Environmental Hazards 6 (2005) 123-133ENVIRONMENTALHAZARDSPapersDisaster threat: Preparedness and potential response of the lowestincome quartileBrenda D. Phillipsa, William C. ~ e t z ~ l * ,Leslie A. ~ i e v e s ~"Depur~~iletl~of Poli~iculScience. Oklulronlu SIuie Universiij~,5368 Mtiill Sciences Bliilling. Siill~~~tirer.O K 74074. USAb~ecisionuntl Inforn~tiiionSciences Division, Arqonne Ntilioncil Lohororor)~.9700 So~iilrCuss Auenlie. Arqonne, I L 60439. USAAbstractFor a community to manage hazards successfully, those who are responsible for planning and implementing responses to a disasterthreat situation must understand the social and economic realities of populations at risk. A random sample survey of residents in thevicinity of a US Army chemical weapons storage depot in Alabama confirnms that those in the lowest quartile of household income (i.e.,less than US $25,000in 1999)differ in important ways from the rest of the sample. Using econonlic status as a grouping variable resultedin identifying a concentration of individuals with special needs. This group differed significantly from the remainder of the sample as todemographic and attitudinal characteristics, hazard knowledge and concerns, emergency preparedness, and emergency decision-makingand their likelihood of taking protective actions. Respondents in the lowest income quartile reported greater restrictions in physicalabilities, fewer community contacts, a heightened concern about area hazards, and limited resources for taking preparedness andresponse actions.02006 Elsevier Ltd. All rights reserved.Ke)~~~~ords:Populations at risk; Economic status; Special-needspopulation; Emergency planning: Decision making; Protective action;Chemical weapons;Technological hazards1. IntroductionThis study examines the preparedness and potentialresponse behaviors reported by members of households ina northeastern Alabama, US community facing the threatof a possible chemical release from a US Army depot. Thestudy occurred in the context of state emergency responseIplanning for the unlikely event of a release during storageor incineration of a portion of the national chemicalweapons stockpile. In this study, economic strata of thesample were analyzed to identify any population-specific_ needs-anhbehauiors-witkrele~~ncefar-emegency-rgsponse-planning. Within the low-income group, additional demo-graphic variables emerged that suggest increased potentialrisk.This study illustrates the value of learning about acommunitys at-risk populations in advance of a potentialevent so that appropriate emergency preparedness activities*Corresponding author Fax: + 1 6302525327.E-tltuil tidciress: wmetzLn alil.gov (W.C. Metz).and response planning can take place. In the discussionthat follows, we situate our work in existing literature,describe the contextual setting in which the study occurred,outline our methods and findings, and offer recommenda-tions for research, policy and practice. Given that chemicalhazards in many communities have increased over time andthat evacuation compliance can reduce deaths (Cutter,1991), understanding potential behaviors of those presum-ably most vulnerable should prove insightful. Strategies forenhanced preparedness and targeted outreach campaignscan then be developed and implemented to facilitate and-rei-nforceappr-opr-iaie-pwtec-t-i-2. Literature reviewIn this examination, household preparedness serves asthe dependent variable. Existing literature finds that, ingeneral, households fail to prepare for most hazards andare especially unprepared for chemical hazards (Tierneyet al., 2001). Studies have yet to understand householdpreparedness among those presumed at highest risk, that is,1464-28671s-seefront matter Q 2006 Elsevier Ltd. kll rights reserved.doi:10.1016/j.hazards.2006.05.001
