DISASTERS AND COMMUNITY RESILIENCE: SPANISH FLU AND THE FORMATION
OF RETAIL COOPERATIVES IN NORWAY
Graduate School of Business
Stanford, CA 94305-5015
HENRICH R. GREVE
1 Ayer Rajah Avenue
We are grateful for comments from Peter Bearman, Saum Jha, Jerry Kim, Kinga Makovi, Grant
Miller, and Damon Phillips. The Norwegian Cooperative Association and Vidar Øverland of the
National Archive provided the data on the cooperative founding dates. Census data are provided
by the Municipality Database of the Norwegian Social Science Data Service (NSSD). NSSD is
not responsible for the analysis or interpretation given here.
Forthcoming in Academy of Management Journal
DISASTERS AND COMMUNITY RESILIENCE: SPANISH FLU AND THE FORMATION
OF RETAIL COOPERATIVES IN NORWAY
Why are some communities resilient in the face of disasters, and why are others unable to
recover? We suggest that two mechanisms matter: the framing of the cause of the disaster, and the
community civic capacity to form diverse non-profits. We propose that disasters that are
attributed to other community members weaken cooperation and reduce the formation of new
cooperatives that serve the community, unlike disasters attributed to chance or to nature, which
strengthen cooperation and increase the creation of cooperatives. We analyze the Spanish Flu, a
contagious disease that was attributed to infected individuals, and compare it with spring frost,
which damaged crops and was attributed to nature. Our measure of resilience is whether the
community members could form retail cooperatives – non-profit community organizations. We
find that communities hit by the Spanish Flu during the period 1918-1919 were unable to form
new retail cooperatives in the short and long run after the epidemic, but this effect was reduced
over time and countered by civic capacity. Implications for research on disasters and institutional
legacies are outlined.
Keywords: Disasters; Community Resilience; Institutional Legacies
From 2005 to 2015, there was an annual average of 367 disasters, and a total of 76,424
deaths and 173,241,621 affected people (Center for Research on the Epidemiology of Disasters).
A key question in the study of disasters is why some communities are more resilient than others
(van der Vegt, Essens, Wahlström, & George, 2015). A disaster is a “potentially traumatic event
that is collectively experienced, has an acute onset, and is time delimited; disasters may be
attributed to natural, technological, or human causes” (McFarlane & Norris, 2006: 4). Although
resilience is commonly understood as an individual’s ability to bounce back from a traumatic
setback, disaster researchers emphasize social resilience – the capability of a community to
withstand and recover from a disaster. In an early but prescient article, Carr (1932: 207-210)
observed that “Not every windstorm, earth-tremor, or rush of water is a catastrophe…It is the
collapse of the cultural protections that constitutes the disaster proper.”
Social resilience after a disaster is of theoretical import for organizational researchers
seeking to study institutional legacies – institutions that affect communities over time. To date,
the literature on institutional legacies has shown how path dependent events create legacies, and
how they are perpetuated through organization building (see Greve & Rao, 2014 for a review).
However, path dependent events can also erode pre-existing institutional legacies. Nonetheless,
this has received very little attention in organizational research. Indeed, there is a parallel gap in
the more applied literature on disasters. While there is evidence that sociable communities, places
where members are willing to help one another, are more resilient and ride out heat waves,
extreme cold, or hurricanes and tsunamis more easily than socially impoverished communities
(see Aldrich, 2012; Kleinberg, 2003; Sampson, 2012 for reviews), there has been little attention
devoted to understanding how disasters themselves reduce sociability. More specifically, the
literature on disasters is subject to three important limitations.
First, reported research on the impact of disasters overlooks how causal attributions of a
disaster shape a community’s response to it. The literature has tended to treat disasters as the
stimulus and relief operations as the response, without paying attention to how the interpretation
and framing of a disaster underlie resilience and response. Just as social movements have frames
(Snow & Benford, 1988), disasters too have frames that help community members interpret the
disaster. Like social movement frames, disaster frames hinge on the nature of attributions and
blame. Since the early work by Kelley (1971) and Miller and Ross (1975), it has been well
understood that individuals trace outcomes such as a disaster to an act of nature, or to other
people. In fact, the label ‘natural disaster’ does not necessarily mean that that the disaster is
attributed to nature, it could be interpreted as being man-made. Quarantelli and Dynes (1977)
suggest that disasters are not physical events but are socially constructed instead. For example,
Blocker and Sherkat (1992) found that a flash flood in Tulsa was interpreted by residents as the
outcome of poor management by the Army Corps of Engineers rather than a result of freak
rainfall levels. An advantage of emphasizing causal attributions of the source of a disaster in a
community is that it frees the researcher from the need to make a priori classification of disasters.
A second limitation is that accounts of community resilience in the face of disasters
emphasize the number of non-profits within a community as a source of resilience rather than the
civic capacity to create non-profits. Since Putnam, Leonardi, and Nanetti (1993), disaster
researchers have emphasized how interpersonal ties in the affected communities cushion them
from the adverse impact of disasters. However, as Sampson, MacIndoe, McAdam, and
Weffer-Elizondo (2005) point out, although many non-profits in a community can matter because
they can facilitate social contact and subsequent mobilization, what is more important is the
extent to which the community has the capacity to create a diverse stock of non-profits that can
hedge against unexpected events.
Finally, while there is an abundance of rich case studies that chronicle disaster relief
operations in communities (e.g., Quarantelli, 2005), there is a dearth of quantitative research on
the adaptive capacity of communities – their ability to bounce back to pre-disaster levels, or to do
better. Taylor (1978: 276) lamented that “work in the field of disaster studies needs some exercise
of the sociological imagination in the use and development of research techniques and
procedures.” More recently, van der Vegt et al. (2015: 971-977) urge organizational researchers
to embrace the ‘grand challenge’ of studying the role of organizations in natural disasters, and
recommend the use of time-series designs, wherein longitudinal effects of disruptions can be
better understood in the field than in laboratory experiments.
These considerations supply the motivation for our study. We suggest that when a disaster
frame attributes harm to other community members, suspicion and distrust set in and undermine
resilience. This contrasts with a disaster frame which makes attributions to nature. In such a case,
the sense of shared fate is created which enhances cooperation and resilience. Contagious
diseases are a very important instance of a disaster that undermines community resilience, and are
the focus of our research. As a contrast, we examine weather shocks as disaster attributed to
nature. We argue that contagious disease outbreaks that are transmitted person to person are
attributed to other individuals in the community. This lowers within-group social integration,
fosters distrust, and creates cooperation deficits that lower resilience. Weather shocks have the
usual effect of disasters documented in the literature: they create a sense of shared fate within
groups and lead to risk-sharing arrangements (Campbell, 1958) and solidarity (Erikson, 1994).
Empirically, we study the Spanish Flu in Norway, and our indicator of resilience is whether
community members are able to come together to form retail cooperatives after the exogenous
shocks the Spanish Flu created. Our contrast case is spring frost. Retail cooperatives were general
grocery stores organized on the basis of principles enshrined in the cooperative movement, and
were a form of mutual help and risk-pooling in communities. Their founding and operation are a
classic collective action problem (i.e., Olson, 1965) because the costs are borne by volunteers,
while the benefit accrues to the community at large, as is the case of many other social ventures.
COMMUNITY RESILIENCE: FRAMES AND CIVIC CAPACITY
In an early study, Meyer (1982: 520) used the term ‘resiliency’ to label an organization’s
ability to respond to an adverse event (a strike in this case). Wildavsky (1988: 77) described
resilience as the “capacity to cope with unanticipated dangers after they have become manifest,
learning to bounce back”. Sutcliffe and Vogus (2003) observe that two requirements to test
resilience in communities are a) exposure to threat or an adverse event and b) achievement of
positive adaptation. This means that researchers ought to define threats or adverse events clearly
and delineate positive adaptation. One indicator of resilience is the ability to build new
cooperatives to solve a community’s problem not just immediately after a disaster, but long after
it. Can a community return to pre-disaster rates of building new organizations or improve on it
after a disaster? Below, we argue that disaster frames and pre-existing civic capacity are the
well-springs of resilience in communities.
Although it is clear that resilient organizations and communities respond better to disasters
than their less resilient counterparts, what has received very little attention is whether
communities gain or lose resilience through how the disaster is framed. Erikson (1994) classified
disasters into man-made and natural disasters, and contended that man-made disasters led to
distrust of existing arrangements, whereas natural disasters strengthened cooperation in human
communities. However, Erikson (1994) did not pay attention to disaster framing per se by
members of the community. Even if based on a priori classification, his conjecture underlines the
link between causal attribution of the source of a disaster and the long-term resilience of
Bateson (1956) described a frame as a meta-communicative device that defines the
parameters to understand events. Goffman (1974) developed frame analysis to show how people
interact with each other and their environments by making inferences. In the social movement
literature, Gamson, Fireman, and Rytina (1982) described how ‘injustice frames’ were essential
for people to become motivated to protest against unjust uses of authority by government
officials. Frames "enable individuals to locate, perceive, identify and label occurrences" (Snow,
Rochford, Worden, & Benford, 1986: 464); and "selectively punctuate and encode objects,
situations, events, experiences and sequences of actions within one's present and past
environment" (Snow & Benford, 1992: 137). If frames are mental schemes or assemblages,
framing is the process by which individuals make sense of their world. For social movements,
Snow et al. (1986) argued that what matters in frames are two sets of attributions: diagnostic
attributions and prognostic attributions. Diagnostic attributions place blame or responsibility by
answering the questions “What went wrong? Who or what is responsible?” Prognostic framing
addresses the question of “What do we do? What must be done?” Diagnostic frames naturally are
antecedents of prognostic frames.
