., & Moens,
G. F. G. (1990). Health and inequities in Belgium. Social
Science & Medicine,31(3), 237–248.
¨, P., Valkonen, T., & Martelin, T. (1997). Contribution
of deaths related to alcohol use to socioeconomic variation
in mortality: Register based follow up study. British Medical
Journal,315, 211–216.
Mamelund, S.-E. (1998). Spanskesyken i Norge 1918–1920:
Diffusjon og demografiske konsekvenser. Hovedoppgave i
Samfunnsgeografi høsten 1998. Unpublished Masters thesis,
Institutt for Sosiologi og Samfunnsgeografi, Universitetet i
Oslo, Oslo.
Mamelund, S-E. (2003a). Can the Spanish influenza pandemic
of 1918 explain the baby boom of 1920 in neutral Norway?
Population-E,59(2), 229–260.
Mamelund, S-E. (2003b). Spanish influenza mortality of ethnic
minorities in Norway 1918–1919. European Journal of
Population,19(1), 83–102.
Mamelund, S.-E. (2004). An egalitarian disease? Socioeconomic
status and individual survival of the Spanish Influenza
pandemic of 1918–19 in the Norwegian capital of Kristia-
nia, Memorandum,No06/2004, Department of Economics,
University of Oslo (see pdf format at www.oekonomi.uio.-
McCracken, K., & Curson, P. (2003). Flu downunder: A
demographic and geographic analysis of the 1919 pandemic
in Sydney, Australia. In H. Phillips, & D. Killingray (Eds.),
The Spanish influenza Pandemic of 1918–19. New perspec-
tives (pp. 110–131). London: Routledge Social History of
Medicine series.
Myhre, J. E. (1990). Oslo Bys Historie. Hovedstaden Christiania:
fra 1814 til 1900. Bind 3. Oslo: J.W. Cappelen Forlag A S.
Noymer, A., & Garenne, M. (2000). The 1918 influenza
epidemic’s effect on sex differentials in mortality in the
United States. Population and Development Review,26(3),
Oslo Byarkiv (Oslo City Archive), Anmeldte døde i Oslo
Oslo Byarkiv (Oslo City Archive), folketellingene for Kristiania
1 februar 1918 og 1 februar 1919.
Oxford, J. S., Lambkin, R., Sefton, A., Daniels, R., Elliot, A.,
Brown, R., et al. (2005). A hypothesis: The conjunction of
soldiers, gas, pigs, ducks, geese and horses in Northern
France during the Great War provided the conditions for
the emergence of the ‘‘Spanish’’ influenza pandemic of
1918–1919. Vaccine,23, 940–945.
Oxford, J. S., Sefton, A., Jackson, R., Johnson, N. P. A. S., &
Daniels, R. S. (1999). Who’s that lady? Nature Medicine,
5(12), 1351–1352.
Pickett, K. E., & Pearl, M. (2001). Multilevel analyses of
neighbourhood socioeconomic context and health out-
comes: A critical review. Journal of Epidemiology and
Community Health,55, 111–122.
Prescott, E., & Vestbo, J. (1999). Socioeconomic status and
chronic obstructive pulmonary disease. Thorax,54,
Regidor, E., Calle, M. E., Navarro, P., & Domı
´nguez, V.
(2003). The size of educational differences in mortality from
specific causes of death in men and women. European
Journal of Epidemiology,18, 395–400.
Rice, G. (1988). Black November. The 1918 influenza epidemic in
New Zealand. Wellington: Allen & Unwin/Historical
Rognerud, M. A., & Stensvold, I. (1998). Oslo-helsa. Utreding
om helse, miljø og sosial ulikhet i bydelene. Oslo: Oslo
˚l sykehus.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940 939
Schiøtz, C. (1920). Barns legemsutvikling ved de enkelte
Kristianiaskoler. Nordisk hygienisk tidsskrift,5, 289–298.
Scrimshaw, N. S., Taylor, C. E., & Gordon, J. E. (1959).
Interactions of nutrition and infection. American Journal of
the Medical Sciences,March, 367–403.
