Divorce following the September 11
Catherine L. Cohan
The Pennsylvania State University
Steve W. Cole
HopeLab Foundation and the University of California, Los Angeles
The Pennsylvania State University
We investigated the effect of the September 11, 2001 terror-
ist attack on marital stability. Previous research showed rates
of divorce changed in opposite directions following natural
disaster versus terrorist disaster. Using a prospective, longi-
tudinal design and time series analysis, we examined rates of
divorces ﬁled by month, with respect to the World Trade Center
attack in New York City (NYC). To examine whether effects
radiated beyond NYC according to geographic proximity or
psychological proximity, we examined four other counties of
varying distance from NYC. Results showed geographic and
psychological proximity effects. Following a major manmade
disaster characterized by death, divorce rates decreased in
NYC and Bergen County, NJ, geographically proximal locales,
and in Los Angeles and Philadelphia, psychologically proximal
KEY WORDS: disaster • divorce • September 11 • time series analysis
Journal of Social and Personal Relationships Copyright © 2009 SAGE Publications
(www.sagepublications.com), Vol. 26(4): 512–530. DOI: 10.1177/0265407509351043
This research was supported by a grant from the National Institutes of Health (R03
HD044694–01). We wish to thank Mark Alleva, Robert M. Boyke, Darrin Goldman, Barry
Goldstein, James McElligott, Peter Sorrento, and James Stadel for their generous assistance
with data collection. Address correspondence to Catherine Cohan, The Pennsylvania State
University,Survey Research Center, 406 Marion Place, 330 Building at Innovation Park, Suite
105, University Park,PA 16802, USA [e-mail: CLC18@psu.edu].Paul Mongeau was the Action
Editor on this article.
When spouses vow to stay together “for better and for worse” they are
probably not anticipating that “worse” might mean the experience of a
natural or man-made disaster. Disasters affect individuals, families, com-
munities, and in the case of the September 11, 2001 terrorist attack – an
entire nation. A primary focus of disaster research has been to identify the
mental health consequences for individuals following natural and techno-
logical disasters. However “.. . the experience (of disaster) cannot be
expressed entirely in diagnoses of psychopathology” (Vlahov, 2002, p. 295).
An exclusive focus on individual mental health outcomes will underestimate
the full psychosocial scope of a disaster for adults in close relationships.Our
primary goal is to determine whether the September 11 disaster precipi-
tated changes in rates of marital instability in and beyond New York City
(NYC). The nearly 3000 victims of the World Trade Center (WTC) attack
resided primarily in New York and New Jersey. Individuals in the family-
building years (ages 20 to 44) were disproportionately affected (Johnson,
Stress and marital instability
Depending on the context, disaster can trigger marital instability or soli-
darity. Rates of divorce can increase following major social, political, and
economic events. Crisis situations can generate economic problems, mental
health problems, and marital distress that can disrupt the family. Major
sociopolitical events such as World War II and the Vietnam War preceded
an increase in divorce rates (Lipman-Blumen, 1975;South, 1985). Poor econ-
omic circumstances also hasten the breakdown of marriages. For example,
signiﬁcant income loss and unemployment are related to increased risk of
divorce (South, 1985;Yeung & Hofferth, 1998). Thus post-September 11 job
losses in NYC (Woodberry, 2002) and the worst unemployment in over
three years (Dillon & Schoolman, 2002) may have damaged marriages.
The disaster literature suggests an effect on marital instability via mental
health problems, which are associated with marital distress (Beach &
O’Leary, 1993; Davila & Bradbury, 1998; McLeod, 1994) and in turn divorce
(Karney & Bradbury, 1995). Mental health problems, such as post-traumatic
stress disorder (PTSD) and depression, are common after a disaster (e.g.,
Rubonis & Bickman, 1991), primarily in the subsequent 12 months (Norris
et al., 2002). Mass violence disasters have a more negative effect on adjust-
ment than natural disasters, presumably because of the greater trauma
and demoralization due to murderous intent (Norris et al., 2002). Post-
September 11 research showed mental health consequences in NYC and
across the country. One to two months after the attack, rates of probable
PTSD were higher than normal in NYC (Galea et al., 2002; Schlenger et
al., 2002) but were within normal limits in other major metropolitan areas
(Schlenger et al., 2002).Depression was two times the normative rate (Galea
et al., 2002), and substance use increased signiﬁcantly in NYC (Vlahov et
al., 2002). Three national surveys conducted within weeks of September
11 showed Americans across the country felt distressed after the attack
(Schlenger et al., 2002; Schuster et al., 2001; Silver, Holman, McIntosh,
Cohan et al.: Divorce following September 11 513
Poulin, & Gil-Rivas, 2002). Distress increased with proximity to NYC and
degree of loss. More television viewing about the attacks was related to dis-
tress, suggesting that the vicarious disaster exposure was potent.Although,
on average,the elevated distress among Americans across the United States
after September 11 did not reach clinical levels, it could still have an im-
portant effect on daily activities, such as strained social relationships and
decreased work productivity (Stein et al., 2004).
The elevated mental health problems in NYC may have translated into
increased marital distress and consequently increased marital disruption
and divorce. Of particular relevance to the current research is the Cohan
and Cole (2002) study that examined the effect of Hurricane Hugo in 1989
on divorce rates in South Carolina. Using population-level vital statistics
data for all 46 counties from 1975 to 1997, ARIMA time series analyses
showed a geographic proximity effect such that the divorce rate increased in
1990, the year following the hurricane, in the 24 counties declared Federal
disaster sites but not in the other 22 non-affected counties. In sum, marital
stability is vulnerable to disasters,major sociopolitical events, and economic
downturns, all characteristics of the September 11 attack.
Stress and marital solidarity
Two perspectives suggest that marital solidarity may occur after disaster.
As the original author of attachment theory, Bowlby (1969/1982) theorized
that, in response to threat,universal behavioral strategies evolved for infants
to maintain proximity to their caregivers for the purpose of security, safety,
and survival. Bowlby (1988) also maintained that the attachment system is
active throughout the life span.Adults exhibit proximity and support seeking
in response to stress that can be conceptualized as attachment behavior
(Hazan & Shaver, 1994). Under severe threat, Bowlby posited that danger
triggers efforts to be physically close to a “trusted person” (1969/1982,
p. 207). “When sirens scream of approaching disaster, minds turn to loved
ones. If they are near enough, mothers run to protect their children, and
men seek their families. They huddle together and support one another
through the stress” (Hill & Hansen,1962, p.186). In sum, attachment theory
suggests that, at extreme levels of danger, adults’ and children’s modal
response is proximity seeking to provide physical safety and emotional
comfort. Afﬁliation following extreme threat may translate into maintain-
ing marital bonds by being less likely to divorce. Given the emphasis on
direct threat and survival, attachment theory would predict decreases in
divorces in affected areas but not in non-affected ones.