Just as movements have frames that are amplified through media exposure, disasters too can
be framed through attribution by the members of a community. Such attributions may be
generated in informal conversation and spread through media coverage to other individuals.
Indeed, a long line of research suggests that individuals are ‘naïve’ psychologists who attribute
outcomes to the environment, to themselves, or to other people (Kelley, 1971; Miller & Ross,
1975). The perception of responsibility for disasters and accidents not only shapes emotions that
are aroused, but also profoundly influences how people respond. Thus, an early study by Veltfort
and Lee (1943) reported that citizens of Boston spent time and effort to find the responsible
parties for a night club fire that killed nearly 500 people, but did not acknowledge that such
disasters may occur by chance.
Attribution activity in framing a disaster as manmade or a chance outcome has to do with
self-protection motives. As Walster (1966: 74) noted “If a serious accident is seen as the
consequence of an unpredictable set of circumstances, beyond anyone's control or anticipation, a
person is forced to concede the catastrophe could happen to him. If, however, he decides that the
event was a predictable, controllable one, if he decides that someone was responsible for the
unpleasant event, he should feel somewhat more able to avert such a disaster.” Shaver (1970)
suggested that what matters was situational and personal similarity to the victim. When
individuals are situationally similar to the victim (e.g., engaged in the same activity), they are
likely to deny personal similarity, blame the victim and conclude that they would have made
different choices. In contrast, when they are personally similar to the victim, they are likely to
attribute the event to chance.
Parker, Brewer, and Spencer (1980) studied a bushfire in Southern California and treated it
as a ‘natural randomization process’. Among the homes close to the height of the fire, chance
factors such as wind direction, traffic congestion and other factors shaped which homes were
burnt and which were not. Victims and those not victimized were equivalent in this natural
experiment, but Parker et al. (1980) found that those whose houses were burnt attributed it more
to bad luck and less to effort compared to those whose houses did not burn. But individuals in
communities can attribute a natural event to people regardless of whether they were victimized or
Irrespective of the individual motivations underlying attribution, these attributions and
disaster framing influences helping behaviors. Zagefka, Noor, Brown, de Moura, and Hopthrow
(2011) found that when presented with disasters with clearly natural causes and disasters with
human fault, potential donors were more likely to donate to the naturally caused disaster and less
to manmade disasters. Below, we describe how contagious diseases are likely to be attributed to
infected individuals, and undermine cooperation, and how weather shocks are likely to be
attributed to chance or to God, and reinforce feelings of solidarity and foster resilience.
Contagion Frame. Diverse strands of work in social psychology, history, and sociology
underlie the argument that contagious disease outbreaks lead to blaming either the infected or
another social group as the culprit. Evolutionary psychologists have argued that contagious
diseases evoke xenophobia as a threat management mechanism. Foreigners are seen as carriers of
pathogens and ignorant of local norms that prevent the spread of diseases. But contagious
diseases not only create out-group antagonism, they also weaken the bonds of social integration
within a group or community (e.g., Crandall & Schaller, 2005; Schaller, 2011; Schaller &
Neuberg, 2012). Moral psychologists postulate that disgust avoidance is a universal motive
(Haidt, McCauley, & Rozin, 1994) and build on the ‘law of contagion’, wherein mere contact is
enough to transfer the essence of a contagious object or individual to others (Rozin & Nemeroff,
1990). For example, aversion to AIDS, as measured by responses to wearing sweaters worn
previously by a (hypothetical) person with AIDS, was found to be a composite of aversion to
strangers, infection, and criminals (Rozin, Markwith, & McCauley, 1994). The transfer of
impurity is categorical rather than ordinal (Haidt & Algoe, 2004), and the impurity survives even
after the source of contamination has been removed (Frazer, 1911; Haidt, Rozin, McCauley, &
Colonial accounts of infectious diseases underline how diseases produce cooperation
deficits by fostering anxiety, rumor, suspicion, and eventually panic (Peckham, 2015). Thus,
Carroll (2015) documents how the onset of bubonic plague in 1894 in Canton triggered draconian
sanitary interventions by the government, and precipitated anti-foreigner rumors and sentiments
in Canton among the Chinese. In societies where there are different ethnic populations, minority
groups may be stigmatized in the disaster framing. Also, in ethnically homogenous societies, the
infected members of the communities may be framed by the media, government and community
members as the source of the disease, which could result in social disintegration and loss of
cooperation. A noteworthy feature of contagious disease outbreaks is that they lead to the ‘fear of
the other’, thereby adding a contagion of fear to the spread of the disease. Das (2007: 119) notes
that the rumors are the motors of suspicion and avers that “language becomes communicable,
infectious, causing things to happen almost as if they happened in nature.” Recently, Dutta and
Rao (2015) found that mere exposure to discourse on contagious flu led individuals to oppose the
legalization of illegal immigrants. In addition to discourse itself, actual mortality from a
contagious disease generates fear, weakens social connections, and leads to anger against
authorities, self-isolation, and loss of trust in institutional arrangements (Edelstein, 1988; Picou,
Marshall, & Gill, 2004). Therefore:
Hypothesis 1. The greater the mortality due to a contagious disease outbreak in a
community, the less likely are foundings of cooperatives in the short-term.
A fast-growing body of research suggests that contagious disease outbreaks can have
persistent long-term effects on economy and society. More generally, acts of nature generate
institutional legacies that reshape community capacity and motivation for cooperative behaviors
(Greve & Rao, 2014). For example, Acemoglu, Jackson and Robinson (2012) suggested that
European colonizers confronted a choice of whether to settle in the country or exploit it from afar,
and decided this on the basis of diseases. They migrated to healthier places and created inclusive
political institutions with strong property rights that led to economic growth. By contrast, when
settler mortality due to disease was high, they relied on extractive institutions that undermined
economic development (see Acemoglu & Robinson, 2012). Similarly, Voigtländer and Voth
(2012) note that Jews were blamed for the Black Plague, and use plague-era pogroms to predict
anti-Semitic violence six centuries later in 1920’s Germany. The argument is that anti-Semitic
pogroms in the 1300’s created a residue of prejudices that underpinned the tragedy of the 1920’s
and 1930’s in Germany, and continued up to World War II.
These arguments suggest that contagious disease outbreaks are likely to trigger not only
fear of the other (infected individuals) but also generates epistemic anxiety that becomes
incorporated into the local culture of a community. Erikson (1994: 148) observes that toxic
disasters “never end…An all clear is never sounded. The book of accounts is never closed.”
Molotch, Freudenburg, and Paulsen (2000: 794) suggest that a community comprises a local
culture that is drawn upon by citizens to think and feel how their community ought to be
organized and propose that they “harbor memory traces…through which something like a social
structure can transpose itself from one time to the next or one institutional realm to the next.” It is
almost as though biological contamination creates corrosive communities (Freudenburg, 1997).
Hypothesis 2. The greater the mortality due to a contagious disease outbreak in a
community, the less likely are foundings of cooperatives in the long-term.
The central problem with the community corrosion created by a disaster with a contagion
frame is the lowered interpersonal trust. Community corrosion weakens the community’s
networks and hampers the mobilizing efforts needed to form any organization (McEvily, Perrone,
& Zaheer, 2003), and that are especially complex when seeking to found a cooperative, which are
created to solve a collective action problem. Central steps in rebuilding trust are reconstitution of
network ties and recreation of forums in the community. Reconstitution of ties was addressed
earlier in the literature on accidentally broken interlocks, which found that new ties were made,
but did not closely resemble the original network structure (e.g., Palmer, 1983; Palmer, Friedland,
& Singh, 1986). Similarly, disaster research has found slow reconstitution of intra-community
ties can occur in disasters attributed to nature (Elliott, Haney, & Sams-Abiodun, 2010). Against
these negative findings, work has shown that tie formation is a common occurrence in
communities, so we should expect that communication and trust building will occur over time.
Associations have been examined in the literature on institutional legacies, wherein
voluntary organizations are seen as meeting places that enable people to come together to realize
collective goals (Greve & Rao, 2014). A key process here is that voluntary organization founding
is a path-dependent and recursive process, where early foundings build organizational templates,
train founders and workers, and produce success stories that help later founding efforts (Greve &
Rao, 2012; Iversen & Soskice, 2009). These effects can also derive from political movements that
have a component of collective action or specific organizational forms seen as valuable
(Haveman, Rao, & Paruchuri, 2007; Schneiberg, King, & Smith, 2008). Although network tie
reconstitution and creation of associations both have random and path-dependent components,
they should over time combine to reduce the effect of any one specific disaster. Thus we
Hypothesis 3. The long-term effect of the mortality due to a contagious disease outbreak in
a community on foundings of cooperatives is weaker than the short-term effect.
An important reason why resilient communities adapt to disasters better than their less
resilient counterparts is that a diverse non-profit sector leaves the community with significant
civic capacity, including experienced founders and workers, dense social connections, and trust
of others. Wildavsky (1988: 70) argues, for example, that to be resilient is to be vitally prepared
for adversity which requires “improvement in overall capability, i.e., a generalized capacity to
investigate, to learn, and to act, without knowing in advance what one will be called to act upon.”
