Why did they "choose" to stay? Perspectives of Hurricane Katrina observers and survivors.

Abstract

Models of agency--powerful implicit assumptions about what constitutes normatively "good" action--shaped how observers and survivors made meaning after Hurricane Katrina. In Study 1, we analyzed how 461 observers perceived survivors who evacuated (leavers) or stayed (stayers) in New Orleans. Observers described leavers positively (as agentic, independent, and in control) and stayers negatively (as passive and lacking agency). Observers' perceptions reflected the disjoint model of agency, which is prevalent in middle-class White contexts and defines "good" actions as those that emanate from within the individual and proactively influence the environment. In Study 2, we examined interviews with 79 survivors and found that leavers and stayers relied on divergent models of agency. Leavers emphasized independence, choice, and control, whereas stayers emphasized interdependence, strength, and faith. Although both leavers and stayers exercised agency, observers failed to recognize stayers' agency and derogated them because observers assumed that being independent and in control was the only way to be agentic.

Full text

Page 1

Research Article
Why Did They ‘‘Choose’’
to Stay?
Perspectives of Hurricane Katrina Observers and Survivors
Nicole M. Stephens,1MarYam G. Hamedani,1Hazel Rose Markus,1Hilary B. Bergsieker,2and
Liyam Eloul1
1Stanford University and2Princeton University
ABSTRACT—Models of agency—powerful implicit assump-
tions about what constitutes normatively ‘‘good’’ action—
shaped how observers and survivors made meaning after
Hurricane Katrina. In Study 1, we analyzed how 461 ob-
servers perceived survivors who evacuated (leavers) or
stayed(stayers)inNewOrleans.Observersdescribedleav-
ers positively (as agentic, independent, and in control)
and stayers negatively (as passive and lacking agency).
Observers’ perceptions reflected the disjoint model of
agency, which is prevalent in middle-class White contexts
and defines ‘‘good’’ actions as those that emanate from
within the individual and proactively influence the envi-
ronment. In Study 2, we examined interviews with 79
survivors and found that leavers and stayers relied on
divergent models of agency. Leavers emphasized inde-
pendence, choice, and control, whereas stayers empha-
sized interdependence, strength, and faith. Although both
leavers and stayers exercised agency, observers failed to
recognize stayers’ agency and derogated them because
observers assumed that being independent and in control
was the only way to be agentic.
Etched in Americans’ collective memory of Hurricane Katrina
are images ofsurvivors standing on rooftops awaiting help. With
these images came the claim that survivors failed to take ap-
propriate actions. Responding to the rising death toll in New
Orleans, Federal Emergency Management Agency Director
Michael Brown said, ‘‘That’s going to be attributable a lot to
people who . . . . chose not to leave’’ (CNN Weather, 2005).
Similarly, Secretary of Homeland Security Michael Chertoff
explained, ‘‘Officials called for a mandatory evacuation. Some
people chose not to obey that order. That was a mistake on their
part’’ (CNN Transcripts, 2005). Brown and Chertoff assumed
that survivors who stayed ‘‘chose’’ not to evacuate and were
therefore to blame for their suffering. We suggest that these
commonreactionstosurvivorsweregroundedinandlegitimized
by a powerful, yet often tacit, set of assumptions about what
constitutes normatively ‘‘good’’ action—a model of agency. As
Brown’s and Chertoff’s comments suggest, this implicit model
led observers to interpret action in a context-specific way that
fostered a lack of empathy for survivors who stayed.
Attributiontheoryandsystem-justifyingtendencies—suchas
victim blaming, stereotyping, and belief in a just world—pro-
vide useful frameworks for understanding observers’ responses
to survivors (Adams, O’Brien, & Nelson, 2006; Napier, Man-
disodza, Andersen, & Jost, 2006). Attribution theory predicts
that observers will locate the causes of survivors’ divergent
outcomes in their individual attributes (Ross & Nisbett, 1991),
butitdoesnotanswerthequestionofwhyobservers’perceptions
took the particular forms they did. Highlighting another im-
portant feature of the explanatory sequence, the research we
report here addresses not observers’ explanations of the causes
of survivors’ behavior (i.e., attribution), but instead how ob-
servers perceive and make sense of survivors’ actions. In two
studies, we contrasted observers’ and survivors’ perspectives to
illuminate the contextually derivedmodels ofagency that shape
howpeople makesense ofbehaviorandwhatpeopleperceive as
sensible, culturally appropriate action.
We suggest that observers’ responses to Katrina survivors
were predominantly grounded in the disjoint model of agency—
the most prevalent model in mainstream middle-class White
contexts (Markus, Uchida, Omoregie, Townsend, & Kitayama,
2006; Savani, Markus, & Conner, 2008). The disjoint model
assumes that agency emanates from within the individual and
defines ‘‘good’’ actions as those that influence the environment
Address correspondence to Nicole Stephens, Stanford Univer-
sity, Department of Psychology, Stanford, CA 94305-2310, e-mail:
nstephen@stanford.edu.
PSYCHOLOGICAL SCIENCE
878
Volume 20—Number 7 Copyright r 2009 Association for Psychological Science