Singh, G. K., & Siahpush, M. (2001). All-cause and cause-
specific mortality of immigrants and native born in the
United States. American Journal of Public Health,91(3),
Statistisk sentralbyra
˚(SSB). (1917). Dyrtidens virkninger pa
levevilkaarene. 1ste del. NOS VI 105. Kristiania: Aschehoug.
Statistisk sentralbyra
˚(SSB). (1918a). Dyrtidens virkninger pa
levevilkaarene. 2den del. NOS VI 124. Kristiania: Asche-
Statistisk sentralbyra
˚(SSB). (1918b). Lønninger og levevilkaar i
Norge under verdenskrigen. NOS VI 141.
Statistisk sentralbyra
˚(SSB). (1919). Husholdningsforbruket før
og under krigen. Meddelelser fra Det Statistiske Centralbyra
1918,36(12), 129–144 NOS. Kristiania.
Statistisk sentralbyra
˚(SSB). (1920). Undersøkelser over hush-
oldningsforbruket i perioden 27 Januar til 20 april 1919.
Meddelelser fra Det Statistiske Centralbyra
˚,37(11), 164–173
NOS. Kristiania.
Statistisk sentralbyra
˚. (1955). Økonomisk Utsyn 1900–1950
(Economic Survey 1900–1950). Samfunnsøkonomiske Stu-
dier nr. 3, Statistisk sentralbyra
˚, Oslo.
Stevenson, T. H. C. (1921). The incidence of mortality upon the
rich and poor districts of Paris and London. Journal of the
Royal Statistical Society,LXXXIV(Part 1), 90–99.
Sydenstricker, E. (1931). The incidence of influenza among
persons of different economic status during the epidemic of
1918. Public Health Reports,46(4), 154–170.
Tomkins, S. (1992). The failure of expertise: Public health
policy in Britain during the 1918–19 influenza epidemic.
Social History of Medicine,5, 435–454.
Townsend, P., & Davidson, N. (1982). Inequalities in health:
The Black Report. Harmondsworth: Penguin Books.
Vallin, J., Mesle
´, F., & Valkonen, T. (2001). Trends in mortality
and differential mortality. Strasbourg: Council of Europe
van Hartesveldt, F. R. (Ed.). (1992). The 1918–1919 pandemic
of influenza. The urban impact in the Western World. New
York: Edwin Mellen Press.
Vaughan, W. T. (1921). Influenza. An Epidemiological Study.
The American Journal of Hygiene. Monograph Series No.
1. Baltimore: The American Journal of Hygiene.
Waitzman, N. J., & Smith, K. R. (1998). Phantom of the area:
Poverty-area residence and mortality in the United States.
American Journal of Public Health,88(6), 973–976.
Winter, J. M., & Robert, J.-L. (1997). Capital cities at war,
Paris, London, Berlin 1914–1919. Cambridge: Cambridge
University Press.
Zylberman, P. (2003). A holocaust in an holocaust: The
Great War and the 1918 Spanish influenza epidemic in
France. In H. Phillips, & D. Killingray (Eds.), The
Spanish influenza Pandemic of 1918–19. New perspectives
(pp. 191–201). London: Routledge Social History of
Medicine series.
S.-E. Mamelund / Social Science & Medicine 62 (2006) 923–940940
... Kautokeino or Karasjok had outbreaks within the timeframe of the three last waves described in Norway and internationally but did not experience the wave in the spring or summer of 1918 that was prominent in Scandinavian cities and the Eastern seaboard of the USA [19,20]. One-tenth of Norway escaped the summer wave in 1918, and these were usually rural areas without larger cities/towns and major communication routes and networks such as Karasjok and Kautokeino [16]. ...
Full-text available
The 1918–20 pandemic influenza killed 50–100 million people worldwide, but mortality varied by ethnicity and geography. In Norway, areas dominated by Sámi experienced 3–5 times higher mortality than the country’s average. We here use data from burial registers and censuses to calculate all-cause excess mortality by age and wave in two remote Sámi areas of Norway 1918–20. We hypothesise that geographic isolation, less prior exposure to seasonal influenza, and thus less immunity led to higher Indigenous mortality and a different age distribution of mortality (higher mortality for all) than was typical for this pandemic in non-isolated majority populations (higher young adult mortality & sparing of the elderly). Our results show that in the fall of 1918 (Karasjok), winter of 1919 (Kautokeino), and winter of 1920 (Karasjok), young adults had the highest excess mortality, followed by also high excess mortality among the elderly and children. Children did not exhibit excess mortality in the second wave in Karasjok in 1920. It was not the young adults alone who produced the excess mortality in Kautokeino and Karasjok. We conclude that geographic isolation caused higher mortality among the elderly in the first and second waves, and among children in the first wave.