Terror management theory posits that reminders of mortality may increase
afﬁliation.According to the theory,a fundamental aspect of the human con-
dition is an existential fear of death (Pyszczynski, Greenberg, & Solomon,
1999; Solomon, Greenberg, & Pyszczynski, 1991). In response to reminders
of death, close relationships can buffer that fear through literal protection,
emotional comfort, and as a symbol of ongoing life (Mikulincer, Florian,
& Hirschberger, 2003, 2004). To reduce death anxiety after being primed
with thoughts of death, people may be motivated to protect serious close
514 Journal of Social and Personal Relationships 26(4)
relationships with long-term committed bonds by avoiding “interpersonal
conﬂicts that can destroy these meaningful bonds which involve a strong
sense of connectedness and commitment and provide a sense of continuity
and lastingness” (Mikulincer et al., 2003, p. 25). Laboratory research showed
that undergraduates primed with reminders of death reported greater desire
for intimacy with and greater commitment to their current partner (Florian,
Mikulincer, & Herschberger, 2002; Mikulincer & Florian, 2000), compro-
mised their standards for a partner (Hirschberger, Florian, & Mikulincer,
2003), and sat close to others rather than alone (Wisman & Koole, 2003).
Given that death and the threat of death are central to the September 11
disaster (Pyszczynski, Solomon, & Greenberg, 2003), terror management
theory would also support the prediction that the September 11 disaster
would lead to an increase in marital solidarity and a decrease in divorces.
Given the emphasis on reminders of death, terror management theory would
predict decreases in divorces at large because affected and non-affected
areas were all exposed to graphic reminders of death.
In contrast to the Cohan and Cole (2002) study that showed increased
divorces in the year following natural disaster, a study of divorce rates and
the 1995 Oklahoma City bombing showed the opposite result. After the
bombing, divorce rates decreased (i) in the 7 counties in and around Okla-
homa City compared to the other 70 more geographically distant counties
in Oklahoma and (ii) in the 13 metropolitan Oklahoma counties compared
to the nonmetropolitan ones (Nakonezny, Reddick, & Rodgers, 2004). The
pattern of change in divorce rates showed a proximity effect as well as a
metropolitan effect. Interestingly, counties geographically distant from
Oklahoma City but similar in terms of metropolitan status also showed the
decline in divorce. This result is so striking because effects are restricted to
geographic proximity with most disasters (Galea & Resnick, 2005). The
metropolitan effect may have occurred if residents of other metropolitan
counties saw themselves as similar to those in Oklahoma City and thus felt
vulnerable. In other words, though geographically distant they presumably
experienced a psychological proximity to the bombing and behaved similar
to people with direct exposure to the bombing.
Divorce as a function of proximity to disaster
Where we observe change in divorce rates and the direction of that change
are two aspects of the spatial analysis that can provide theoretical insight
into how the September 11 disaster affected people and their marriages.
Proximity effects are the norm for natural and technological disasters such
that those directly affected suffer negative consequences but those beyond
do not (Galea & Resnick, 2005). The detectable (albeit subclinical) eleva-
tions in distress far from the site of the September 11 attack (Schlenger et
al., 2002; Schuster et al., 2001; Silver et al., 2002) were remarkable in terms
of disaster research and lend credence to the possibility that marriages
across the country far from the direct exposure of the attacks may have
been affected as well. Studies of mental health reactions to September 11
and the Oklahoma City bombing divorce study (Nakonezny et al., 2004)
Cohan et al.: Divorce following September 11 515
suggest that responses to terrorist disasters are affected by two kinds of
proximity – geographic proximity and psychological proximity. Observed
change may occur according to a spatially-ordered process like exposure to
trauma in which there are changes in close physical proximity to the affected
areas but not elsewhere, as with natural disasters. For example, mental
health problems occur as a result of direct exposure and related social and
economic disruptions and dissipate as one goes farther from the disaster
site (Galea et al., 2002; Hanson, Kilpatrick, Freedy, & Saunders, 1995).
The notion of relative risk appraisal helps to explain how psychological
proximity may occur. Marshall et al (2007) posited that relative risk ap-
praisal may have mediated the September 11 disaster and responses to it,
especially for those with indirect exposure. People tend to overestimate
their risk of personal harm from rare, novel, and frightening events – char-
acteristics of terrorism. Indirect exposure, such as through visual images of
the disaster, can generate appraisals of higher risk and greater distress
(Marshall et al., 2007). Perceived risk may be increased by relative proxim-
ity, perceived similarity, or a unique threat. Citizens of Los Angeles, who
felt “a dark shiver of vulnerability,” had a particular reason to feel at greater
risk (Ness, Schevitz, & Aguila, 2001, p. A24). Three of the four hijacked
airplanes were bound for Los Angeles. People in other urban areas with
high-proﬁle landmarks, such as the Sears Tower in Chicago, also feared
attack (Yednak & Mellen, 2001).Those with indirect exposure who perceive
a high risk for terrorist attack may feel a psychological proximity to the
WTC attack that results in similar responses as those with direct exposure
and an actual degree of threat.
The current study and hypotheses
To examine whether the September 11 terrorist attack had a signiﬁcant effect
on marital stability, we used a prospective longitudinal design to examine
population-level vital statistics divorce data by county. We examined coun-
ties with varying degrees of spatial proximity from NYC. We examined
county data for NYC, Bergen County, NJ (a commuter county across the
Hudson River from NYC), and three major counties [Cook (Chicago)
County in Illinois, Los Angeles County in California, Philadelphia County
in Pennsylvania]. Population-level vital statistics data allowed us to examine
an objective behavioral outcome (divorce) indicative of marital instability
and to observe a major family transition that occurs infrequently in small
samples.To document overt action toward divorce, vital statistics data have
the advantage over survey data of avoiding problems with inaccurate recol-
lections for the dates of separation and divorce (Bumpass & Raley, 2007).
The prospective design provided a baseline before the disaster against
which to compare subsequent divorce rates.