One dimension of this mechanism is how a more diverse stock of non-profits in a
community impacts its ability to acquire and process information, learn, and match volunteers to
social problems. Organizational diversity allows people in a community to meet each other in a
variety of venues, which lowers the cost of mobilizing volunteers. For instance, after the Indian
Ocean tsunami affected southern India in 2004, Aldrich (2011) compared six villages in Southern
India in the same district, and found that those with fishermen caste councils and parish councils
did better than others in securing aid and mobilizing efforts. Similarly, Olshansky, Johnson, and
Topping (2006) document how “machizukuri” (town-building) organizations played a vital role in
Kobe after the earthquake of 1995, by organizing temporary housing, parking, and even a local
currency to help retailers. Kleinberg (2003) studied the effect of a silent disaster – the 1995 heat
wave in Chicago which had a death toll comparable to that of Hurricane Katrina. This study
showed that a Latino neighborhood had far fewer deaths than an adjacent African-American
neighborhood because the sheer variety of neighborhood organizations created overlapping
networks that allowed people to check in on the elderly. In each of these cases, a disaster had
reduced short-term effects as a result of the non-profit organization diversity.
The theory behind these findings can be adapted to our research question because we are
primarily interested in a contagious disease’s long term effects on the community, although
short-term mobilization also matters. Contagious diseases weaken social bonds, as they come with
disaster frames that perceive others as potential sources of harm, which can occur even if a
community has many pre-existing non-profit organizations. After all, it is natural to avoid the
various voluntary organizations in a community when individuals seek to protect themselves from
contagious disease. This is problematic because voluntary organizations are not just meeting
places; they enhance a community’s civic capacity. Indeed, communities with high civic capacity
are especially adaptive and able to form organizations to handle many kinds of community needs
(Sampson et al., 2005). For a community that faces an unusual event such as a serious contagious
disease, this ability to organize and re-organize in response is especially important. Thus we
Hypothesis 4. The greater the prior civic capacity of a community, the lower is the effect of
mortality due to a contagious disease outbreak on its ability to build cooperatives in the
short-term and the long-term.
SPANISH FLU IN NORWAY
The Spanish Flu is the most widely studied contagious disease because of its overall impact,
as it led to more than 20 million deaths worldwide (Taubenberger, Reid, Krafft, Bijwaard, &
Fanning, 1997). Its characteristic W-shaped mortality curve over age with a peak around 30 years
of age constituted a demographic shock that removed adults in prime working age from their
families and communities. We focused on Norway for a number of reasons. First, public health
records show that the Spanish Flu infected 1.2 million Norwegians, and 15,000 deaths were
recorded in a population of 2.6 million, leading to the death of 0.6 percent of the population
(Mamelund 1998).1 If the Spanish Flu had decimated a large percentage of the population,
adverse effects on organization building would have been tragic but entirely understandable.
After all, communities that lose a large fraction of their people simply have no one to build new
organizations. The Spanish Flu in Norway is interesting precisely because it removed no more
than 0.6 percent of the population. Second, there are high quality data on flu mortality in
Norwegian archives, and we also have high quality data on cooperatives and other community
organizations. As an exogenous event, the Spanish Flu assigned communities into categories with
different exposure to flu mortality. Although communities on the coast were more likely than
communities inland to get the Spanish Flu, the order in which towns got it and its severity was
The origins of the Spanish Flu in the 20th century have been widely debated in the literature.
Oxford et al. (2005) trace the onset of the Spanish Flu to a British military base in Etaples in the
north of France. This base was occupied by 100,000 soldiers, covered 12 square kilometers, and
was located next to sea marshes with migratory birds. The congestion of individuals and animals
and extensive use of gases in World War I that were mutagenic might have been the incubator,
and many soldiers were admitted between December 1916 and March 1917 for acute respiratory
infection. Alternatively, Frémeaux (2006) notes that 50,000 soldiers from Annam (Vietnam, Laos,
and Cambodia) were in France, and experienced Annamite pneumonia. In the U.S., writers trace
it to an outbreak at the Fort Riley military base in Kansas (Vaughan, 1921). These narratives
suggest a relationship between war and the Spanish Flu, implying that soldiers were the conduits
1This is equal to the annual malaria death rate of Gabon, and is six times the annual gun death rate of the
US, but differs from both of these in being an anomalous spark of deaths rather than a chronic death rate.
for the transmission of the virus.
The disease got its name because it was first reported as widespread in the general
population in Spain, though the earlier reporting there was because Spain had freer press than the
nations participating in WWI. The Madrid newspaper, El Sol, published the first headline on May
22, 1918 and noted that the disease observed amongst the public was similar to those observed in
prisons; during May and June 1918, there were 275 influenza related deaths, thereby, leading to a
rate of 0.42/1000 inhabitants (Erkoreka, 2009). Even if Spain was neutral in WWI, it allowed
Portuguese troops, workers, and merchandize to have passage towards France, near Irun and
Hendai, and this region may well have been the place where the virus mutated before spreading
to Europe (Erkoreka, 2009). Mortality rates varied by country, but were estimated at 1.1 percent
overall (Ansart et al., 2009).
Spanish Flu Contagion Frame
Norway too was a neutral country in World War 1. Nonetheless, the Norwegian Navy was
mobilized, and the first cases may have been in the coastal areas. Indeed, in earlier outbreaks of
diseases such as radesyge, a skin disease in Norway that afflicted peasants, the origin story that
was most popular was about a Russian naval ship near Stavanger (Lie, 2007). The Spanish flu
affected Norway in three waves: the first was in the summer of 1918, the second occurred later in
the year, and the final wave was in 1919. It was later that epidemiologists saw the three waves as
similar, and the pandemic was treated as a single cataclysmic event.
Initially, in Europe, newspapers depicted the Spanish Flu as a case of the “chills” and as the
flu became more virulent, it was depicted as a plague or treated as an enemy. The pattern of news
coverage in England is illustrative. For example, in the first week of June, the Daily Mail in
England had a headline titled “Is Influenza Coming” and advised its readers that it was no worse
than a cold, but urged readers to prepare their ‘defenses.’” By late July 1918, it was described by
the Salford Reporter in England as a ‘new foe’ and chemists reported panic buying of quinine and
other medicines. By late, October, the Daily Mail and Times increased their coverage of flu,
reporting the deaths of firemen, policemen, and even young girls working in factories. The Times
intoned that “Fear is the mother of infection… The alarmists and defeatists are the allies of the
epidemic.” English newspapers also carried the exhortation of Arthur Newsholme, the Chief
Medical Officer of the Local Government Board, who urged citizens to regulate their behavior by
not ‘carrying on’ and exposing others. The Times succinctly said “Dr. Newsholme emphasizes
that the control over the disease can be secured by the active cooperation of each member of the
community.” Yet, there was anxiety and fear among the English public prompting the Manchester
Guardian to observe that people were “fighting shy of theatres, kinemas (sic) and all kinds of
In Norway, the Spanish Flu induced Edvard Munch to paint his celebrated self-portraits
suffering the flu and after recovery from the flu. The persistent printing of death notices in
newspapers created gloom. One woman recalled that “Aftenposten [a major newspaper of the
capital] was full of death notices. I read them all every day, and it was awful.” At the peak of the
Spanish Flu mortality, the health authority responded to the many death notices in newspapers by
issuing a statement urging people not to panic. Another woman remembered that “everyone was
afraid of everyone else, the contagion was everywhere” (Mamelund, 1998). The strong fear of
contagion from others was consistent with popular understanding of how diseases spread. For
example, tuberculosis was still a wide-spread disease, and was (correctly) seen as contagious
person to person. Norway was the first nation to enact a Tuberculosis Law that allowed forcible
isolation of the infected (in 1900), and it had built tuberculosis sanatoriums, away from towns, to
treat patients and isolate them from others.
Consistent with the populace fear of contagion from others, countermeasures against the
Spanish flu included closures of theatres and public assembly places (Borza, 2001). Health
authorities banned public meetings, closed churches and schools, and advised people to cover
their mouths and isolate themselves. Borza (2001) also implies that there was variability in the
enforcement of these countermeasures, in part because some of the medical establishment saw
the epidemic as unstoppable, and preferred letting it run its course before winter set in and
increased mortality in the affected. It was well known that winter flu was the most deadly.
Norway was a comparatively low-mortality nation because the Spanish Flu removed only
0.6 percent of the population. While many of the deaths were in the 25-40 age group,
communities did not necessarily lose many individuals who were in the most likely age for
community organization formation. Noymer (2007) shows that in Norway, the young people who
were killed tended to be those suffering from tuberculosis, so there was a sharp drop in
tuberculosis deaths in the years after the Spanish Flu. Mamelund (2004) also shows that while
there was a drop in births due to the deaths of pregnant mothers and lowered cohabitation, there
was a spike in births of children in the years following the Spanish flu. As a result, the Spanish
Flu did not sharply reduce the population.
The Spanish flu affected some communities more than other, and it was sudden and so
made behavioral adaptation unlikely. The contagion followed civilian transportation links such as
rail lines and ship routes (Mamelund, 1998). Doctors were required to report Spanish Flu cases in
Norway, and the results were made available at the medical district level (corresponding to one or
more municipal units), which gives excellent granularity of the data (Mamelund, 1998, 2003).