Page 2

accordingtoindividualmotives,goals,andpreferences(Markus
& Kitayama, 2003). According to this model, a good Hurricane
Katrinasurvivorshouldhaveinfluencedthesituation,overcome
situational constraints, and—through independence, choice,
and control—found a way to evacuate.
The disjoint model is not the only model of agency. As a
growing literature on culture and agency has revealed, there are
many ways toact in and respond tothe world (Holland&Quinn,
1987;Morling, Kitayama, &Miyamoto,2002; Morris, Menon,&
Ames, 2001). Defined in a socioculturally neutral manner,
agency is acting in the world (Markus & Kitayama, 2003) and
need not—as is often assumed in Western middle-class con-
texts—involve influence or a commitment to an individualist
modelofthepersoninwhichagencyisaforcethatderivessolely
from within the individual (Gould, 1999). Agency can also in-
volve adjusting the self to the world. Prior research has identi-
fied a conjoint model of agency, which assumes that agency is
‘‘responsive to obligations and expectations of others, roles, and
situations’’(Markus &Kitayama, 2003,p.7)and defines ‘‘good’’
actions as those that adjust to the environment and promote
interdependence with other people. Although a conjoint model
isoftenassociatedwithEastAsiancontexts(Miller,2003),some
elements of this type of model, such as an emphasis on con-
necting with others, also pervade working-class American
contexts (Markus, Ryff, Curhan, & Palmersheim, 2004; Ste-
phens, Markus, & Townsend, 2007).1
SOCIAL CLASS AND AGENCY
Contextsthatarestratifiedbysocialclassandrace,andthatvary
substantially intheir resources, provide one important source of
models of agency (Markus & Kitayama, 2003). Survivors who
evacuated prior to Katrina (leavers) lived in primarily middle-
class White contexts, whereas survivors who stayed (stayers)
lived in primarily working-class Black contexts (Dyson, 2006).
Comparedwithstayers,leavershadmoreeducationandincome,
greater access to news, more reliable transportation, and more
geographically extended social networks (Lieberman, 2006).
Given the influence-enabling resources (i.e., material ad-
vantages and cultural capital, including knowledge, skills, and
advantages based on societal status) associated with middle-
class White contexts (Bourdieu & Passeron, 1977/1992), we
hypothesized that leavers’ firsthand accounts of their experi-
ences would emphasize choice, independence, and control—
themes associated with a disjoint model. By contrast, people in
working-class Black contexts often lacked the necessary re-
sources to evacuate and effectively enact a disjoint model. De-
spite these constraints, we hypothesized that stayers were not
passive but agentic (i.e., acting in the world), in ways that were
appropriate to their contexts. Building on prior research, we
anticipated that narratives of stayers would reflect elements of a
conjoint model, including an emphasis on interdependence and
connection with others (Nobles, 1991; Stephens et al., 2007), as
well as emphases on being strong and maintaining faith in God
(Ryff, Singer, & Palmersheim, 2004; Snibbe & Markus, 2005).
Hurricane Katrina provides an opportunity to examine how
people in different contexts make meaning in response to the
same historically significant event. We present two studies that
illuminate Katrina observers’ and survivors’ models of agency.
In Study 1, we analyzed how two samples of observers—relief
workers and lay observers—perceived leavers and stayers. Giv-
entheprevalenceofthedisjointmodelinmainstreamAmerican
contexts,wepredictedthatbothsamplesofobserverswouldrely
on the disjoint model, and thus perceive leavers as influencing
agents and stayers as lacking agency.
In Study 2, we examined survivors’ accounts of their own ex-
periences. Because stayers’ and leavers’ contexts differed sub-
stantiallyinresourcesandopportunitiesforaction,wepredicted
that leavers’ accounts would emphasize themes associated with
disjoint agency, and stayers’ accounts would emphasize themes
associated with conjoint agency. Thus, we expected to find that
leavers and stayers were both agentic, but in different ways, and
that observers derogated stayers because they assumed that in-
fluencing the environment was the only way to be agentic.
STUDY 1
For Study 1, we recruited observers to complete an on-line sur-
vey. If observers view survivors through the disjoint model, they
should view leavers, who conformed to this model, as sensible
and influencing agents, and they should view stayers, who de-
viated from this model, as not sensible and lacking agency.
Method
Participants
To examine whether direct contact with survivors affected ob-
servers’ perceptions of them, we selected two samples of ob-
servers. First, we recruited 144 relief workers who had direct
contact with survivors and spent an average of 3.5 weeks in the
hurricane-threatened area. To obtain a diverse sample, we re-
cruited participants through Red Cross Listservs, as well as
through forums and advertisements on Web sites for Katrina
relief workers. This sample included employees and volunteers
from nonprofit and governmental organizations (e.g., Salvation
Army), doctors, counselors, firefighters, and police officers.
Second, we recruited lay observers (161 adults and 156 stu-
dents)whohadnodirectcontactwithsurvivorsandobservedthe
consequences of the disaster from afar. We recruited adult lay
observersthroughon-lineadsandstudentlayobserversthrough
dorm Listservs. Because the adult and student samples did not
differ in any analyses, we present results of analyses in which
1Models of agency are not fixed properties of people. They derive from the
context and change upon exposure to contexts where other models are prevalent
(Heine, Lehman, Markus, & Kitayama, 1999).
Volume 20—Number 7
879
N.M. Stephens et al.