The literature suggests the September 11 disaster will affect divorce in
different directions and with different patterns. The natural disaster and
stress literatures suggest that divorces will increase following September 11
in a pattern characterized by geographic proximity. If geographic prox-
imity and direct exposure explain the results, we would expect the most
516 Journal of Social and Personal Relationships 26(4)
pronounced increases in the divorce rate in NYC, perhaps increases in
Bergen County, NJ where there was loss of life but no material damage,
and no changes in divorce rates in urban counties not directly affected. On
the other hand, the attachment and terrorist management theory literatures
suggest that following a terrorist disaster divorces will decrease locally or
at large, respectively. Furthermore, behavioral responses to the September
11 attack may be mediated by psychological proximity and relative risk
appraisals (Marshall et al., 2007). If so, we would expect to ﬁnd changes in
divorce rates in locales with greater perceived risks. In this respect, Los
Angeles serves as a theoretically discriminating case. Because three of the
four hijacked airplanes were going to Los Angeles, the perceived risk among
citizens of Los Angeles may have been greater than other large urban areas
that were geographically closer to the attacks, like Chicago. Put another
way, the impact of September 11 on divorce rates in Los Angeles has the
potential to distinguish the effects of spatial versus psychological distance
on marital dissolution because those two distances are dissociated for Los
Angeles but overlap for NYC and Bergen County, NJ.A psychological prox-
imity effect would be evidenced by observed changes in Los Angeles, which
was indirectly involved in the attack, but not in Chicago, which was not.
Data and design
The purpose of the current study is to examine how a violent man-made
disaster, the September 11 terrorist attack, affected divorce rates. Our ﬁrst
goal was to determine if the September 11 attack had any detectable impact
on divorce rates in NYC. Our second goal was to determine whether attack-
related changes in divorce were uniformly distributed across other urban
counties, or whether they showed a spatial gradient that was most pro-
nounced in the immediate vicinity of “ground zero.” Our ﬁnal goal was to
determine whether any gradient in dissolution rates was driven primarily
by spatial proximity (i.e., direct disruption of social or economic life) or by
psychological proximity (i.e., greater perceived risk). We assessed these
questions by comparing pre- and post-disaster divorce rates using longitu-
dinal population-level data in counties that varied in geographic distance
from NYC. We examined Bergen County, NJ because it experienced the
greatest number of deaths from September 11 among New Jersey commuter
counties (New Jersey Center for Health Statistics) but did not experience
the physical destruction that NYC did. We examined Cook (Chicago)
County Illinois, Los Angeles County California, and Philadelphia County
(Pennsylvania) because they were the comparison sites for the Schlenger
et al. (2002) study of psychological reactions to the September 11 attack
and varied systematically in spatial distance from NYC.
The current study included two important empirical reﬁnements over
the Cohan and Cole (2002) and the Nakonezny et al (2004) studies. We
examined (i) monthly rather than annual divorces and (ii) rates of divorces
Cohan et al.: Divorce following September 11 517
ﬁled rather than divorces completed. First, the Cohan and Cole study
examined annual rates of divorce and the Nakonezny study examined
annual rates and rates averaged over ﬁve years post-disaster. To increase
temporal precision, the current study examined shorter periods of time –
divorce statistics by month. [See Rodgers, St. John, & Coleman (2005) for
a time series analysis of births by month following terrorist attack.] For
most counties we examined rates of divorces ﬁled by month from January
1995 to December 2005. There were two exceptions. Data on divorces ﬁled
for Philadelphia were available from August 1995 to December 2005, and
data on divorces ﬁled for NYC were only available from January 2000 to
Second, the Cohan and Cole (2002) and the Nakonezny et al. (2004)
studies examined rates of divorces completed. The length of time it takes to
obtain a divorce varies widely by jurisdiction and by couple from months
to years. Therefore when rate of divorces completed is the outcome, it is
not possible to identify when those divorces were initiated relative to the
disaster. For example, an annual measure of divorces completed in 2002
would include divorces initiated prior to September 11 and divorces initi-
ated after September 11, which would mar the integrity of the goal to under-
stand the temporal association between disaster and divorce behavior. To
examine precisely when a couple took action to end their marriage legally
with respect to the onset of disaster, the current study examined rates of
divorces ﬁled. Research on the Marital Instability Index (Booth, Johnson,
& Edwards, 1983) supports the validity of ﬁling for divorce as a measure of
marital instability. People who took overt actions toward divorce, such as
ﬁling for divorce, were more likely to divorce compared to those who
contemplated divorce (Booth, Johnson, White, & Edwards, 1985).
Monthly divorce rates were computed by dividing the number of monthly
divorces ﬁled at the county level by annual population estimates for that
county. County population estimates were obtained from the U.S. Census
Bureau. County-level divorce data were obtained from the following sources:
Circuit Court of Cook County, Los Angeles County Superior Court, Center
for Health Statistics of the New Jersey Department of Health and Senior
Services, New York County Supreme Court, and the Philadelphia Family
Court. The demographic covariate data were obtained from the following
sources: age, education, and race were obtained from the American Com-
munity Survey of the U.S. Census Bureau; income was obtained from the
Bureau of Economic Analyses of the U.S. Department of Commerce; and
unemployment rates were obtained from state labor agencies.
Data analytic strategy
Divorce rates were analyzed as in a previous study of disaster-induced
changes in marital outcomes using Autoregressive Integrated Moving
Average (ARIMA) time series methods that incorporate spatial dose-
response proﬁles (Cohan & Cole, 2002) and adjust for seasonal effects,
secular trends, and autocorrelation. ARIMA regression models analyze the
association between an outcome sequence (e.g., divorce rates over time)
518 Journal of Social and Personal Relationships 26(4)
and a sequence of explanatory variables (e.g., the occurrence of a terrorist
attack). Time series data often show autocorrelation (correlations among
successive residuals) and ARIMA models explicitly model autocorrelation
to prevent correlated residuals from violating the independence assump-
tions underlying statistical inference (Cryer, 1986). Once autocorrelation
among residuals is eliminated, accurate conﬁdence intervals and signiﬁcance
levels can be calculated for regression coefﬁcients summarizing the relation-
ship between outcome incidence and time (i.e., the magnitude of secular
trend, a slow steady change in incidence over time) or between outcome
incidence and a transient event, such as the September 11 terrorist attack.
Under the ARIMA approach,outcome incidence is ﬁrst plotted as a func-
tion of time in order to identify secular trends, outliers, heteroscedasticity,
and any other features of the data that might complicate modeling (e.g.,
changing patterns of secular trend, changes in outcome variability over
time). A preliminary model incorporating any visible secular trend is then
ﬁt to the data, and the residuals are examined for mutual independence
using autocorrelation functions and partial autocorrelation functions (see
Cryer, 1986).If residual autocorrelation is present, two types of parameters
can be added to the model to capture autocorrelation in the deterministic
portion of the model and thereby remove it from the stochastic residuals.
Autoregressive (AR) terms relate outcomes at time tto those one lag back
at t-1 (or two lags back, or more), and are indicated by partial autocorrela-
tion among outcomes. Moving average (MA) terms relate residuals at time
tto those one or more lags back, and are indicated by partial autocorrela-
tion among residuals. To remove secular trends, analysis may focus on the
integrated (I) series, which is obtained by taking the difference between
each observation and the one preceding it (the ﬁrst difference). Because
time series data for each county may show different temporal behavior,
different ARIMA models are appropriate for each county series analyzed.