Indeed, the Spanish flu may be seen as a natural experiment and exogenous source of variation
that assigns municipalities to treatment and control groups. Using data from Mamelund (2003),
we re-analyze the mortality of each community, finding (like he did) that communities with a
high proportion of the Sami and Kven minorities were likely to be hit hard,2 and that deaths
occurred in crowded areas and coastal zones. The other community variables did not have effects.
This analysis is available from the authors.
COOPERATIVES IN NORWAY
The dissolution of the union of Denmark and Norway led to a significant increase in
organizing following the constitution of 1814, which gave greater freedom of expression and
interpersonal association. Myhre (2008: 27) writes that “contemporaries spoke with awe about the
colossal “spirit of association.” Many of these associations followed principles of mutual
organization and were engaged in economically important activities for communities. Prominent
steps were that the first village fire mutual was founded in 1825, the first savings bank was
established in 1822, and mutuals devoted to general insurance were founded soon after. Political
parties and trade unions emerged much later.
These organizations were part of a social movement that sought to build on the virtues of
community collective action to solve problems of risk sharing (Lorange, 1935). Early mutual
insurance organizations saw insurance as a form of community solidarity, hence restricting
customer access to individuals within the same community. The community focus led to greater
trust and denser social ties among the insured individuals, as well as better monitoring of
behaviors, all of which are important for the viability of insurance organizations (Rotschild &
Stiglitz, 1976). Village fire insurance mutuals insured buildings against fire, which was essential
2Sami are an indigenous people; Kven are immigrants from Finland following a period of starvation.
Mamelund (2003) notes that the higher mortality among the Sami could be because their lower exposure
to an earlier, slower-moving flu epidemic left them more at risk toward the Spanish Flu.
for farmers. They limited membership to individuals from the local municipality, in part because
they formalized a pre-existing custom of giving alms to victims of fire within a village (F‘rden,
1967), and many other mutual insurance firms were also limited to people from a municipality,
especially those that provided maritime insurance.
Savings banks were established to provide savings and loan services for the local
community. The social movement backing them saw the encouragement of frugality through
regular savings as the primary goal, and lending was also done in order to finance the interest
paid on savings accounts. Savings banks were established through collection of funds either from
wealthy and philanthropic community members such as leading farmers and merchants, or by
broad mobilization of community members. Savings banks were similar to mutuals in
governance, as the depositors maintained control rights and fund contributors received interest
rather than dividends. Savings banks served a need for saving and lending that was unmet in most
communities, as they were often established before the first entry of a commercial bank.
Our empirical indicators of community organizing efforts are new non-profit organizations,
in the form of foundings of retail cooperatives. The cooperative movement in Norway started in
the 1870s when laws giving monopoly rights on retail trade to urban municipalities were
loosened, but the initial wave of cooperative foundings had few and often short-lived foundings.
It was only after the creation of the National Cooperative Association in 1906 that the cooperative
movement accelerated (Debes, 1931; Lange, 2006). They had robust diffusion overall with 1,124
founding events from 1905 to 1955. Differences in the speed at which communities obtained a
retail cooperative and the final count of cooperatives relative to the human population suggest
community-level differences in collective action capacity. Retail cooperatives are still in
operation in Norway and account for one-quarter of all grocery retail sales by revenue.
Retail cooperatives are important economically and as representatives of a greater
movement of cooperative organization at the local level. Various forms of community
cooperatives focused on mutual organization has been a central part of local economic
development. Schneiberg et al. (2008) observe that risk-pooling mutuals in insurance and banking,
and cooperatives in grain elevators, fruit processing, dairy businesses, and consumer retail
underlay the rise of American capitalism. Iversen and Soskice (2009) argue that economies with a
greater cooperative organization in the 19th century developed a more egalitarian distribution of
income, had proportional representation systems, and therefore, a network of non-market
organizations and regulations in the 20th century.
Retail cooperatives were general grocery stores, so they differed from mutual insurers and
savings banks in that they did not address unmet needs in the community, but instead were a
collective solution to a need already met by private merchants. They were backed by a social
movement that was aware of the mid-19th century English cooperative system (Debes, 1931) and
the Rochdale Principles of open membership, democratic control, and distribution of surplus (e.g.,
Birchall, 1997). The Norwegian cooperatives shared this ideology, and their formation was
encouraged by a social movement to make communities less dependent on private merchants
(Greve & Rao, 2012). That meant an initial focus on a limited range of essential goods, and an
emphasis on low prices through reduced margins. Equally important, they were against the
private merchant practice of selling goods on credit with high interest rates that made customers
dependent on the merchant (Debes, 1931). Many early cooperatives still sold on credit because
their customers had seasonal incomes and had difficulty paying by cash. This was especially true
early on, because the rural economy was undergoing an economic transition from self-sufficient
farming, and hence a cash-poor economy, to specialized production, giving greater availability of
money but also greater need for retail goods all year round (Hodne & Grytten, 2000).
The Norwegian cooperative movement had some distinctive characteristics shaped by its
economic and political situation. Economically, Norway had a high population concentration in
rural areas (Hodne & Grytten, 2000), and the cooperative movement was correspondingly strong
there. Politically, the cooperative movement started just when Norway had emerged from a
period of conflict in which the rural areas had seen rebellions against the legal restrictions
imposed by the monopoly of the state church and merchant towns as centers with trade privileges.
Although freedom of religion and trade was established, the confrontations between rural
communities and the privileged merchant class in cities became just as important for the
Norwegian cooperative movement as the confrontation between labor and capital owners that
characterized cooperative movements in other nations (Debes, 1931).
The cooperative movement stated clearly their intent to not just serve members through
profit distribution and demographic governance, but also to shift the power balance from
merchants to consumers. As stated in a 1919 issue of Kooperatøren, a publication by the national
cooperative association, “Each cannot do much, but when all buyers join together they become a
decisive power” (Lange, 2006: 14). This ideology was continued through time, for example in
1937 Kooperatøren characterized cooperatives as “capitalism turned upside down. Members, not
capital, are on top” (Lange, 2006: 198).
As a result of the anti-merchant nature of the cooperative movement and the realities of
competing with a growing number of private merchants in the countryside, the cooperative
movement and the private merchant association tangled in a series of legislative battles that also
involved the major political parties, usually pitting the conservative party (Right) on the private
merchant side against the Labor and Left parties on the cooperative side (Lange, 2006). For
example, private merchants wanted removal of a legal clause allowing tax-free sales to
membership, which benefited cooperatives that sold exclusively to their members. Cooperatives
wanted exemption from the anti-trust law so they could continue the vertical integration of
wholesale and production of goods promoted by the national cooperative association.
Both the legislative and lobbying battles and the competition that took place in each
community were underpinned by the cooperative movement’s view of itself as the “third way”
between the extremes of capitalism and state rule of the economy (Debes, 1936; Lange, 2006).
The cooperative movement saw itself as the only democratic approach, because it associated
capitalism with market power and cartel building, and state rule with centralized control of the
provision and sale of goods. Cooperatives were consumer controlled, by democratic rule, and
hence represented consumer preferences, and were consumer owned and profit sharing in
proportion to purchases, and hence represented consumer economic interests. This reasoning
underpinned their appeal for exemption from the anti-trust law: they argued that it was
meaningless to cover cooperatives by anti-trust legislation because they were intrinsically
anti-trust. It also reflected the cooperative movement’s liberalist ideology, which was against any
state control of the consumer-governed part of the economy that it represented.
The private merchants and their association saw things differently (Lange, 2006). They saw
the establishment of cooperative wholesale and production as creation of market power that
private merchants could not have, as neither private wholesalers nor retailers could unite to form
production units in the same way that the cooperative association did. Even manufacturers were
threatened by cooperative manufacturing of goods that directly targeted theirs, such as flour,
margarine, and tobacco products. Also, the distinction that cooperatives drew between them and
state rule was lost on private merchants, who saw the cooperative movement’s vertical integration,
standardization of goods, and use of practices such as “agitation weeks” and “buy cooperative”
slogans as closely resembling socialist movements.
In each community, the formation of a retail cooperative would be motivated by a mixture
of agreement with the “third way” ideology and the concrete incentive of buying from a store that
would return its profits to customers. However, although this motivation fully explains why an
individual would join an existing cooperative as a member, the formation of a cooperative is a
community-level collective action problem that can fail to occur even if all community members
would have joined after foundation (Olson, 1965). This makes cooperative formation in a
community a good measure of the community ability to engage in civic action, which in turn
depends on trust and social integration.
DATA AND METHODOLOGY
The Cooperative Association in Norway provided data on cooperative foundings, and we
supplemented and cross-checked them with annual reports from its predecessor organization and
a monograph on cooperatives in Norway (Debes, 1931). An 1896 report to the parliament
provided data on village fire insurer foundings. Data on insurers came from the annual report of
the Norwegian Insurer Association and two monographs on the insurance industry (Lorange,
1935; Wigum, 1993). The Savings Bank Association provided data on savings banks foundings.
Mamelund (2003) provided data on Spanish flu mortality by medical district. Data from the
decennial censuses were drawn from the Municipality database of the Norwegian Social Science
Data Service (NSSDS). Because there were some boundary changes during this period, the data
were adjusted by recalculating the census data to 1950 boundaries using a procedure developed
by the NSSDS. Cooperatives were coded to 1950 municipality boundaries by using the postal
code of their modern location or a match of the geographical name of their founding location to a
modern postal code. The data contain 630 municipalities, but were reduced to 597 due to
changing coastal access and medical districts that were difficult to match to municipalities. These
data were previously used to show that communities with past experience with voluntarism had
more foundings of community organizations (Greve & Rao, 2012).