... Ensuring health equity, especially for vulnerable populations in less developed settings with poor health systems is essential for the current COVID-19 pandemic and future global health threats [5][6][7][8][9][10][11]. At communitylevel, it is known that the severity of illness and clinical outcomes can be affected by the concentration of comorbidities in susceptible groups in communities [12][13][14][15], and through disparities in access to health care for preventive measures or prompt diagnosis and treatment [15,16]. ...
Full-text available
Background Ensuring health equity, especially for vulnerable populations in less developed settings with poor health system is essential for the current and future global health threats. This study examined geographical variations of COVID-19 mortality and its association with population health characteristics, health care capacity in responding pandemic, and socio-economic characteristics across 514 districts in Indonesia. Methods This nationwide ecological study included aggregated data of COVID-19 cases and deaths from all 514 districts in Indonesia, recorded in the National COVID-19 Task Force database, during the first two years of the epidemic, from 1 March 2020 to 27 February 2022. The dependent variable was district-level COVID-19 mortality rate per 100,000 populations. The independent variables include district-level COVID-19 incidence rate, population health, health care capacity, and socio-demographics data from government official sources. We used multivariable ordinal logistic regression to examine factors associated with higher mortality rate. Results Of total 5,539,333 reported COVID-19 cases, 148,034 (2.7%) died, and 5,391,299 (97.4%) were recovered. The district-level mortality rate ranged from 0 to 284 deaths per 100,000 populations. The top five districts with the highest mortality rate were Balikpapan (284 deaths per 100,000 populations), Semarang (263), Madiun (254), Magelang (250), and Yogyakarta (247). A higher COVID-19 incidence (coefficient 1.64, 95% CI 1.22 to 1.75), a higher proportion of ≥ 60 years old population (coefficient 0.26, 95% CI 0.06 to 0.46), a higher prevalence of diabetes mellitus (coefficient 0.60, 95% CI 0.37 to 0.84), a lower prevalence of obesity (coefficient -0.32, 95% CI -0.56 to -0.08), a lower number of nurses per population (coefficient -0.27, 95% CI -0.50 to -0.04), a higher number of midwives per population (coefficient 0.32, 95% CI 0.13 to 0.50), and a higher expenditure (coefficient 0.34, 95% CI 0.10 to 0.57) was associated with a higher COVID-19 mortality rate. Conclusion COVID-19 mortality rate in Indonesia was highly heterogeneous and associated with higher COVID-19 incidence, different prevalence of pre-existing comorbidity, healthcare capacity in responding the pandemic, and socio-economic characteristics. This study revealed the need of controlling both COVID-19 and those known comorbidities, health capacity strengthening, and better resource allocation to ensure optimal health outcomes for vulnerable population.
... Regional differences in the spread and impact of smallpox have already been examined by various other historical studies (Duncan et al. 1993;Sköld 1996;Duncan and Duncan 1997;Davenport et al. 2018;Rafferty et al. 2018), but we ask the question how inequalities were expressed in a demographic and spatial sense in one specific city, namely Amsterdam. Other epidemics such the bubonic plague and the Spanish Flu are being studied through the lens of social inequalities (Mamelund 2006;Bengtsson et al. 2018;Alfani 2021;Fourie and Jayes 2021), but smallpox has so far not received such scrutiny. This may in part be due to the idea that smallpox struck lower and higher social classes fairly equally due to its mode of transmission, namely air and close personal contact (Snowden 2019: 97-101). ...