When reporting results, we describe the ARIMA models in terms of the
number of AR terms, the degree of integration, and the number of MA
terms. Thus an ARIMA (2, 1, 0) model employs two autoregressive (AR)
parameters, is ﬁt to the ﬁrst difference of the series (I), and employs no
moving average terms (MA). We selected as the most appropriate model
that which was most parsimonious (i.e., has the fewest AR or MA terms)
and ﬁt the data well (i.e., no signiﬁcant autocorrelation remained in the
Exploratory analyses of divorce rate time series for several of the coun-
ties studied here revealed two periods of increased divorce rates prior to the
September 11 terrorist attack (Cohan, Cole, & Schoen, 2007). One period
occurred between April and October 1997 and a second between January
and April 2001, which coincided with transient changes in federal immigra-
tion policy. To be cautious and to remove the possible inﬂuence of “policy
spikes” from the underlying secular trends on divorce, we included in our
initial ARIMA models indicator variables capturing those systematic inﬂu-
ences (i.e., one parameter taking a value of 0 prior to April 1997, 1 between
April and October of 1997, and 0 thereafter, and a second indicator taking
Cohan et al.: Divorce following September 11 519
a value of 0 prior to January 2001, 1 between January and April 2001, and
0 thereafter). Inclusion of indicator variables capturing those systematic
effects ensures that (i) their impact is removed from the stochastic series of
residuals, and (ii) subsequent assessment of divorce rates following Septem-
ber 2001 are not confounded by policy-induced disturbances.
Three considerations guided where we would look for changes in rates
of divorces ﬁled. First, we started with September 2001, the month of
greatest impact, given the onset of post-disaster effects is prompt rather
than delayed (Norris et al., 2002). Second, if there was a signiﬁcant effect
for September 2001, we would follow up with an examination of October
2001 and so on for as long as the previous month showed signiﬁcant changes.
Third, we used visual inspection of the time series.
After describing the general behavior of an outcome series with an appro-
priate ARIMA model controlling for federal immigration policy timing, we
examined the effect of the September 11 terrorist attack by adding to the
model an indicator variable that took the value 0 prior to September 2001,
the value 1 during September 2001, and the value 0 thereafter. Indicator
variables representing persisting effects in subsequent months were also
added to gauge the length of time over which the outcome series departed
from its basal behavior following the terrorist attacks. Residual autocorre-
lations were examined following the addition of all monthly indicator vari-
ables to ensure that the ﬁnal model continued to ﬁt the data well. In the
context of the ﬁnalized model, statistical characteristics of each indicator
variable (parameter estimate, standard error, pvalue) were interpreted to
assess the statistical signiﬁcance and effect size of the September 11 terror-
ist attack on marital dissolution rates.
Following basic analysis of the divorce time series for each county studied,
we also carried out “spatial dose-response analyses” and “psychological
dose-response analyses” comparing the magnitude of change in divorce
rates in highly proximal counties versus more distal counties. If there is a
September 11 effect we would expect any signiﬁcant change in NYC to
remain signiﬁcant after controlling for divorce rates in a more distal county.
On the other hand, if some other national sociodemographic trend accounts
for any change in NYC then we would expect signiﬁcant changes to become
non-signiﬁcant after controlling for another urban county. These analyses
took the time series of divorce rates in a proximal county as the primary
outcome while controlling for the parallel time series of divorce rates in a
distal county. This approach is equivalent to the analysis of the difference
in divorce rates in proximal versus distal counties (Cohan & Cole, 2002).
To the extent that the September 2001 indicator variable remained statis-
tically signiﬁcant after controlling for correlated changes in distal county
divorce rates, we concluded that a signiﬁcant dose-response relationship
existed (i.e., the magnitude of September 11-associated change in divorce
rates in the proximal county signiﬁcantly exceeded that observed in the distal
county). In initial analyses, NYC served as the proximal county, and varia-
tions in the selection of the distal counties (e.g., Philadelphia, Los Angeles)
served to distinguish between the effects of spatial versus psychological
520 Journal of Social and Personal Relationships 26(4)
proximity. Subsequent pair-wise comparisons among other counties (e.g.,
analysis of the Los Angeles divorce series, controlling for that of Philadel-
phia) served to further reﬁne conclusions from the primary analyses.
We used the SAS PROC TIMEPLOT to examine the data and PROC
ARIMA to estimate models by conditional least squares.For a more detailed
description of ARIMA modeling, see Cryer (1982). All signiﬁcance levels
were based on two-tailed tests.
Did the September 11 attack affect divorce rates in New York
Monthly rates of divorces ﬁled in New York County (Manhattan) between
January 2000 and December 2005 are shown in Figure 1. In ARIMA time
series analyses, results revealed a statistically signiﬁcant decrease in NYC
divorce ﬁlings during September 2001 (p= .0013). The magnitude of change
was substantial, with ﬁling rates falling by –.24 per 1000 residents (32.4%)
below the value expected based on background secular trends as estimated
by the best-ﬁtting ARIMA (2,1,1) model. Table 1 presents results of time
series analyses for all counties analyzed. First, sub-normal divorce ﬁlings
persisted only for September 2001. October 2001 rates did not differ sig-
niﬁcantly from that expected on the basis of secular trends and seasonal
variation (p= .56). Thus, NYC divorce rates dropped in a signiﬁcant and
punctate fashion in the immediate aftermath of the September 2001 terror-
ist attacks. Second, inspection of the New York County time series sug-
gested that after an initial decline in divorces ﬁled in September 2001, there
might be an increase in divorces in 2002. Because we did not have an a priori
reason or previous empirical evidence to examine a portion of the year, we
tested for change in divorce rates across all of 2002 by aggregating the 2002
monthly data. There was a non-signiﬁcant up tick in divorces in New York
County in 2002, (p= .1437, +.0468, SD = .0315).
Although the pronounced transient decline in NYC divorce rates could
be predicted by psychological theories, an alternative explanation is the
effects of civil service disruption. The New York County Supreme Court
where divorces are ﬁled was closed for one week after September 11 (James
McElligot, Chief Matrimonial Clerk, New York County Supreme Court,
personal communication, August 2005). Although the 32.4% decline in
divorce ﬁlings observed exceeds the 23.3% which would normally accrue
in NYC during one week, it is conceivable that lack of access or opportunity
could contribute to the decline observed.To distinguish whether the decline
in divorces was attributable to psychosocial inﬂuences of the disaster versus
civil service disruption, we carried out parallel analyses in Bergen County,
NJ, a commuter county for NYC where many of the dead had resided.