Hypothesis tests. We relied on two independent variables for hypothesis tests. One was
municipality Spanish flu mortality per 1000 individuals as reported by Mamelund (2003). We
follow his procedure of adding mortality attributed to the Spanish Flu and pneumonia because
some Spanish Flu cases had pneumonia as the final cause of death. In preliminary runs we also
used discounting of this mortality to examine whether its effect changes over time, but found that
a constant effect had good fit to the data.
The other was based on a Blau (1977) measure of organizational diversity (one minus the
sum of squared proportions) of non-profit forms in 1917 for each community. We treated civic
capacity as the unobservable associated with the diversity of non-profits, so we use the control
function approach (Wooldridge, 2010). We modeled the Blau measure as a function of the same
variables as those used to model cooperative founding, and extracted a Pearson Residual for
Cooperative Form Diversity in 1917 to account for the municipality’s organizational
competences prior to the onset of the Spanish Flu. We get similar results when we include the
Blau measure of non-profit diversity instead of this Pearson residual.
Spring Frost as a Contrast Case. In Norway, farming was a main occupation for many
families and a side occupation for others, and low temperatures during the spring planting season
were a significant problem for those reliant on farming. More generally, variability in climate
ensues from low-frequency processes that have long cycles and last a generation or more and lead
to fluctuations in groundwater or soil erosion, and high frequency processes that have shorter
cycles and underlie year-to-year variations. Temperature fluctuations affect yearly variations in
crop yields (Lobell & Field, 2007) and crop failure rates (Mendelsohn, 2007).
Unlike the Spanish Flu, climate variability did not produce fatalities, but it led to economic
hardship. In agriculture, weather variability in the early growing season matters more than
weather at any other time of the year (Durante, 2010). Durante (2010) conducted a large-scale
analysis of the relationship between climate variability in the growing season and risk sharing
arrangements, and found a strong relationship. In Norway, spring frost was likely to induce
farmers to develop shared storage and insurance solutions. Spring frost was such a well-known
source of hardship that Norwegian sources rarely mention it, but foreign visitors noted its
severity and the use of solutions such as community grain banks in response (Laing, 1852:
168-169). Greve and Rao (2012) found that greater spring frost made Norwegian communities
more likely to found mutual banks and insurance organizations in the 19th century.
To account for the expected positive effects of Spring Frost on cooperation, we use data on
annual deviations from the monthly mean temperature in April for five geographical regions in
Norway obtained through the climate database of the Norwegian Metrological Service. We form
a variable as the (negative) sum of below-zero deviations in the three years before the focal year.
Higher values on this variable indicate more severe recent spring frost.
Control Variables. Variables were entered to control for other characteristics that may affect
the founding rates of community organizations. War is an indicator variable set to one during
either of the two world wars. Latitude indicates how far north the center of the community is
located, and captures the lower economic resources of communities in the north. Rural is an
indicator for whether a community was classified as rural. Coastal is an indicator variable created
because communities located alongside the coastline and fjords may have a higher propensity to
be affected by the Spanish Flu than others. Additionally, coastal communities had traditions of
mutual help and sharing within the community such as lot-based distribution of the catch among
fishermen (Lorange, 1935). However, mutual aid was also practiced in inland communities. To
control for community size, the logarithm of the human population is entered. To control for
spatial access (larger areas creating demand for more stores) and possibly also heterogeneous
communities, the logarithm of the municipality area is entered.
We control for population diversity because a large body of research suggests that diversity
impedes collective action through lowering trust of strangers. Religious dissenters is the
proportion of community members that belonged to churches other than the official state church.
Proportion poor is the proportion of the community that received aid for the poor, and indicates
both community poverty and the potential for civic action to be oriented primarily toward poverty
relief. Mortality is the 5-year mortality from all causes, and is entered as a contrast to the Spanish
flu mortality. Occupational diversity was formed as the Blau index of the occupations farmer,
maritime, merchant, artisan, laborer, and other. Norway experienced significant emigration,
primarily to the USA, between 1820 and World War I, with most of the emigration occurring
between 1865 and 1885. Emigration drains the communities of resources and disrupts community
networks, so we enter a variable indicating the emigration rate. Emigration is calculated as
emigration divided by births in the same time period. A 10-year span is chosen for the time range
of this variable. It is updated every 5th year from 1905 to 1920 (the last year with emigration data
available), and the 1920 value is retained for the rest of the study period to indicate the
community disruption through emigration.
The linear and squared density of cooperative stores in the community were entered to take
into account density dependence in the founding rates (Carroll & Hannan, 2000). Although
density dependence is mainly intended to capture change in competition and legitimacy in a
single population over time, it can also be used in panel data to measure cross-sectional
differences among communities (Greve, 2002). We computed a measure of failures of
cooperatives to discern if prior failures reduced confidence in the form. This variable is defined
as all failures of cooperatives prior to the focal year in the same community. Foundings did occur
before 1905, but most ended up in failures because the cooperative stores behaved more like
commercial stores (Debes, 1931; Lange, 2006).
In view of Greve and Rao’s (2012) finding that communities that had early experience with
voluntarism had more foundings of community organizations, we entered controls for how early
the municipality founded each of the three community organizations village fire insurer, other
mutual insurer, and savings bank. Each early founding variable was calculated as the logarithm of
(1900-t), where t is the year in which the first founding of the focal organizational form took
place, and it is set to zero if the municipality has no founding of the form by year 1900. Hence,
high numbers mean early founding.
Similarly, some unobserved third variable might simultaneously influence both Spanish flu
mortality and the founding of new retail cooperatives, so the relation we observe can be an
artifact driven by an unidentified common cause. One solution to the problem of endogeneity is
the control function approach, which is a two-stage method introduced by Wooldridge (2007) and
shown to be consistent for several distributions (Wooldridge, 2010). The control function method
contains two stages and is also referred to as the two stage residual inclusion approach. In the
first stage, we predict Spanish flu mortality and extract a Pearson residual from that equation. In
the second stage, we include the Pearson residual when predicting the formation of retail
cooperatives to assess if it extinguishes the effect of Spanish flu mortality per se. We show the
descriptive statistics in Table 1.
=== Insert Table 1 about here ===
Dependent Variable and Models
The dependent variable for the founding analysis is an indicator of founding a cooperative
association in a municipality in the focal year. Accordingly, we use a complementary log logistic
model for the founding probability, which (along with the time-period indicators) makes the
model a discrete-time equivalent of the piece-wise exponential hazard rate model (Allison, 1982).
In a robustness check, a logit model with municipality random effects is applied, and found to
have consistent findings.
Table 2 presents the evidence on the different forms of disasters on the founding of
cooperatives. We begin the analysis with Model 1 that tests the short-term Hypothesis 1 by
showing the 3-year effect of the Spanish flu. The findings show a negative and significant effect
of Spanish flu mortality, consistent with Hypothesis 1. There is a positive and significant effect of
the spring frost, thus suggesting that spring frost was a contrast case to Spanish Flu. These
findings support our initial claim that contagious disease reduces cooperative foundings in the
short term, opposite of the positive effect of weather shocks that are framed as acts of nature.
Interestingly, the 3-year effect is actually weaker than the effect of the next time interval as
shown in Model 2, suggesting that the disruption of founding processes resulting from distrust
was built up over longer time than Model 1 captures.
Model 1 also has a set of control-variable results that merit mention, and that are retained
with very little variation also in subsequent models. The foundings of cooperatives are
significantly lower in rural communities and communities with many poor, but are higher in
communities with dense populations, large area, and early founding of village mutuals and
savings banks. Cooperative density in each community has a significant inverted-U shaped
relationship with the founding rate. Cooperative failures increase subsequent foundings. The
Pearson residual for organizational diversity in the cooperative sector is positive and significant,
suggesting that unobservables associated with communities are important causal influences. This
variable helps establish that subsequent findings correctly control for the community capability
and propensity to form cooperatives before the Spanish flu and weather shocks. Similarly, the
Pearson residual for Spanish Flu mortality also has a positive effect. After establishing this
contrast between the effects of the contagion frame and the act of nature frame in the short term,
we go on to examine the duration of the effects.
=== Insert Table 2 about here ===
We next present the analyses of the formation of retail cooperatives in 15-year intervals
staggered by 5 years in order to examine whether there is a gradual decline in the effect of the
Spanish Flu, In order to use the data on the pre-flu period as a control condition, each analysis
retains the 1905 through 1919 observations as well. Model 2 shows that Spanish flu mortality has
a significant negative effect on the formation of retail cooperatives until 1929. This supports
Hypothesis 2 that a contagious disease shock has a lengthy negative effect on the founding rate of
community organizations. Similarly, Model 3 relates to the 1925-1939 time period and Model 4
to the 1930-1944 time period, and in each of them Spanish flu mortality has a significant negative
effect on the formation of retail cooperatives, supporting Hypothesis 2 on its long-term effect.
Model 5 (1935-1949) fails to support Hypothesis 2 on the Spanish Flu effects.