socially and economically vulnerable populations (Quar-antelli, 1984). For this study, we define vulnerability as"the characteristics of a person or group and their situationthat influence their capacity to anticipate, cope with, resistand recover from the impact of a natural hazard" (Wisneret al., 2004, p. l I).Key independent variables that indicate social vulner-ability include income, gender, race and ethnicity, age,geographic location, homeownership, education, healthstatus and special needs (Heinz Center, 2002; Enarsonet al., 2003; Tierney et al., 2001; Sattler et al., 2000;Drabek, 1996). Logically, ones personal context, particu-larly the lack of key resources such as health, education orincome increases vulnerability (Fothergill and Peek, 2004).Demographic variables serve as predictors as well, withones age (very young or very old), ones gender (beingfemale), or ones racelethnicity (i.e., not belonging to thedominant group) increasing vulnerability (Pine et al., 2002;Fothergill et al., 1999; Mulilis, 1999; Phillips, 1993; Perry,1985; Morrow and Phillips, 1999; Quarantelli, 1993).Situational variables also increase vulnerability; for exam-ple, living in a rural area or being a renter increases risk(Morrow, 1999). Further, having special needs such as adisability exacerbates vulnerability to a threat (VanWilligen et al., 2002; Dow and Cutter, 2002).Presumably, each additional disabling condition thataffects an individual increases hislher vulnerability. Thus, alow-income elderly woman in poor health would presum-ably be more vulnerable than a younger low-incomewoman in good health. Similarly, a poor, rural rentermay lack income or shelter options (Heinz Center, 2002).Separating economic effects from other variables, such asrace, disability, and age, remains a significant challenge(Fothergill and Peek, 2004; Tierney et al., 2001). Unfortu-nately, only minimal research has been focused on thelikely role played by combinations of these factors (Tierneyet al., 2001).Further compounding demographic, situational andcontextual variables, vulnerable populations also appearto receive, perceive and interpret risk differently bothwithin and across groups (Lindell and Perry, 2004; Perryand Mushkatel, 1986). For example, persons that are deafmay not even receive warnings to the general population(Wood and Weisman, 2003). Racial and ethnic minoritiesmay not find authorities transmitting risk messages to becredible (Perry and Mushkatel, 1986; Legates and Biddle,1999). Women may want to respond immediately to riskcommunications, whereas men may be more likely toeschew a warning (Enarson and Morrow, 1998).How communities choose to manage the increasedvulnerability of some population groups carries importantimplications for those facing chemical hazards (Mileti,1999). Effective disaster planning requires understandingboth the vulnerability issues and the potential behavioralresponses of those at risk (Tierney et al., 2001; Drabek,1996; Lindell and Perry, 1992; Auf der Heide, 1989;Drabek, 1986; Quarantelli, 1985; Cutter et al., 2003a;Cutter et al., 2003b; Dynes, 1994). Understanding andassessing the prevalence of vulnerabilities within a specificcontext is the purpose of this examination, an effort withimplications for future research and for the practice ofemergency management.3. ContextA community in northeastern Alabama provided anopportunity to identify those at-risk and to investigatelevels of household preparedness and anticipated reactionsto a potentially disastrous event (Metz et al., 2005, 2002).The Anniston Army Depot (ANAD) is one of seven USArmy depots located within the continental USA thatcurrently host stockpiled chemical weapons that are bothbeing incinerated and awaiting destruction. ANAD storesmore than 2000tons of chemical weapons agent (nerve andvesicant agents). ANAD is situated approximately 70 mileseast of Birmingham and 100 miles west of Atlanta,Georgia. Fig. 1 shows the area around the depot with thedesignated response zones and population centers. Ap-proximately 31,000 households, or 75,000 persons, arelocated within the Immediate Response Zone (IRZ), with aradius extending 13-16 km from the depot. The IRZincludes two counties, and the major population center isthe City of Anniston, with a 2004 population estimated at24,000. In the unlikely event of an accident that carriespotentially deadly chemicals off-site, residents would haveonly a short time to receive a warning and undertake aprotective action (i.e., evacuation or sheltering).In assuring community leaders of maximum protectionfor the general population from the consequences of anagent release off the depot, the US Army and the FederalEmergency Management Agency sought to mitigate theeffects of an accident to the extent practicable, defined asensuring individual risk of no greater than "one fatality in2,500,000 years" (US Department of the Army and FederalEmergency Management Agency, 2000). This study andFig. I . Anniston Army Depot and Surrounding Area.