Although disaster exposure was similar to NYC residents, this New Jersey
county suffered no disruption in civil services.ARIMA time-series analysis
of monthly divorce ﬁlings in Bergen County, NJ showed a statistically
Cohan et al.: Divorce following September 11 521
522 Journal of Social and Personal Relationships 26(4)
Rate of divorces ﬁled by month 2000 to 2005 for New York County (Manhattan)
Month and year
Cohan et al.: Divorce following September 11 523
signiﬁcant decrease in September 2001 (p= .021), followed by a slight (but
non-signiﬁcant) return toward basal rates in October 2001 (p= .191). Thus,
transient decreases in divorce ﬁlings in close proximity to the September
2001 terrorist attack were not solely attributable to the effects of civic
disruption on ﬁling opportunity.
Divorce rates in other urban counties
To determine how broadly divorce rates were affected in the aftermath of
the terrorist attacks, we examined county divorce ﬁlings for Cook, Los
Angeles, and Philadelphia Counties. As shown in Table 1, Philadelphia and
Los Angeles showed marked and statistically signiﬁcant declines in
divorces ﬁled in September 2001 (p= .013 and .016, respectively). Divorce
rates showed a subsequent trend upward in October 2001, and this effect
was statistically signiﬁcant in Los Angeles (p= .009) but not in Philadel-
phia (p= .236). The results for Los Angeles are consistent with the hypoth-
esis that perturbations in divorce following the terrorist attacks were driven
by psychological proximity rather than direct physical effects of the attacks.
Contrary to a strict metropolitan effect, there was no change in Cook
County (Chicago) for the rate of divorces ﬁled in September 2001.
We conducted additional tests to evaluate the veracity of the results. To
rule out the alternative explanation that sociodemographic conditions
drove the results and to test internal validity, we added ﬁve demographic
covariates to our analyses that are related to divorce, following those
included in the Nakonezny et al. (2004) study of divorce after the Oklahoma
City bombing. After controlling for trends in age, education, income, race,
and unemployment, September divorce declines remained statistically signi-
ﬁcant in all cases in which they were originally signiﬁcant (Bergen County,
New Jersey; Los Angeles County, California; New York County, New York;
Philadelphia County, Pennsylvania); all ps < .05) and non-signiﬁcant in
Results of ARIMA analyses for September 2001 divorce rates
Location Model Raw (SE) Zp
New York County, NY 2,1,0 –.243 (.092) –2.65 .010
Bergen County, NJ 3,1,0 –.077 (.033) –2.34 .021
Cook County, IL (Chicago) 2,0,2 –.009 (.015) –0.56 .579
Los Angeles County, CA 3,1,2 –.046 (.019) –2.44 .016
Philadelphia County, PA 2,0,0 –.041 (.016) –2.54 .013
New York (controlling for Philadelphia) 2,1,0 –.236 (.084) –2.83 .006
New York (controlling for Los Angeles) 2,1,0 –.249 (.091) –2.73 .008
Los Angeles (controlling for Philadelphia) 3,1,2 –046 (.018) –2.45 .016
Philadelphia (controlling for Los Angeles) 3,1,0 –.034 (.018) –1.87 .065
Note: The numbers in the Model column represent the number of Autoregressive terms, degree
of Integration (differencing) of the original time series, and number of Moving Average terms.
Cook County (p= .32). To examine potential threats to statistical conclusion
validity, residuals were veriﬁed to show non-signiﬁcant autocorrelation (by
Ljung-Box test over 1–24 month lags), heteroscedasticity (by Portmanteau
Q-test), and no departure from normal distribution (by Bera-Jarque test)
following the estimation of each model.
Divorce rates as a function of spatial and psychological proximity
Signiﬁcant decreases in divorces ﬁled in Los Angeles and Philadelphia in
September 2001 suggest that indirect exposure appeared to play a central
role in post-attack divorce dynamics.We next sought to determine whether
such experiences propagated as a function of spatial proximity or psycho-
logical proximity. We gauged the effects of spatial proximity by comparing
the intensity of September 2001 divorce perturbations in NYC versus the
progressively more distant cities of Philadelphia and Los Angeles. As
expected,ARIMA models assessing each pair-wise difference between cities
showed the magnitude of perturbation in NYC was signiﬁcantly greater than
that in Philadelphia (analysis of NYC rates controlling for Philadelphia; p=
.001) and Los Angeles (NYC controlling for Los Angeles; p= .008; see
Table 1), supporting the view that proximity to the terrorist attack rather
than a broader sociodemographic trend accounted for the change.
The pattern of results showing signiﬁcant change in September 2001
divorces for Philadelphia but not Chicago supports a spatially-driven
psychological gradient in which the impact of the terrorist attacks decays
in proportion to geographic distance from the site of the attacks. However,
the signiﬁcant change in divorces in Los Angeles is contrary to a purely
geographic effect and is consistent with a psychologically-driven gradient
in which psychological proximity to the attack determines the magnitude
of behavioral impact. Philadelphia and Los Angeles were both removed
from the site of the direct attack and thus showed a less dramatic impact
than NYC. However, the relative magnitude of divorce perturbations in
those two cities provides an opportunity to discriminate the effects of
spatial gradients from psychological symbolic gradients because they show
different relative ordering in spatial versus symbolic distance. Philadelphia
is relatively close to NYC in geographic terms (about 100 miles away), and
a strictly geographic gradient would imply greater perturbation in divorce
rates in Philadelphia than in Los Angeles to the extent that relative prox-
imity increases perceived risk. On the other hand, Los Angeles had a unique
risk and was “closer”to the terrorist attacks than Philadelphia because three
of the hijacked planes were destined for Los Angeles, whereas none were
going to Philadelphia.Thus,comparison of the relative magnitude of divorce
perturbations in Los Angeles and Philadelphia provides an opportunity to
determine whether the magnitude of impact decays in direct proportion to
geographic distance, in which case Philadelphia rate changes would exceed
Los Angeles rate changes, or in direct proportion to psychological distance,
in which case Los Angeles rate changes would exceed Philadelphia rate
change. Consistent with the latter hypothesis, the empirical magnitude of
divorce perturbation in Los Angeles signiﬁcantly exceeded that observed
524 Journal of Social and Personal Relationships 26(4)
in Philadelphia (see Table 1).The percent alteration in Philadelphia divorce
rates was smaller than that of Los Angeles in relative (15.4% drop from
historical September norms vs. 16.6%, respectively) and in absolute terms
(–.03 ﬁlings/1000 married individuals vs. –.06, respectively). Consistent with
the hypothesis that Los Angeles divorce rates in September 2001 contained
a distinct dynamic component not present at the same time in Philadelphia,
ARIMA analysis of Los Angeles rates continued to indicate a signiﬁcant
decline in September 2001 after controlling for concurrent changes in
Philadelphia rates (p= .016). Philadelphia divorce rates did not provide a
signiﬁcant increment to predictive power above and beyond the basic effects
of the September 2001 indicator variable (p= .167). In contrast, changes in
Philadelphia divorce rates were rendered marginal by controlling statisti-
cally for Los Angeles rates (p= .065), whereas Los Angeles rates remained
a signiﬁcant predictor of the Philadelphia time series (p= .012). These
results imply that the impact of the terrorist attacks on marital dissolution
rates decayed more so as a function of the strength of the psychological
proximity to the attack than as a function of geographic distance.