Comparison of the coefficient estimates for Spanish Flu mortality for Models 2 through 5
shows a gradual decline in its negative effect, consistent with our theory of a loss of trust, and
with gradual buildup of trust taking place in the following decades. The chi-square estimates
below Model 2 through 5 tests for equality of the Spanish Flu coefficient of each model against
that of its predecessor model (a stricter test than testing against Model 1), and shows that all but
one of the tests are significant. Interestingly, it is the first of the long-term tests that does not
reach significance, suggesting that the loss of trust stays at the same level for about 20 years after
the Spanish Flu, and then is gradually recovered. This supports Hypothesis 3 on the weakening of
the Spanish Flu effect over time, although the weakening does not set in immediately. In these
longer time intervals, spring frost either retains significance of its positive effect or loses
Models 6 and 7 repeat the analyses of Models 2 and 5, but without the years 1905-1919 as a
control condition. This gives a replication using only post-flu data, which could in principle give
different results for two reasons. First, there could be unmeasured differences between the pre-
and post-flu period that affect the analyses and give biased estimates of variables in the early
models. Second, the restriction of variance of the Spanish Flu mortality variable in the post-flu
data (it is never zero, unlike in the data before 1920) could make this variable estimate less
precise, though it should not introduce bias. Because we are unable to distinguish which of the
two effects would account for a difference in the estimates of the two models, significant
differences of the coefficient estimates would be problematic. However, the Spanish flu mortality
coefficient estimate of Model 2 and 6 are very similar, and the Spanish flu mortality coefficient
estimate of Model 7 is actually more in support of Hypothesis 1 than that of Model 5. The
findings support Hypotheses 1 and 2, and the test shown below supports Hypotheses 3, although
at a marginal level. We maintain that our interpretation that the analyses show a gradual
dissipation of the Spanish Flu effect over time, and that it was gone (or nearly so) by the
1935-1949 interval. The Spanish Flu has a measurable effect over a generation long (25 years)
time span, as we can detect its effect as far as the 1930-1944 interval.
Models 8 and 9 enter interaction variables of Spanish flu mortality and the residual of the
diversity model. We use the residual as a proxy for civic capacity because the residual shows the
extent to which a community has a more diverse set of community organizations than expected
from its characteristics. In both models, Spanish flu mortality retains a negative effect, which is
fully significant in the early time interval (1920-34) and marginally significant in the late time
interval (1935-49). Consistent with Hypothesis 4, the interaction effect has the opposite sign and
is significant, so pre-existing civic capacity served to reduce the negative effect of Spanish Flu
mortality on cooperative founding.
Figure 1 shows the magnitude of the Spanish Flu mortality effect. Panel 1a is based on
Models 2 through 4,3 and displays how the founding rate of a community with the average
founding rate in the pre-Spanish flu period would see its cooperative founding rate change as a
function of Spanish flu mortality. The horizontal axis is between 0 and 20 deaths per 1000 (the
actual range is 2 to 28, with a mean of 6), and the figure shows that in the next 15 years, the
predicted founding rate falls by 58% from zero to the mean mortality rate of 6, and by 93% if the
mortality rate hits 18. This effect is weakened as time passes, so the drop in founding rate at the
average mortality rate of 6 is 48% in 1926-40, and 31% in 1931-45. Panel 1b is based on Model 8,
and shows the interaction effect of the diversity residual in the first time period. In this model, the
drop in cooperative founding rate for communities with an average (zero) residual is 66%, but the
3 Model 5 has an insignificant coefficient estimate, so we do not plot it.
founding rate is reduced by 76% when the residual is two standard deviations below, and is
reduced by 52% when the residual is two standard deviations above. These plots show that 1) the
Spanish flu strongly affected founding rates, 2) the effect was reduced over time, and 3) the effect
was smaller in communities that had a high civic capacity before the Spanish flu. Figure 1
strongly supports our theory. For comparison, we also calculated the effects of spring frost in
models 1 and 2, which saw the greatest effects. Spring frost also had substantial effects, with an
increase in the founding rate of up to 230% in Model 1 and 155% in Model 2. However, even a
sequence of years with very high spring frost would not be enough to overcome the effect of high
Spanish Flu mortality (e.g,, 0.07 x 2.3 = 0.16, so a reduction to one-sixth of the founding rate).
DISCUSSION AND CONCLUSIONS
Our paper extends the literature on how communities and organizing efforts interrelate by
showing that community founding of cooperatives can experience long-term effects of a disaster.
Reported research has usually emphasized either the occurrence of disasters (e.g., Dutta, 2016),
or the magnitude of disasters (e.g., Tilcsik & Marquis, 2013). Our study suggests that there is an
important difference in the effects on the community and local organizations not just based on
occurrence and magnitude but also the causal framing of the disaster. We advanced the idea of a
disaster frame and documented a contrast between disasters attributed to other community
members (members infected by a disease in our case) and hardships attributed to nature (spring
frost). To date, the research on disasters has either embraced an a priori classification of disasters
into natural versus man made, or treated disasters as the stimulus and disaster relief operations as
the response. What has been missing is the disaster interpretation and framing by community
members that shape both response and community resilience. Our study suggests that just as
movements have frames, disasters too have frames based on how members attribute their cause.
This insight renders a priori classification superfluous and instead directs attention to how
community members perceive and affix responsibility. In the short term, it would be difficult to
empirically disentangle the effects of the social framing of contagion from those of the actual
disease contagion, but our analysis of long-term effects make the argument for a distinct social
Our primary interest is in contagious disease outbreaks (such as the Spanish Flu) as an
instance of disasters attributed to other community members, and we use weather shocks as a
contrasting example of disasters attributed to nature. We find that disasters attributed to other
community members weaken cooperation, increase suspicion and distrust of the other, and lead to
a long-term (with a declining effect) reduction in organization building. By contrast, disasters
attributed to an act of nature evoke a sense of shared fate which fosters cooperation, although
with short term effect. These findings suggest that disasters are not merely physical events but
socially constructed as well.
Moreover, our findings also expand the reach of research on civic capacity by qualifying the
conclusion that sociable communities are more likely to help themselves and be resilient
(Sampson, 2012). Sampson, Raudenbush, and Earls (1997) pointed out the abundance of
non-profit organizations as an incubator of sociability and resilience. While a number of
organizational scholars are interested in long-lasting effects of past conditions and events on
communities and organizations (e.g., Greve & Rao, 2012; Tilcsik & Marquis, 2013), our analyses
suggest that organizational diversity matters. The Pearson residual of cooperative form diversity
pertains to community unobservables that drive organizational founding. This had a significant
positive effect on the formation of retail cooperatives decades after the founding of the village
insurers, mutual banks, and mutual insurers we used to construct this measure. Diversity of
organizational forms means diversity of routines and competences, which provides security to
The literature on disasters has tended to focus on events such as floods, earthquakes,
chemical spills, and tornadoes (Erikson, 1994). Contagious disease outbreaks have been given
short shrift. We document the organizational consequences of the Spanish Flu by showing that it
lowered rates of forming new retail cooperatives not just in the short run but also in the long run.
By doing so, our study adds to the public health literature by showing that epidemics of
contagious diseases have long-lasting effects on the organizational vitality of communities. This
means that policy responses to epidemics should be broadened to include support for founding
new organizations. The typical response to pandemics include isolation and treatment, home
quarantines, closure of schools, cancellation of large-scale public meetings, and other steps to
reduce social density (Center for Disease Control and Prevention, 2007). While these immediate
responses are entirely practical, policy planners should also consider how a pandemic impairs the
social infrastructure of a community over the long-term, and undertake initiatives to foster the
building of community organizations. After all, if it is sociable communities that survive disasters
by helping themselves, investments in enhancing the social infrastructure of communities too
Our results extend the literature on institutional legacies by clearly showing that contagious
diseases are negative shocks to cooperation in communities with long run effects. The Spanish
Flu in Norway in 1917-1919 killed 0.6 percent of the population, but took out a swath of young
people. Even taking this selective mortality into account, its effect on the founding of retail
cooperatives was disproportionately strong and lengthy, resulting in an institutional legacy of
weaker collective action overall and especially in the most hard hit communities. The cooperation
deficit impaired the ability of Norwegian communities to build new community organizations for
25 years after the onset of the Spanish Flu. Our analyses also extend the work of Greve and Rao
(2012) by demonstrating that exposure to climate shocks increased the formation of retail
cooperatives even after accounting for the negative cooperation shock from the Spanish Flu.
The findings also add to the corpus of work on entrepreneurship. For long, as Aldrich and
Ruef (2012) note, scholars have focused on the traits of entrepreneurs and not rates of
entrepreneurship. During the last two decades, organizational ecologists have identified how
density-dependent legitimation and competition in populations of organizations affects rates of
entrepreneurship. More recently, the work has taken a cognitive turn, and suggested that whether
a form becomes a widely accepted category determines rates of entrepreneurship (Hannan, Polos,
& Carroll, 2007). Our study may be seen as an effort to understand categorization shocks and
how they affect norms of cooperation in a community. So whereas the Spanish Flu led to other
community members being labeled as threats and fostered distrust, weather shocks created a
common fate and had positive effects on norms of cooperation. Indeed, norms of cooperation can
affect the ease of entrepreneurship activity.