B.D. Pl~illipsei ul. / Enuironmeniul Huzurcls 6 (2005) 123-133 125special emergency preparation efforts that followed were The survey collected the usual socio-demographic dataundertaken in an effort to afford persons with special needs along with attitudinal and other data relevant to emergencya level of protection equivalent to that afforded the general planning. The approximately 100 questions, 25 of whichI population. had multiple parts, required an average of 1 h and 20minto complete. Quality control was maintained by a 20%I supervisory sampling of on-going phone surveys, and a4. Methodologysupervisor callback for each realized in-person interviewensured survey consistency, as did data entry using aArgonne National Laboratory, working with the Uni-versity of Alabama at Birmingham (uAB), conducted a10% random sample survey of the estimated 31,000households in the IRZ from May through November1999. A major goal was to provide baseline data to theAlabama Emergency Management Agency, the sponsor,on persons with special needs. The Chemical StockpileEmergency Preparedness Program (CSEPP) PlanningGuidance (see Federal Emergency Management Agencyand US Department of the Army, 1996) defined personswith special needs as persons with sensory, mobility, ormental impairment and/or individuals with special equip-ment needs because of medical conditions; chronically illpersons; individuals who do not own or have access to anautomobile; and unattended children who would needassistance in the event of an accident.With a goal of surveying 3100 IRZ households, a sampleof approximately 4300 addresses was extracted from acommercial database so that after deleting businesses,nonfunctioning numbers, addresses lacking phone num-bers, and some cases of dubious eligibility, a sample set of3159 households remained. During the survey process,2776 persons were interviewed. In most cases, therespondents identified themselves as either the head ofhousehold or the special-needs persons caregiver. Some8O0/0 of the respondents were interviewed by telephone; theremainder, who could not be reached by telephone despitemultiple attempts, were interviewed in person. Afterdeleting incomplete surveys and responses of personsdetermined to be residing outside of the IRZ boundaries,2460 surveys remained-a response rate of almost 80%.While there were a number of no-contact cases due to listedaddresses no longer being occupied by the listed person, aclear majority of the households that refused interviewslived in locations marked by poverty conditions. Inaddition, some interviews were refused in cases where ayounger family member or relative was acting as guardianof an older or dependent person. The socio-demographicvariables of gender, ethnicity, and median householdincome were roughly similar among the respondinghouseholds and the general population of the two countiescovered, as compared to the US Census 1998 GeneralProfile. Researchers were unable to profile the socio-demographic variables of the non-responders, nor theirintended behaviors.UAB took primary responsibility Tor survey design and construction.while Argonne National Laboratory assumed responsibility for datacollection and analysis.computer-assisted telephoning system that allowed statis-tical analyses of interviewer exchanges. Seventeen surveycases were listed as "missing", that is, where the sourcecould not be identified, producing a less than 1.0%processing error.To identify key variables and possible relationships, thedata were explored using frequencies, cross-tabulations,and correlation analysis. This data exploration identified asegment of the sample, distinguished by household incomelevel, which appears to differ substantially from theremainder of the sample in terms of emergency-relatedknowledge, preparedness, and response characteristics.These quantitative data rank among the few demographicprofiles available for at-risk populations. Given the resultsof the data exploration effort, the analysis that followsfocuses on the households in the lowest quartile of theincome distribution, defined as those with householdincomes lower than $25,000 in 1999. Descriptive statisticsfor household income for the 2178 respondents whoreported income are given in Table I . The mean incomefor the entire sample population was almost $39,000.Table 2 compares the demographic profile of the lowestincome quartile to the sample as a whole. The lowestquartile includes 43% of the sample households withspecial-needs individuals who reported that they requireassistance in an emergency. Within this group there was ahigh proportion of persons over age 64 who self-rated theirhealth as poor or fair; these are designated the "frailelderly" in the analysis. Together, households with frailelderly or with persons indicating a need for evacuationassistance and households with persons in both of thesecategories account for about 34% of households in thelowest income quartile compared to 16% of the entiresample.We, conducted statistical analyses using t-tests fordifferences in quartile means and z2-tests for differencesTable IDescriptive statistics Tor 1999 household income Tor the sample (11 = 2,178cases for which income was reported)Statistic Va lue ($1000~)MetinStandard deviationMinimumLowest quartileMedianUpper quartileMaxiinuln