To study the effect of an exogenous shock on marriage, the current study
examined the effect of the September 11 terrorist disaster on an objective
indicator of marital instability, rates of divorces ﬁled by month, for NYC.
The time series methods adjusted for seasonal effects, secular trends, and
autocorrelation. To examine whether effects radiated beyond “ground zero”
according to spatial proximity or psychological proximity, we also examined
Bergen County, NJ (a commuter county for NYC) and three metropolitan
counties of varying spatial distance (Cook County in Illinois, Los Angeles
County in California, and Philadelphia County in Pennsylvania). Two pre-
vious studies of disaster and divorce make different predictions about the
direction of change in rates of divorce. Following natural disaster, divorces
increased according to geographic proximity (Cohan & Cole, 2002), whereas
following a man-made bombing disaster, divorces decreased according to
geographic proximity and metropolitan status (i.e., psychological proxim-
ity; Nakonezny et al., 2004).
The results of the current study are consistent with the observation that
“. . . a disaster is more than an individual-level event, but it is also a
community-level event with potential psychological consequences even for
those persons who experience no direct losses” (Norris, 2002, p. 310). In
NYC there was a signiﬁcant decrease in the rate of divorces ﬁled in Sept-
ember 2001. Similarly, there was a signiﬁcant decrease in divorces ﬁled in
September 2001 in Bergen, Philadelphia, and Los Angeles Counties. There
were no changes in Cook County (Chicago). Similar to a study of divorce
following the 1995 Oklahoma City bombing, people with geographic and
psychological proximity to the September 11 attack, who would have ﬁled
for divorce in September 2001 had the disaster not occurred, decided to
Cohan et al.: Divorce following September 11 525
postpone or forego a divorce. The consistency across locales in the timing
and direction of the change in divorces ﬁled lends credence to the robust-
ness of the effect.
The decline in divorces following terrorist disaster is in contrast to the
increase in divorces following natural disaster (Cohan & Cole, 2002) and
suggests there were different dynamics for marital stability for different
disasters. Hurricane Hugo involved massive physical damage that required
signiﬁcant rebuilding but minimal loss of human life (32 deaths in South
Carolina; South Carolina Governor’s Ofﬁce, 1991) because the storm was
predicted and people evacuated. When the context of the disaster centered
on chronic post-disaster rebuilding stress, marriages were more likely to
unravel. In contrast, the September 11 disaster killed approximately 3000
people without warning by a malicious human agent. When the context of
the disaster centered on death, fewer people ﬁled for divorce.
An important ﬁnding for the disaster literature is that the effect of mass
violence can move people to alter their behavior, even through indirect
exposure to the disaster. The decline in divorce rates in NYC and Bergen
County, NJ are consistent with a geographic proximity effect. The relative
geographic proximity of Philadelphia to NYC (e.g., easy access by road or
train, personal connections) and the unique threat of the crash of United
ﬂight 93 in Pennsylvania may have contributed to increased perceived risk
and the subsequent drop in divorces ﬁled there. The signiﬁcant result for
Philadelphia together with the nonsigniﬁcant result for Chicago suggests
that the impact of September 11 on divorce dissipated with spatial distance.
However, the nonsigniﬁcant result for Chicago coupled with the signiﬁcant
change in Los Angeles suggests there was also a psychological proximity
effect and that it was more nuanced than simply perceived risk due to
metropolitan status. For the disaster literature, the decline in divorces ﬁled
in Los Angeles County in September 2001, on the opposite side of the
country from NYC, is a remarkable exception to the geographic proximity
effect found with most disasters. The results suggest that a mass violence
disaster can affect behavior when the exposure is vicarious and the prox-
imity is psychological (Galea & Resnick, 2005). Because three of the
hijacked airplanes were going to Los Angeles, the threat of death was
presumably higher for that county than a geographically closer county like
Chicago.The pattern of results in the current study of mass violence corrob-
orates an earlier study on divorce following the Oklahoma City bombing
(Nakonezny et al., 2004) such that when there was speciﬁc intent to kill
large numbers of people, divorces declined in areas geographically
proximal as well as areas that were geographically distant but psychologi-
cally proximal. Thus vulnerability to mass violence deaths, with or without
accompanying infrastructure destruction, was associated with a decrease in
divorces ﬁled, divorces that would have been ﬁled had the September 11
attack not occurred.
Examination of monthly rates of divorces ﬁled in the current study yielded
greater precision compared to two previous studies of disaster and divorce
that examined annual rates of divorces completed, with implications for
526 Journal of Social and Personal Relationships 26(4)
reﬁning theory (Cohan & Cole, 2002; Nakonezny et al., 2004). We found a
signiﬁcant one-month decline in divorces.The decline in NYC at the height
of the trauma is consistent with an attachment theory interpretation. Our
pattern of results is similar to Bowlby’s observation that family members
will stay in close proximity for “days or weeks” following a disaster because
afﬁliation is comforting during disaster (Bowlby, 1973, p. 167). Marital
preservation appears to be an immediate response to mortal threat, but
relaxes once the threat is less acute. Under conditions of extreme stress,
uncertainty, and threat, people maintain the status quo and refrain from
making a major life change. “Goals, during crisis periods, are easily rank
ordered and commonly focused upon survival or sustenance of the social
system and/or individuals within it” (Lipman-Blumen, 1975, p. 891). Inten-
tionally or unintentionally, humans may take quick measures to conserve
resources following a severe exogenous shock. In a parallel ﬁnding that
reﬂects unconscious efforts to conserve resources under stress, the odds of
a male birth fell signiﬁcantly in NYC and California in the three months
after the September 11 attacks, consistent with other studies of major
stressors and the human sex ratio in the population (Catalano, Bruckner,
Marks,& Eskenazi, 2006:Catalano, Bruckner, Gould, Eskenazi,& Anderson,
2005). From an evolutionary perspective, fetal loss of weak males may
increase the chance that females will survive to reproduction in a stressful
The decline in divorces in Philadelphia and Los Angeles was consistent
with the terror management theory position that reminders of death can
lead to actions to counter the existential fear of death. Increased commit-
ment to a romantic partner may be one way to mitigate those fears. How-
ever, the length of the signiﬁcant decline appears more like the attachment
theory idea of maintaining security and stability rather than the terror
management theory idea of increased commitment, which implies a longer
effect. Furthermore, the non-signiﬁcant effect in Chicago,which was exposed
to September 11-related images of death, runs counter to the terror manage-
ment theory idea that reminders of death should bolster marriages at large.