Some natural extensions merit attention. We concentrated on contagious diseases and
climate shocks in 20th century Norway, in which the former was attributed to humans and the
latter to an act of nature. In other times and places, these same disasters may both be attributed to
God, and we also need to understand the consequences of this attribution. Moreover, future
research needs to understand the effect of infectious diseases, which are unlike contagious
diseases because they do not spread person to person, but with an agent such as a mosquito or a
rodent. Other studies have found that a dearth of rainfall aggravated crime, such as violence by
men against their wives to extract dowry in India (Sekhri & Storeygard, 2011), and that floods or
drought incite people to attack elderly women by labeling them as witches (Miguel, 2005). These
studies primarily focus on climate effects in poor countries, and it is possible that weather shocks
are amplified by poverty, ethnic heterogeneity, and the lack of institutions. A comparative
analysis of different climate shocks and how they are mediated by institutions beckons for
researcher attention. We studied the effect of exogenous shocks on the birth rates of organizations
with a community focus, but researchers also ought to look at whether they have corresponding
effects on the death rates of organizations. Indeed, a viable idea is to assess whether organizations
built during shocks survive a trial by fire (Swaminathan, 1996) and have longer lives than
organizations founded under more munificent conditions.
We began this paper with the concern for disaster consequences voiced by Erikson (1994)
and the attention to community sociability shown by Putnam et al. (1993). It is but fitting to close
with a nod to de Tocqueville who said that “In democratic countries, the science of association is
the mother of science: the progress of all of the rest depends on the progress it has made.” This is
a call for further study of events that have long lasting effects on community and cooperation, as
expressed through the formation of cooperative associations. Such associations are the basis of
civic society, and understanding them is essential to understand the prospects of our societies.
Acemoglu, D., & Robinson, J. 2012. Why Nations Fail: Origins of Power, Poverty and Prosperity.
New York: Crown Publishers (Random House).
Aldrich, D. P. 2012. Building Resilience: Social Capital and Post-Disaster Recovery. Chicago,
IL: University of Chicago Press.
Allison, P. D. 1982. Discrete time methods for the analysis of event histories. In S. Leinhardt
(Ed.), Sociological Methodology 1982: 61-98. San Francisco: Jossey Bass.
Ansart, S., Pelat, C., Boelle, P. Y., Carrat, F., Flahault, A., & Valleron, A. J. 2009. Mortality
burden of the 1918–1919 influenza pandemic in Europe. Influenza and Other
Respiratory Viruses, 3(3): 99-106.
Bateson, G. 1956. The message, 'This is play.'. Group processes, 2: 145-241.
Birchall, J. 1997. The International Co-operative Movement: Manchester University Press.
Blau, P. M. 1977. Inequality and Heterogeneity. New York: The Free Press.
Blocker, T. J., & Sherkat, D. E. 1992. In the eyes of the beholder: Technological and naturalistic
interpretations of a disaster. Organization & Environment, 6(2): 153-166.
Borza, T. 2001. Spanskesyken i Norge 1918-19. Tidskrift for Den norske legeforening, 30(121):
Campbell, D. T. 1958. Common fate, similarity, and other indices of the status of aggregates of
persons as social entities. Behavioral science, 3(1): 14-25.
Carr, L. J. 1932. Disaster and the sequence-pattern concept of social change. American Journal
of Sociology: 207-218.
Carroll, G. R., & Hannan, M. T. 2000. The demography of corporations and industries.
Princeton, N.J.: Princeton University Press.
Carroll, J. M. 2015. Slow Burn in China: Factories, Fear, and Fire in Canton. In R. Peckham (Ed.),
Empires of Panic: Epidemics and Colonial Anxieties: 35-55. Hong Kong: Hong Kong
Center for Disease Control and Prevention. 2007. Interim pre-pandemic planning guidance:
community strategy for pandemic influenza mitigation in the United States.
Crandall, C. S., & Schaller, M. 2005. Social psychology of prejudice: Historical and
contemporary issues. Lawrence, KS: Lewinian Press.
Das, V. 2007. Life and Words: Violence and the Descent into the Ordinary: Univ of California
Debes, I. 1931. Forbruker Kooperasjonens Historie i Norge, del I. Oslo: Norges Kooperative
Debes, I. 1936. Forbruker Kooperasjonens Historie i Norge, del II. Oslo: Norges Kooperative
Durante, R. 2010. Risk, Cooperation and the Economic Origins of Social Trust: An Empirical
Investigation. SSRN Working Paper Series.
Dutta, S. 2016. Creating in the crucibles of nature's fury: Associational diversity and local social
entrepreneurship after natural disasters in California, 1991-2010. Administrative Science
Dutta, S., & Rao, H. 2015. Infectious diseases, contamination rumors and ethnic violence:
Regimental mutinies in the Bengal Native Army in 1857 India. Organizational Behavior
and Human Decision Processes, 129: 36-47.
Edelstein, M. R. 1988. Contaminated communities: The social and psychological impacts of
residential toxic exposure. Boulder CO: Westview Press.
Elliott, J. R., Haney, T. J., & Sams-Abiodun, P. 2010. Limits to Social Capital: Comparing
Network Assistance in Two New Orleans Neighborhoods Devastated by Hurricane
Katrina. Sociological Quarterly, 51(4): 624-648.
Erikson, K. 1994. A New Species of Trouble: The Human Experience of Modern Disasters.
New York: Norton.
Erkoreka, A. 2009. Origins of the Spanish Influenza pandemic (1918–1920) and its relation to the
First World War. Journal of molecular and genetic medicine: an international journal
of biomedical research, 3(2): 190.
F‘rden, K. 1967. Forsikringsvesenets historie i Norge 1814-1914. Oslo: Wesmanns
Frazer, J. G. 1911. The Golden Bough: A Study in Magic and Religion: London: Macmillan.
Frémeaux, J. 2006. Combats et épreuves des peuples d'outre-mer. 14-18 éditions. Cahors: Les
colonies dans la Grande Guerre: .
Freudenburg, W. R. 1997. Contamination, corrosion and the social order: An overview. Current
Sociology, 45(3): 19-39.
Gamson, W. A., Fireman, B., & Rytina, S. 1982. Encounters with Unjust Authority: Dorsey
Goffman, E. 1974. Frame Analysis: An Essay on the Organization of Experience: Harvard
Greve, H. R. 2002. An ecological theory of spatial evolution: Local density dependence in Tokyo
banking, 1894-1936. Social Forces, 80(3): 847-879.
Greve, H. R., & Rao, H. 2012. Echoes of the Past: Organizational Foundings as Sources of an
Institutional Legacy of Mutualism. American Journal of Sociology, 118(3): 635-675.
Greve, H. R., & Rao, H. 2014. History and the present: Institutional legacies in communities of
organizations. Research in Organizational Behavior, 34(0): 27-41.
Haidt, J., & Algoe, S. 2004. Moral amplification and the emotions that attach us to saints and
demons. Handbook of experimental existential psychology: 322-335.
Haidt, J., McCauley, C., & Rozin, P. 1994. Individual differences in sensitivity to disgust: A scale
sampling seven domains of disgust elicitors. Personality and Individual differences,
Haidt, J., Rozin, P., McCauley, C., & Imada, S. 1997. Body, psyche, and culture: The relationship
between disgust and morality. Psychology & Developing Societies, 9(1): 107-131.
Hannan, M. T., Polos, L., & Carroll, G. R. 2007. Logics of Organization Theory: Audiences,
Codes, and Ecologies. Princeton: Princeton University Press.
Haveman, H. A., Rao, H., & Paruchuri, S. 2007. The winds of change: The progressive
movement and the bureaucratization of thrift. American Sociological Review, 72(1):
Hodne, F., & Grytten, O. H. 2000. Norsk økonomi i det 19. århundre. Bergen: Fagbokforlaget.
Iversen, T., & Soskice, D. 2009. Distribution and Redistribution: The Shadow of the Nineteenth
Century. World Politics, 61(3): 438-486.
Kelley, H. H. 1971. Attribution in Social Interaction. Morristown, NJ: General Learning Press.
Kleinberg, E. 2003. Heatwave: The Social Autopsy of Disaster. Chicago, IL: University of
Laing, S. 1852. Journal of a Residence in Norway during the years 1834, 1835, & 1836.
London: Longman, Brown, Green, and Longmans.
Lange, S. 2006. Organisert Kjopekraft: Forbrukersamvirkets Historie i Norge. Oslo: Pax.
Lie, A. K. 2007. Origin Stories and the Norwegian Radesyge. Social History of Medicine, 20(3):
Lobell, D. B., & Field, C. B. 2007. Global scale climate–crop yield relationships and the impacts
of recent warming. Environmental research letters, 2(1): 014002.
Lorange, K. 1935. Forsikringsvesenets historie i Norge inntil 1814. Oslo: Wesmanns
Mamelund, S.-E. 1998. Diffusjon av influensa i Norge under spanskesyken 1918-19. Norsk
Epidemiologi, 8(1): 45-58.
Mamelund, S.-E. 2003. Spanish Influenza Mortality of Ethnic Minorities in Norway 1918–1919.
European Journal of Population, 19: 83-102.
McEvily, B., Perrone, V., & Zaheer, A. 2003. Trust as an Organizing Principle. Organization
Science, 14(1): 91-103.
McFarlane, A. C., & Norris, F. H. 2006. Definitions and concepts in disaster research. Methods
for disaster mental health research: 3-19.
Mendelsohn, R. 2007. What causes crop failure? Climatic change, 81(1): 61-70.
Meyer, A. D. 1982. Adapting to environmental jolts. Administrative science quarterly: 515-537.
Miguel, E. 2005. Poverty and witch killing. The Review of Economic Studies, 72(4): 1153-1172.
Miller, D. T., & Ross, M. 1975. Self-serving biases in the attribution of causality: Fact or fiction?
Psychological bulletin, 82(2): 213.