Page 3

these two groups were combined. Table 1 presents demographic
information about the Study 1 participants.
Procedure
We used responses on two within-subjects tasks as our primary
dependent variables. After completing these tasks, participants
also reported demographic information and answered questions
about their political orientation and religious views.
First,aperson-descriptiontaskassessedobservers’perceptionsof
survivors.Weaskedparticipantstoprovidethreewordsdescribing
people who evacuated from the hurricane-affected area and three
wordsdescribingpeoplewhostayedinthehurricane-affectedarea.
We counterbalanced the order of these two questions.
Second, a vignette task assessed whether participants per-
ceived survivors’ actions as sensible. Each participant read two
vignettes, one about a leaver and one about a stayer (in coun-
terbalanced order). Personal characteristics and family struc-
ture were consistent across the vignettes (i.e., a ‘‘friendly,
responsible, andhardworking’’survivor‘‘lives with two kids and
a spouse’’). In the leaver vignette, survivor ‘‘K’’ had resources
and evacuated (i.e., went to another state to ‘‘stay with a friend
untilthehurricanepassed’’).Inthestayervignette,survivor‘‘D’’
lacked resources and stayed (i.e., ‘‘didn’t have any close friends
or family to stay with who lived outside of the hurricane-
threatened area’’). Participants were asked, ‘‘Given the situa-
tion,towhatextentdidthesurvivor’sbehaviormakesense?’’The
response scale ranged from 1 (not at all) to 4 (a lot).
Results
Person-Description Task
Two research assistants who were blind to our hypotheses coded
each word for valence (positive or negative) and thematic con-
tent (mean k 5 .93). To capture the thematic content of the
person descriptions, we created a code for each distinct theme
that occurred in at least 5% of responses.
Although observers described stayers as lacking resources
and leavers as having resources (see Table 2), they still viewed
leavers as influencing agents and stayers as lacking agency. As
hypothesized, descriptions of leavers were more likely than
descriptions of stayers to include positive attributes that reflect
the cultural ideals of a disjoint model of agency. For example,
leavers were described as independent (e.g., self-reliant, in
control) and responsible (e.g., hardworking, conscientious).
Descriptions of leavers were also more likely than descriptions
ofstayerstorefer tohigh-arousalemotions (e.g.,angry,agitated)
and action-requiring states (e.g., being prepared, planning). By
contrast, observers were more likely to describe stayers as
having negative attributes that together connote inaction and
lack of agency. For example, stayers were described as passive
(e.g., lazy, dependent), irresponsible (e.g., careless, negligent),
andinflexible(e.g., stubborn,uncompromising).Descriptionsof
stayerswerealsomorelikelythandescriptionsofleaverstorefer
to low-arousal emotions (e.g., sad, depressed) and inactive
states, such as being unprepared (e.g., disorganized, ill-equip-
ped) and defeated (e.g., hopeless, devastated).
To assess whether participants’ perceptions were moderated
bytheirexperienceaseitherareliefworkeroralayobserver,we
used the valence coding to create a positivity index. For each
response (comprising three descriptors), we subtracted the
number of items coded as negative from the number coded as
positive. As expected, observers described leavers more posi-
tively(M50.72)thanstayers(M5?1.75),t(460)519.83,p<
.001. A repeated measures analysis of variance with between-
subjects factors revealed a significant interaction between
sample (relief worker vs. lay observer) and survivor group
TABLE 1
Demographic Characteristics of Observers in Study 1
Variable
Relief workers
(n 5 144)
Lay observers
Adults (n 5 161)Students (n 5 156)
Gender52% female,
48% male
39.9
81% female,
19% male
36.0
73% female,
27% male
20.8 Age (mean)
Race-ethnicity
White
Latino
Asian
Black
Other
Middle classa
Annual household income
(median; thousands of dollars)b
Number of children (mean)
86%
4%
2%
1%
7%
99%
63%
7%
12%
5%
12%
98%
44%
6%
19%
10%
21%
85%
50–75
2.5
50–75
1.6
100–200
0.0
aMiddle class was defined as having some college education (for relief workers and adult lay observers) or as having
at least one parent with a college education (for students).bHousehold income was assessed on a categorical scale.
880
Volume 20—Number 7
Perspectives of Katrina Observers and Survivors