Full-text available
The complex relationship between the history of infectious diseases and social inequalities has recently attracted renewed attention. Smallpox has so far largely escaped this revived scholarly scrutiny, despite its century-long status as one of the deadliest and widespread of all infectious diseases. Literature has demonstrated important differences between rural and urban communities, and between cities, but has so far failed to address intra-urban disparities due to varying living conditions and disease environments. This article examines the last nationwide upsurge of smallpox in the Netherlands through the lens of Amsterdam’s 50 neighborhoods in the period 1870–72. We use a mixed methods approach combining qualitative spatial analysis and OLS regression to investigate which part of the population was affected most by this epidemic in terms of age and sex, geographic distribution across the city, and underlying sociodemographic neighborhood characteristics such as relative wealth, housing density, crude death rate, and birth rate. Our analyses reveal a significant spatial patterning of smallpox mortality that can largely be explained by the existing social environment. Lacking universal vaccination, the smallpox epidemic was not socially neutral, but laid bare some of the deep-seated social and health inequalities across the city.
... Ωστόσο, τα ευρήματα αρκετών μελετών υποδεικνύουν διαφορές ανάμεσα στις κοινωνικο-οικονομικές ομάδες αναφορικά με τη συχνότητα και τη σοβαρότητα οξέων λοιμώξεων. Εν προκειμένω, οι αναλύσεις της πανδημίας της ισπανικής γρίπης του 1918 -1919 και της επιδημίας γρίπης του 2009 αναδεικνύουν ότι οι κοινωνικο-οικονομικά κατώτερες ομάδες διέτρεχαν μεγαλύτερο κίνδυνο νόσησης και θανάτου σε αντίθεση με τις κοινωνικο -οικονομικά πιο εύπορες ομάδες (Bengtsson et al., 2018, Mamelund, 2006, Rutter et al., 2012. Ωστόσο, τέτοια κοινωνικά επιδημιολογικά πρότυπα δεν είναι ούτε σταθερά ούτε παρόμοια για κάθε συγκείμενο και μπορεί να ποικίλλουν και να μετασχηματίζονται κατά τη διάρκεια μιας επιδημίας (Mamelund, 2018). ...
Full-text available
Η πανδημία της COVID-19 τα τελευταία δυόμιση έτη μαστίζει την ανθρωπότητα, δημιουργώντας τεράστιες πιέσεις στα υγειονομικά συστήματα, στα κοινωνικά κράτη και στις οικονομίες. Ένα από τα σημαντικά ζητήματα, το οποίο δεν έχει μελετηθεί επαρκώς, είναι οι επιπτώσεις που δημιουργούνται από τις κοινωνικές ανισότητες που ενυπάρχουν στα επιμέρους κοινωνικο-οικονομικά συγκείμενα, καθώς και από τις νέες που δημιουργήθηκαν. Η σχέση πανδημίας και κοινωνικών ανισοτήτων δεν είναι ελληνικό φαινόμενο αλλά παρατηρείται σε πολλές χώρες, ιδίως σε εκείνες με ελλείμματα στην οργανωμένη κοινωνική προστασία. Στόχος της συγκεκριμένης μελέτης είναι να σκιαγραφήσει, με τη χρήση συγκεκριμένων δευτερογενών πηγών και δεδομένων, την σχέση κοινωνικών ανισοτήτων και πανδημικής κρίσης σε μικροεπίπεδο και συνακόλουθα, να θέσει το ζήτημα της παγκόσμιας ανισότητας ως μια από τις βασικές συνισταμένες για την επίταση των κοινωνικών προβλημάτων τόσο στις πιο φτωχές χώρες όσο και στις πιο ευάλωτες κοινωνικο-οικονομικές ομάδες (ένα παράδειγμα είναι η αποτυχία ή αργοπορία σχεδίου COVAX). Λέξεις κλειδιά: Κοινωνική ανισότητα, covid-19, πανδημία, κοινωνικό κράτος Παραπομπή ως : Τζαγκαράκης, Σ. Ι., Μελίδης, Μ., Κρήτας Δ., (2022). Κοινωνικές Ανισότητες και COVID-19: Μια επισκόπηση της κοινωνικής διάστασης της πανδημίας (σελ. 86-94). Κεφ. στο Πανταζής, Σ., Μαράκη, Ε., Μπελαδάκης, Δ., κ.α. (επιμ.). Κοινωνία, Εκπαίδευση και Πολιτική Σχέσεις και Ανασχέσεις. Ηράκλειο: Ινστιτούτο Ανθρωπιστικών και Κοινωνικών Επιστημών
... 139 Even historical research into the 1918 Spanish influenza pandemic has documented area-level inequalities in mortality related to deprivation (eg, house hold size and income). [140][141][142] Similarly, our results reflect the findings of reviews of socioeconomic inequalities in COVID-19 at the individual level. For example, an inter national systematic review of inequalities in COVID-19 outcomes found that ethnic minorities and low socioe conomic groups had high risks of COVID-19 infection, hospitalisation, confirmed diagnosis, and death. ...