It suggests that an additional dose of perceived risk was required for change
in divorce behavior in locales without direct exposure. For example, Los
Angeles was the destination location for three of the four airplanes and
Philadelphia is relatively close to NYC and the crash of United ﬂight 93.
There are several cautions to consider when interpreting the results. First,
it is difﬁcult to tell from the population-level data how long the September
11 effect on divorce persisted for the people who did not ﬁle for divorce in
September 2001 and whether they decided to forego divorce permanently
or temporarily. For people whose divorces were deferred temporarily, it is
logical to expect that those divorces ﬁled at a later time were spread out
over a number of months, which would make it impossible to detect a
particular rebound with the time series analyses. In Los Angeles there was
a signiﬁcant increase in divorces in October 2001, which probably repre-
sented the rebound from the divorces deferred in September 2001. So in
Los Angeles it appears that couples temporarily postponed divorces during
Cohan et al.: Divorce following September 11 527
“the brunt of the storm” as predicted by attachment theory rather than
preserved their marriage through a disaster-inspired shift in long-term
commitment, as would be suggested by terror management theory. Second,
individual-level data are necessary to test whether the processes we posited
were the mechanisms for the observed population-level declines in divorce.
Alternative explanations could include urban chaos or downward social
comparison (Buunk, 2006; Taylor, Buunk, Collins, & Reed, 1992), that is,
perceiving one’s distressed marriage as better than nothing in light of the
young, happy families that were devastated by the disaster. Third, we con-
ducted a number of analyses to show that the results were a consequence
of the September 11 terrorist attack. Yet the possibility remains with corre-
lational data that an unexamined variable caused the results.
The September 11 terrorist attack was a disaster like no other in the US.
The study points to the power of a manmade disaster that focused on inten-
tional killing to alter behavior among people directly and indirectly
affected.The pattern of results indicates that the threat of death, even when
experienced vicariously, moved people to alter their path and refrain from
ﬁling for divorce. Future research can examine how the context of disaster
may affect the interpersonal dynamics in marriage to inﬂuence whether
some couples stay together “for better and for worse” following disaster.
Beach, S. R. H., & O’Leary, K. D. (1993). Dysphoria and marital discord: Are dysphoric indi-
viduals at risk for marital adjustment? Journal of Marital and Family Therapy, 19, 355–368.
Booth, A., Johnson, D., & Edwards, J. N. (1983). Measuring marital instability. Journal of
Marriage and the Family, 45, 387–394.
Booth, A., Johnson, D. R., White, L. K., & Edwards, J. N. (1985). Predicting divorce and
permanent separation. Journal of Family Issues, 6, 331–346.
Bowlby,J. (1982). Attachment and loss:Volume 1. Attachment. New York: Basic Books. (Original
work published 1969)
Bowlby, J. (1973). Attachment and loss: Volume 2. Separation: Anxiety and anger. New York:
Bowlby,J. (1988). A secure base: Clinical applications of attachment theory.London: Routledge.
Bumpass, L., & Raley, K. (2007). Measuring separation and divorce. In S. L. Hofferth & L. M.
Casper (Eds.), Handbook of measurement issues in family research (pp. 125–143). Mahwah,
NJ: Lawrence Erlbaum.
Buunk, A. P. (2006). Responses to a happily married other: The role of relationship satisfac-
tion and social comparison orientation. Personal Relationships, 13, 397–409.
Catalano, R., Bruckner, T., Gould, J., Eskenazi, B., & Anderson, E. (2005). Sex ratios in Cali-
fornia following the terrorist attacks of September 11, 2001. Human Reproduction, 20,
Catalano, R., Bruckner, T., Marks, A. R., & Eskenazi, B. (2006). Exogenous shocks to the
human sex ratio: The case of September 11, 2001 in New York City. Human Reproduction,
528 Journal of Social and Personal Relationships 26(4)
Cohan, C. L., & Cole, S. (2002). Life course transitions and natural disaster: Marriage, birth,
and divorce following Hurricane Hugo. Journal of Family Psychology, 16, 14–25.
Cohan, C. L., Cole, S., & Schoen, R. (2007). Federal immigration policy and change in divorce
rates. Manuscript in preparation.
Cryer, J. D. (1986). Time series analysis. Boston, MA: PWS-Kent Publishing Company.
Davila, J., & Bradbury, T. N. (1998). Psychopathology and the marital dyad. In L. L’Abate
(Ed.), Family psychopathology:The relational roots of dyfunctional behavior (pp. 127–157).
New York: Guilford Press.
Dillon, N., & Schoolman, J. (2002,April 22). City can expect job losses to grow. Daily News, 30.
Florian, V., Mikulincer, M., & Hirschberger, G. (2002). The anxiety-buffering function of close
relationships: Evidence that relationship commitment acts as a terror management
mechanism. Journal of Personality and Social Psychology, 82, 527–542.
Galea, S., Ahern, J., Resnick, H., Kilpatrick, D., Bucuvalas, M., Gold, J., et al. (2002). Psycho-
logical sequelae of the September 11 terrorist attacks in New York City. New England
Journal of Medicine, 346, 982–987.
Galea, S., & Resnick, H. (2005). Posttraumatic stress disorder in the general population after
mass terrorist incidents: Considerations about the nature of exposure. CNS Spectrums, 10,
Hanson, R. F., Kilpatrick, D. G., Freedy, J. R., & Saunders, B. E. (1995). Los Angeles County
after the 1992 civil disturbances: Degree of exposure and impact on mental health. Journal
of Consulting and Clinical Psychology, 63, 987–996.
Hazan, C.,& Shaver, P. R. (1994).Attachment as an organizational framework for research on
close relationships. Psychological Inquiry, 5, 1–22.
Hill, R., & Hansen, D.A. (1962). Families in disaster. In G. W. Baker & D. W. Chapman (Eds.),
Man and society in disaster (pp. 185–221). New York: Basic Books.
Hirschberger, G., Florian,V., & Mikulincer, M.(2003). Strivings for romantic intimacy following
partner complaint or partner criticism: A terror management perspective. Journal of Social
and Personal Relationships, 20, 675–687.