Page 4

(leaver vs. stayer), F(1, 459) 5 16.18, p < .001. Relative to lay
observers, relief workers viewed stayers less negatively,
t(306) 5 2.34, p 5 .02, and leavers less positively, t(239) 5
3.13, p 5 .002 (see Fig. 1). In follow-up analyses, we found that
participants’ liberalism, conservatism, and religiosity were not
significant covariates.
Vignette Task
For the vignette task, we compared observers’ ratings of the
extent to which each survivor’s actions made sense. A paired-
samplesttestrevealedthatbothreliefworkersandlayobservers
perceived leavers’ actions as making more sense (M 5 3.87)
than stayers’ actions (M 5 2.65), t(460) 5 25.91, p < .001.
Vignette order did not affect responses.
Discussion
Ashypothesized,bothsamplesofobserversportrayedleaversas
influencing agents and stayers as lacking agency and saw
stayers’actionsaslesssensiblethanleavers’actions.Wesuggest
that these perceptions arose because observers assumed that
influencing the environment through independence and control
was the only way to be agentic. Thus, stayers’ actions, which
deviated from the disjoint model, were viewed as not making
sense and as lacking agency. Notably, observers derogated
survivors who stayed as stupid and passive, despite clearly
recognizingreasonswhystayerscouldnotevacuate(e.g.,lackof
money or transportation).
TABLE 2
Percentage of Descriptions of Leavers and Stayers Coded for Each Category in Study 1
Coding categoryExamples Leavers (%)Stayers (%)
Positive attributes
Smart
Responsible
Independent
Negative attributes
Stupid
Irresponsible
Passive
Inflexible
Active states
High-arousal emotions
Prepared
Inactive states
Low-arousal emotions
Unprepared
Immobilized
Defeated
Luck
Good luck
Bad luck
Resources
Have resources
Lack resources
Demographics
White race
Non-White race
Poor health
Other demographics
Intelligent, sensible, wise
Responsible, cautious, conscientious
Independent, in control, self-reliant
28
13
9
0
0
2
Dumb, ignorant, foolish
Irresponsible, careless, negligent
Dependent, helpless, lazy
Uncompromising, stubborn, strong-headed
0
0
3
6
15
7
9
27
Angry, stressed, agitated
Prepared, planned, organized
26
21
14
0
Sad, depressed, grateful
Ill-equipped, uninformed, disorganized
Isolated, cramped, trapped
Defeated, devastated, hopeless
21
0
3
12
31
10
12
18
Lucky, fortunate, good luck
Unlucky, unfortunate, bad luck
18
2
0
7
Well-off, rich, privileged
Poor, broke, no transportation
26
11
0
52
White, Caucasian, European American
Black, African American, minority
Sick, dying, ill
Elderly, female, male
10
5
2
2
0
18
7
7
Note. McNemar’s tests of homogeneity between percentages of leavers’ and stayers’ descriptions coded for a given cat-
egory were all significant at the p < .001 level, except in the case of ‘‘defeated,’’ p < .05, and ‘‘bad luck,’’ p < .01.
2
1
0
Disaster Relief Workers Lay Observers
0.200.96
−1.53
−1.84
Leavers
Stayers
Positivity Index
−1
−2
Fig. 1. Mean positivity of relief workers’ and lay observers’ perceptions
of Katrina survivors as a function of whether survivors stayed in New
Orleans or left. Error bars denote ?1 SE.
Volume 20—Number 7
881
N.M. Stephens et al.