Full-text available
COVID-19 has exacerbated endemic health inequalities resulting in a syndemic pandemic of higher mortality and morbidity rates among the most socially disadvantaged. We did a scoping review to identify and synthesise published evidence on geographical inequalities in COVID-19 mortality rates globally. We included peer-reviewed studies, from any country, written in English that showed any area-level (eg, neighbourhood, town, city, municipality, or region) inequalities in mortality by socioeconomic deprivation (ie, measured via indices of multiple deprivation: the percentage of people living in poverty or proxy factors including the Gini coefficient, employment rates, or housing tenure). 95 papers from five WHO global regions were included in the final synthesis. A large majority of the studies (n=86) found that COVID-19 mortality rates were higher in areas of socioeconomic disadvantage than in affluent areas. The subsequent discussion reflects on how the unequal nature of the pandemic has resulted from a syndemic of COVID-19 and endemic inequalities in chronic disease burden.
There is a growing concern that inequalities are hindering health outcomes. This paper's primary objective is to investigate the role of relative deprivation and inequality in explaining the daily spread of the Covid-19 pandemic. For this purpose, we use secondary cross-sectional data across 119 (developed and developing) countries from January 2020 – to April 2021. For the empirical analysis, we use a recent dynamic panel data modelling approach that allows us to identify the role of time-invariant variables such as degree of globalisation, political freedom and income inequality on the dynamics of the pandemic and fatality rates across countries. We find that new cases per million and fatality rates are highly persistent processes. After controlling for time-varying mobility statistics from the Google mobility database and region-specific dummy variables, the two significant factors that explain the severity of Covid-19 spread in a country are per-capita Gross Domestic Product (GDP) and Yitzhaki's relative income deprivation index. Lagged value of new cases per million significantly explains cross-country variations in the daily case fatality rates. A higher proportion of the older population and pollution increased fatality rates while better medical infrastructure reduced it.
Life courses can be perceived as interplays between different trajectories, such as education, work, and reproduction, spanning the different life phases of an individual. Along these paths, an individual finds different life course transitions, such as the entrance to the workforce, the establishment of a relationship, or the birth of children. The age and the historical time in which the transitions take place are also important. Transitions can, in turn, be influenced by agency, social relations, coincidence, geographical locations, and historical conditions. These life course elements are elaborated on and illustrated with evidence from relevant research.
Research on the 1918 influenza pandemic often focuses exclusively on pandemic years, reducing the potential long‐term insights about the pandemic. It is critical to frame the 1918 pandemic within the underlying population dynamics, health, and sociocultural context to understand what factors contributed to pandemic mortality and survivorship, with respect to observed inequality, and consequences of the pandemic. Individual death records and censuses from The Rooms Provincial Archives and Memorial University of Newfoundland Digital Archives for three major causes of death—influenza and pneumonia; tuberculosis; and pooled bronchitis, measles, and whooping cough—were collected for three periods in the early 20th century: pre‐pandemic (1909–11), pandemic (March 1918–Janaury 1919), and post‐pandemic (1933–1935). We calculated pooled age‐standardized mortality rates and changes in pre‐ to post‐pandemic mortality rates by region. We fit Kaplan–Meier and Cox proportional hazards models to each period, controlling for age, cause of death, and region. Pandemic mortality was higher than that of pre‐ and post‐pandemic periods. Post‐pandemic mortality was significantly lower than pre‐pandemic mortality in all regions, except Western Newfoundland. Survival was lowest during the pandemic and increased significantly post‐pandemic (p < 0.0001), with no significant differences among regions during the pandemic (p = 0.32). Significant differences in survivorship in 1933–1935 were driven by increasing differences in survivorship for P&I among the regions more than other causes of death. Myopic perspectives of pandemics can obscure our understanding of observed outcomes. Inequalities in respiratory disease mortality are evident in pre‐ and post‐pandemic periods, but these would have been missed in investigations of the pandemic period alone.