Molotch, H., Freudenburg, W., & Paulsen, K. E. 2000. History repeats itself, but how? City
character, urban transition, and the accomplishment of place. American Sociological
Review, 65: 791-823.
Myhre, J. E. 2008. Academics as the Ruling Elite in 19th Century Norway. Historical Social
Noymer, A. 2007. Selective mortality in Norway during the 1918 flu pandemic. Paper presented
at the Annual Meeting - Population Association of America, Princeton.
Olshansky, R. B., Johnson, L. A., & Topping, K. C. 2006. Rebuilding Communities Following
Disaster: Lessons from Kobe and Los Angeles. Built Environment, 32(4): 354-374.
Olson, M. 1965. The logic of collective action. Cambridge, MA: Harvard University Press.
Oxford, J. S., Lambkin, R., Sefton, A., Daniels, R., Elliot, A., Brown, R., & Gill, D. 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(7): 940-945.
Palmer, D. 1983. Broken ties: Interlocking directorates and intercorporate coordination.
Administrative Science Quarterly, 28(March): 40-55.
Palmer, D., Friedland, R., & Singh, J. V. 1986. The ties that bind: Organization and class bases of
stability in a corporate interlock network. American Sociological Review, 51: 781-796.
Parker, S. D., Brewer, M. B., & Spencer, J. R. 1980. Natural disaster, perceived control, and
attributions to fate. Personality and Social Psychology Bulletin, 6(3): 454-459.
Peckham, R. 2015. Empires of Panic: Epidemics and Colonial Anxieties: Hong Kong
Picou, J. S., Marshall, B. K., & Gill, D. A. 2004. Disaster, litigation, and the corrosive community.
Social forces, 82(4): 1493-1522.
Putnam, R., Leonardi, R., & Nanetti, R. 1993. Making Democracy Work: Civic Traditions in
Modern Italy. Princeton, NJ: Princeton University Press.
Quarantelli, E. L. 2005. What is a disaster? A dozen perspectives on the question: Routledge.
Quarantelli, E. L., & Dynes, R. R. 1977. Response to social crisis and disaster. Annual Review of
Sociology, 3: 23-49.
Rotschild, M., & Stiglitz, J. 1976. Equilibrium in competitive insurance markets: An essay on the
economics of imperfect information. Quarterly Journal of Economics, 90: 629-650.
Rozin, P., Markwith, M., & McCauley, C. 1994. Sensitivity to indirect contacts with other
persons: AIDS aversion as a composite of aversion to strangers, infection, moral taint, and
misfortune. Journal of Abnormal Psychology, 103(3): 495.
Rozin, P., & Nemeroff, C. 1990. The laws of sympathetic magic: A psychological analysis of
similarity and contagion.
Sampson, R. J. 2012. The Great American City: Chicago and the Enduring Community Effect.
Chicago, IL: University of Chicago Press.
Sampson, R. J., MacIndoe, H., McAdam, D., & Weffer-Elizondo, S. 2005. Civil society
reconsidered: The durable nature and community structure of collective civic action.
American Journal of Sociology, 111(3): 673-714.
Sampson, R. J., Raudenbush, S. W., & Earls, F. 1997. Neighborhoods and Violent Crime: A
Multilevel Study of Collective Efficacy. Science, 277(5328): 918-924.
Schaller, M. 2011. The behavioural immune system and the psychology of human sociality.
Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1583):
Schaller, M., & Neuberg, S. L. 2012. Danger, Disease, and the Nature of. Advances in
experimental social psychology, 46: 1-54.
Schneiberg, M., King, M., & Smith, T. 2008. Social Movements and Organizational Form:
Cooperative Alternatives to Corporations in the American Insurance, Dairy, and Grain
Industries. American Sociological Review, 73(4): 635-667.
Sekhri, S., & Storeygard, A. 2011. The impact of climate variability on crimes against women:
dowry deaths in India: Tufts University.
Shaver, K. G. 1970. Defensive attribution: Effects of severity and relevance on the responsibility
assigned for an accident. Journal of Personality and Social Psychology, 14(2): 101.
Snow, D. A., & Benford, R. D. 1988. Ideology, frame resonance, and participant mobilization.
International Social Movement Research, 1(1): 197-217.
Snow, D. A., & Benford, R. D. 1992. Master frames and cycles of protest. Frontiers in social
movement theory: 133-155.
Snow, D. A., Rochford, E. B., Worden, S. K., & Benford, R. D. 1986. Frame alignment processes,
micromobilization, and movement participation. American Sociological Review, 51(4):
Sutcliffe, K. M., & Vogus, T. J. 2003. Organizing for resilience. In K. S. Cameron, J. E. Dutton,
& R. E. Quinn (Eds.), Positive Organizational Scholarship: Foundations of a New
Discipline: 94-110. San Francisco: Berrett-Koehler.
Swaminathan, A. 1996. Environmental conditions at founding and organizational mortality: A
trial-by-fire model. Academy of Management Journal, 39(5): 1350-1377.
Taubenberger, J. K., Reid, A. H., Krafft, A. E., Bijwaard, K. E., & Fanning, T. G. 1997. Initial
Genetic Characterization of the 1918 “Spanish” Influenza Virus. Science, 275(5307):
Taylor, V. 1978. Future directions for study. In E. L. Quarantelli (Ed.), Disasters: Theory and
research: 251-280. London: Sage.
Tilcsik, A., & Marquis, C. 2013. Punctuated generosity how mega-events and natural disasters
affect corporate philanthropy in US communities. Administrative Science Quarterly, 58:
van der Vegt, G. S., Essens, P., Wahlström, M., & George, G. 2015. Managing risk and resilience,
Academy of Management Journal, Vol. 58: 971-980: Academy of Management.
Vaughan, W. T. 1921. Influenza: an epidemiologic study. Baltimore: American journal of
Veltfort, H. R., & Lee, G. E. 1943. The Cocoanut Grove fire: A study in scapegoating. The
Journal of Abnormal and Social Psychology, 38(2S): 138.
Voigtländer, N., & Voth, H.-J. 2012. Persecution perpetuated: the medieval origins of anti-Semitic
violence in Nazi Germany. Quarterly Journal of Economics, 127(3): 1339-1392.
Walster, E. 1966. Assignment of responsibility for an accident. Journal of personality and social
psychology, 3(1): 73.
Wigum, S. 1993. "Lykken er d†rens formynder": Historie om gjensidig fiskeriforsikring i
Norge. Bergen, Norway: Hegland Trykkeri.
Wildavsky, A. B. 1988. Searching for safety: Transaction publishers.
Wooldridge, J. M. 2010. Econometric Analysis of Cross Section and Panel Data, second
edition: MIT press.
Zagefka, H., Noor, M., Brown, R., de Moura, G. R., & Hopthrow, T. 2011. Donating to disaster
victims: Responses to natural and humanly caused events. European journal of social
psychology, 41(3): 353-363.
Table 1: Descriptive statistics and correlation coefficients
Variable Mean s.d. Min Max 1 2 3 4 5 6 7 8 9 10 11 12
1 Founding 0.027 0.161 0
2 War year 0.246 0.430 0
3 Latitude 61.97 3.267 58.03
4 Rural comm 0.935 0.246 0
5 Coastal comm 0.559 0.496 0
6 Ln pop. dens. 7.715 0.725 5.09
7 Ln area
10 Occup. dive. 0.541 0.158 0.08
11 Emigration 0.243 0.331 0
12 Mortality 14.231 2.788 3.50
13 Coop density 1.038 2.862 0
14 Coop failure 0.032 0.296 0
15 Village fire 0.666 1.364 0
16 Saving bank 0.478 1.070 0
18 Dive. Pears.
19 Flu Pears.
20 Flu mortality 3.932 3.748 0.00
21 Spring frost 1.458 1.122 0.00
14 Failures 0.22
15 Fire insurer
16 Saving bank
18 Dive. resid. -0.02
19 Flu resid. 0.01
20 Flu mortality 0.14
21 Spring frost -0.02
Table 2: Log logistic model of cooperative founding
Foundings in years 1920-1922 1920-34 1925-1939 1930-1944 1935-1949 1920-34 1935-1949 1920-34 1935-1949
0.378** 0.474** 0.705**
0.058** 0.068** 0.020 0.157** -
Rural community -
-0.520+ -1.136+ -0.150 -
Ln population density
0.173** 0.184** 0.142 0.160*
Proportion dissenters -
1.694 2.892** -0.128 6.264** -
Proportion poor -
-5.210* -5.545** -2.907 -3.342 -
0.025 0.030+ -0.063+ 0.046* -
0.188** 0.041 0.195**
Cooperative density -
-1.159** -0.095 -1.090** -0.052 -
Early village fire
Early savings bank
Early insurance firm
0.081 0.102 0.087
0.551** 0.593** 0.878** 0.855**
3.921+ 16.905** 7.514**
Spring frost (contrast to
0.044 0.067** 0.026 -0.064
Flu mortality x diversity
-12.630** -13.206** -8.062** -17.466** -
Flu mortality reduced
Likelihood Ratio Test 337.32** 394.36** 363.20** 377.84** 488.54** 127.00** 322.98** 403.36** 498.26**
+ p<0.1; * p<0.05; ** p<0.01
Model 1-5 and 8-9 include years 1905-1919 as a pre-flu contrast. Models 6 and 7 do not have contrast years.
Figure 1a: Flu Mortality Effects on Cooperative Founding over Time
Figure 1b: Flu Mortality Interacted with Diversity Residual, 1921-1935