Johnson, K. (2002, September 8).Threats and responses: The families; in bereavement,pioneers
on a lonely trail. The New York Times, pp. A1, A20, A21.
Karney,B. R., & Bradbury,T. N.(1995).The longitudinal course of marital quality and stability:
A review of theory, method, and research. Psychological Bulletin, 118, 3–34.
Lipman-Blumen, J. (1975). A crisis framework applied to macrosociological family changes:
Marriage, divorce, and occupational trends associated with World War II. Journal of
Marriage and the Family, 37, 889–902.
Marshall, R.D.,Bryant, R.A., Amsel,L., Suh,E. J., Cook,J. M., & Neria,Y.(2007).The psychol-
ogy of ongoing threat: Relative risk appraisal, the September 11 attacks, and terrorism-
related fears. American Psychologist, 62, 304–316.
McLeod, J. D. (1994). Anxiety disorders and marital quality. Journal of Abnormal Psychology,
Mikulincer, M., & Florian, V. (2000). Exploring individual differences in reactions to mortality
salience: Does attachment style regulate terror management mechanisms? Journal of
Personality and Social Psychology, 79, 260–273.
Mikulincer, M., Florian, V., & Hirschberger, G. (2003). The existential function of close rela-
tionships: Introducing death into the science of love. Personality and Social Psychology
Review, 7, 20–40.
Mikulincer, M., Florian, V., & Hirschberger, G. (2004). The terror of death and the quest for
love: An existential perspective on close relationships. In J. Greenberg, S. L. Koole, &
T. Pyszczynski (Eds.), Handbook of experimental existential psychology (pp. 287–304). New
Nakonezny, P. A., Reddick, R., & Rodgers, J. L. (2004). Did divorces decline after the
Oklahoma City bombing? Journal of Marriage and Family, 66, 90–100.
Ness, C., Schevitz, T., & Aguila, J. (2001, September 12). L.A. in shock – 3 of the 4 crashed
planes were bound for LAX, San Francisco Chronicle, A24.
Cohan et al.: Divorce following September 11 529
New Jersey Center for Health Statistics. New Jersey Resident Deaths Due to the Terrorist Attacks
on September 11, 2001. http://www.state.nj.us/health/chs/sept11mortstats.htm, accessed 11/
Norris, F. H. (2002). Disasters in urban context. Journal of Urban Health, 79, 308–314.
Norris, F. H., Friedman, M. J., Watson, P., Byrne, C. M., Diaz, E., & Kaniasty, K. (2002). 60,000
disaster victims speak: Part I. An empirical review of the empirical literature, 1981–2001.
Psychiatry, 65, 207–239.
Psyzcynski, T., Greenberg, J., & Solomon, S. (1999). A dual process model of defense against
conscious and unconscious death-related thoughts: An extension of terror management
theory. Psychological Review, 106, 835–845.
Pyszczynski, T., Solomon, S., & Greenberg, J. (2003). In the wake of 9/11: The psychology of
terror. Washington DC: American Psychological Association.
Rodgers, J. L., St. John, C. A., & Coleman, R. (2005). Did fertility go up after the Oklahoma
City bombing? An analysis of births in metropolitan counties in Oklahoma, 1990–1999.
Demography, 42, 675–692.
Rubonis, A. V., & Bickman, L. (1991). Psychological impairment in the wake of disaster: The
disaster-psychopathology relationship. Psychological Bulletin, 109, 384–399.
Schlenger,W. E., Caddell,J.M., Ebert, L., Jordan, B. K., Rourke, K. M.,Wilson, D.,et al. (2002).
Psychological reactions to terrorist attacks: Findings from the National Study of Americans’
Reactions to September 11. Journal of the American Medical Association, 288, 581–588.
Schuster, M.A., Stein, B. D., Jaycox, L. H., Collins, R. L., Marshall, G. N., Elliott, et al. (2001).
A national survey of stress reactions after the September 11, 2001, terrorist attacks. New
England Journal of Medicine, 345, 1507–1512.
Silver, R. C., Holman, E. A., McIntosh, D. N., Poulin, M., & Gil-Rivas, V. (2002). Nationwide
longitudinal study of psychological responses to September 11. Journal of the American
Medical Association, 288, 1235–1244.
Solomon, S., Greenberg, J., & Pyszcynski, T. (1991).A terror management theory of social beha-
vior: The psychological functions of self esteem and cultural worldviews. In L. Berkowitz
(Ed.), Advances in experimental social psychology (Vol. 24, pp. 93–159). New York:Academic.
South, S. J. (1985). Economic conditions and the divorce rate: A time-series analysis of the
postwar United States. Journal of Marriage and the Family, 47, 31–41.
South Carolina Governor’s Ofﬁce (1991). An analysis of the damage and effects of Hurricane
Hugo and status of recovery one year later. Division of Intergovernmental Relations.
Stein, B. D., Elliott, M. N., Jaycox, L. H., Collins, R. L., Berry, S. H., Klein, D. J., et al. (2004).
A national longitudinal study of the psychological consequences of the September 11, 2001
terrorist attacks: Reactions, impairment, and help-seeking. Psychiatry, 67, 105–117.
Taylor, S. E., Buunk, B. P., Collins, R. L., & Reed, G. M. (1992). Social comparison and afﬁlia-
tion under threat. In L. Montada, S. Filipp, & M. J. Lerner (Eds.), Life crises and experiences
of loss in adulthood (pp. 213–227). Hillsdale, NJ: Lawrence Erlbaum.
Vlahov,D. (2002). Urban disaster: A population perspective. Journal of Urban Health, 79, 295.
Vlahov, D., Galea, S., Resnick, H., Ahern, J., Boscarino, J. A., Bucuvalas, M., et al. (2002).
Increased use of cigarettes, alcohol, and marijuana among Manhattan, New York, residents
after the September 11th terrorist attacks. American Journal of Epidemiology,155, 988–996.
Wisman, A., & Koole, S. L. (2003). Hiding in the crowd: Can mortality salience promote afﬁl-
iation with others who oppose one’s worldviews? Journal of Personality and Social Psychol-
ogy, 84, 511–526.
Woodberry, W. (2002, April 25). No 9/11 aid for 9,000 who lost aviation jobs. Daily News,6.
Yednak, C. & Mellen, K. (2001, September 12). Chicago’s high-rise landmarks take security
precautions, Chicago Tribune, 14.
Yeung, W. J., & Hofferth, S. L. (1998). Family adaptations to income and job loss in the U.S.
Journal of Family and Economic Issues, 19, 255–283.
530 Journal of Social and Personal Relationships 26(4)