Full-text available
Objective The COVID-19 pandemic has exacerbated existing health disparities. To provide a historical perspective on health disparities for pandemic acute respiratory viruses, we conducted a scoping review of the public health literature of health disparities in influenza outcomes during the 1918, 1957, 1968, and 2009 influenza pandemics. Methods We searched for articles examining socioeconomic or racial/ethnic disparities in any population, examining any influenza-related outcome (e.g., incidence, hospitalizations, mortality), during the 1918, 1957, 1968, and 2009 influenza pandemics. We conducted a structured search of English-written articles in PubMed supplemented by a snowball of articles meeting inclusion criteria. Results A total of 29 articles met inclusion criteria, all but one focusing exclusively on the 1918 or 2009 pandemics. Individuals of low socioeconomic status, or living in low socioeconomic status areas, experienced higher incidence, hospitalizations, and mortality in the 1918 and 2009 pandemics. There were conflicting results regarding racial/ethnic disparities during the 1918 pandemic, with differences in magnitude and direction by outcome, potentially due to issues in data quality by race/ethnicity. Racial/ethnic minorities had generally higher incidence, mortality, and hospitalization rates in the 1957 and 2009 pandemics. Conclusion Individuals of low socioeconomic status and racial/ethnic minorities have historically experienced worse influenza outcomes during pandemics. These historical patterns can inform current research to understand disparities in the ongoing COVID-19 pandemic and future pandemics.
Digitisation of this thesis was sponsored by Arcadia Fund, a charitable fund of Lisbet Rausing and Peter Baldwin.
Two years after the First World War ended there was a surge in European birth rates, including 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 influenza 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 superior quality in a European context in that registration of population data, including vital statistics, continued normally in Norway undisturbed by the war.
Howard Philips and David Killingray Introduction Part I: Virological and Pathological Perspectives 1. Edwin D. Kilbourne A Virologist's Perspective on the 1918-1919 Pandemic 2. Jeffery K. Taubenberger Genetic Characterisation of the 1918 'Spanish' Influenza Virus Part II: Contemporary Medical and Nursing Perspectives 3. Wilfried Witte The Plague That was Not Allowed to Happen: German Medicine and the Influenza Epidemic of 1918-1919 in Baden 4. Nancy K. Bristow 'You Can't Do Anything for Influenza': Doctors, Nurses and the Power of Gender During the Influenza Epidemic in the United States Part III: Official Responses to the Pandemic 5. Geoffrey W. Rice Japan and New Zealand in the 1918 Influenza Pandemic: Comparative Perspectives on Offical Responses and Crisis Management 6. Mridula Ramanna Coping with the Pandemic: The Bombay Experience Part IV: The Demographic Impact 7. Wataru Iijima Spanish Influenza in China, 1918-1920 8. Kevin McCracken and Peter Curson Flu Downunder: A Demographic and Geographic Analysis of the 1919 Pandemic in Sydney, Australia 9. N. P. A. S. Johnson The Overshadowed Killer: Influenza in Britain in 1918-1919 10. D. Ann Herring and Lisa Sattenspiel Death in Winter: Spanish Flu in the Canadian Subarctic 11. Beatriz Echeverri Spanish Influenza seen from Spain 12. Patrick Zylberman A Holocaust in a Holocaust: The Great War and the 1918 'Spanish' Influenza Epidemic in France 13. Andrew Noymer and Michel Garenne Long-Term Effects of the 1918 'Spanish' Influenza Epidemic on Sex Differentials of Mortality in the USA: Exploratory Findings from Historical Data Part V: Long-Term Consequences and Memories 14. James G. Ellison 'A Fierce Hunger': Tracing Impacts of the 1918-1919 Influenza Pandemic in Southwest Tanzania 15. Myron Echenberg 'The Dog that Did Not Bark': Memory and the 1918 Influenza Epidemic in Senegal Part VI: Epidemiological Lessons of the Pandemic 16. Stephen C. Schoenbaum Tranmission of and Protections against Influenza: Epidemiological Observations Beginning with the 1918 Pandemic and the Implications