Page 5

Relief workers, like lay observers, derogated stayers as
lacking agency. The convergent perspectives of these different
populationsunderscorethepowerandprevalenceofthedisjoint
model of agency in mainstream American contexts. In their
person descriptions, however, relief workers derogated stayers
somewhat less than lay observers did. This divergence could
have occurred because relief workers and lay observers had
different views (e.g., racial attitudes) prior to the hurricane. Our
theory, however, suggests that exposure to the contexts of sur-
vivors shaped relief workers’ models of agency and their per-
ceptions of survivors.
Observers’ derogation of stayers is consistent with the asser-
tion that psychological biases, such as prejudice and belief in a
just world, shaped responses to Katrina. Our purpose, however,
istotakeabroadersocioculturalperspectivebyilluminatingthe
disjoint model that underlies observers’ perceptions and con-
tributes to these more specific biases. Numerous psychological
theoriespredictthatobserverswouldlocatethecauseofstayers’
unfortunate outcomes in stayers’ own individual attributes and
thus view stayers negatively. Models-of-agency theory predicts
more specifically how stayers will be derogated—that they will
bejudgedonthebasisofwhethertheiractionsdeviatedfromthe
disjointmodel’sassumptionsaboutwhatconstitutesnormatively
good action. Given the pervasive negative representations of
African Americans (Oyserman & Harrison, 1998), observers
could have derogated stayers in myriad ways, such as by de-
scribing them as immoral, crazy, dangerous, or devious. How-
ever,guidedbytheassumptionthatnormativelygoodactionsare
those that influence the environment, observers in our study
derogatedstayersusingasocioculturallyspecificsubsetofterms
(e.g., lazy, passive, and careless) that together connote inaction
and the absence of agency.
STUDY 2
Moving beyond observers, Study 2 examined firsthand accounts
of Katrina survivors. Three months after Katrina, we asked
survivors to describe their hurricane-related experiences. We
hypothesized that leavers’ and stayers’ narratives would reflect
divergent models of agency.
Method
Participants
Seventy-nine participants were interviewed for 1 hr and com-
pensated $50. To obtain a diverse sample, we sent study invi-
tations to survivors on a Department of Housing and Urban
Development mailing list and posted flyers in New Orleans and
San Antonio, Texas. We also used on-line advertisements and
Listservstorecruit survivors whoevacuatedbeforeKatrina.Our
samples of stayers and leavers (see Table 3) were demographi-
cally comparable to the overall populations of Katrina survivors
who stayed and evacuated, respectively (Center for American
Progress, 2005). We conducted 57 interviews in person and 22
over the telephone. All interviews were audiotaped with per-
mission and transcribed.
In this study, we contrasted the perspectives of leavers and
stayers. We did so for two reasons. First, such a focus allowedus
tocomparetheresultsofStudy1(observers’perceptionsofleav-
ers and stayers) with survivors’ understandings of themselves.
Second, because most middle-class White participants were
leavers andmost working-classBlack participantswere stayers,
the experiences of leaving and staying provided a conceptually
meaningfulproxyfor socialclassandrace.Further,ourfocuson
leavers and stayers helped to specify some of the social expe-
riences that produce the divergent life outcomes tied to the
categoriesofsocialclassandrace(Helms,Jernigan,&Mascher,
2005; Markus, 2008). It is these differential social experiences
that shape and maintain models of agency.
Procedure
We matched the race of interviewers and survivors to help par-
ticipants feel comfortable sharing their experiences (Schaeffer,
1980). The interviewers asked participants to describe their
experiencesbysaying,‘‘Startfromthebeginning.I’dliketohear
what happened to you before, during, and after the hurricane.’’
Participants subsequently provided demographic information
and answered questions about their well-being and mental
health.
Coding
Three coders who were blind to our hypotheses read the tran-
scripts and identified explicit agency-related themes present
in survivors’ descriptions of what they did and why. Thirteen
non-mutually exclusive codes that each required minimal in-
ferenceandoccurredinatleast10%ofnarrativeswereincluded
in the coding scheme (see Table 4; mean k 5 .90).
TABLE 3
Demographic Characteristics of Leavers and Stayers in Study 2
Variable
Leavers
(n 5 38)
Stayers
(n 5 41)
Gender 71% female,
29% male
40.3
78% White,
22% Black
90%
$35,000
1.4
49%
40%
100%
73% female,
27% male
45.8
29% White,
71% Black
39%
$19,500
2.8
78%
18%
54%
Age (mean)
Race-ethnicity
Middle classa
Annual personal income (mean)
Number of children (mean)
Born in New Orleans
Has flood insurance
Owns a vehicle
aMiddle class was defined as having some college education.
882
Volume 20—Number 7
Perspectives of Katrina Observers and Survivors