Table 2Composition of the lowest quartile of household income compared to thetotal sample, among those reporting incomeCategory" Lowest income quartile Total sampleN %;I N %hFrail elderly 80 15 154 7Frail elderly/special- 58 I I 79 4needs householdSpecial-needs 42 8 112 5householdOther 359 67 1833 85Total 539 101 2178 101"The following definitions were used: (1) Frail elderly = age over 64 andself-rating of health as poor or fair; (2) frail elderly/special-needshousehold = household with someone in categories I and 3; (3) special-needs household = someone in household is unwilling or unable to leaveor evacuate; (4) other = not in categories I through 3.h~ercentagesdo not sum to 100% due to rounding.in the distribution of responses among data categories.Since these statistical tests tend to become increasinglysignificant as sample size increases and the number ofrespondents in this study is large, most test statistics arehighly significant. In this situation, it is the absolutemagnitude of (significant) differences between groupresponses that may be most notable. Percentages given inthe tables that follow represent the percentage of respon-dents for whom household income was reported.5. Research findingsThe research findings focus on demographic andattitudinal characteristics, hazard knowledge and concerns,emergency preparedness, and emergency decision-makingand likelihood of taking protective actions in an emer-gency. Analyses consider the total sample and twosubgroups based on economic status-those with lowestquartile of household income and the remainder (upperthree quartiles) of the total sample.5.1. Demographics and attitudinal characteristicsRelevant demographic characteristics are compared in- ~ - ~ - T a b l - ~ 3 - a s i ~ g 7 ; f e s t s - f o r r h m m - f - c h - ~group.Race and Ethnicity: The lowest income quartile has asignificantly higher proportion (33%) of Black respondentscompared to the remainder of the sample (18%).Gender:Females comprise a significantly greater propor-tion of the lowest quartile (73%) than of the rest of thesample, which is 56% female.Education: The lowest income quartile had significantlyless education: a mean of 1 1.6 years (roughly equivalent tohigh school graduation) compared with 13.6 years for theremainder.Table 3Comparison of demographic characteristics for the lowest income quartilewith the remainder of the sampleResponse !-value 01) Lowest income Remainderquartile (n = 539) (n = 1639)Reported 7.16 (0.00) 32.8% 18.3%race = BlackGender = Female 7.07 (0.00) 73.4% 56.5%Mean years of -13.71 (0.00) 11.6 13.6educationCurrently married -16.80 (0.00) 27.6% 66.5%Single-person 12.10 (0.00) 40.9% 16.5%householdChildren under age -4.12 (0.00) 28.0% 37.7%16Table 4Comparison of health status and activity measures for the lowest incomequartile with the remainder of the sampleQuestion and responses Lowest income Remainderquartile (%)HWIIIIS~~IIIISX 2 = 261 0,= 0.00)Excellent 10.1 21.2Very good 13.8 33.3Good 27.1 28.0Fair 25.9 12.2Poor 23.1 5.3Self-reporlet1 c o n o ~ ~ ~ m i ~ ) ~intiolve~~~enty2 = 36 @ = 0.00)Very involved 19.4 28.9Involved 32.5 36.1Not very involved 31.0 24.2Not involved , 17.2 10.8Feel ltelpless to rleal 1vir11proble~~uof life X2 = 62 @ = 0.00)Strongly agree 4.4 3.9Agree 35.2 20.0So-so 5.3 3.6Disagree 46.1 56.2Strongly disagree 9.1 16.4Marital status: Persons in the lowest income quartile aresignificantly less likely to be currently married (27.6%)than the remainder of the sample (66.5%). The presence ofmarital partners may contribute to higher income status fora household because of the effect of having two wageearners. CCREiSenti-th-lower rates of current marriage,-the lowest income quartile has a substantially higherpercentage of single-person households (40.9%) comparedto the rest of the sample (16.5%).Children: Households in the lowest income quartile aresignificantly less likely to have children under 16 years ofage in the home (28.0%) compared with higher incomehouseholds (37.7%). Investigation of the composition ofthe lowest income quartile households with childrenindicated that households where the parent was under theage of 21 accounted for only a small percentage of thesehouseholds.
B.D. Pllillips el crl. / EnvironmentReported health status and activity measures arecompared across the lowest income quartile and theremainder in Table 4.Health: The survey asked respondents to self-selectamong five categories of health status from excellent topoor. Differences between the two groups were significant(based on a Pearson X2-test),with 49% of the lowestincome quartile reporting."poor" or "fair" health com-pared to 18% of the rest of the sample.Community involvement: Differences in communityinvolvement are also significant, with the lowest incomequartile reporting less involvement. Community involve-ment was defined as peoples informal interactions within acommunity and was operationalized through such self-reported items as attending local events, participating inlocal elections, and going to places of worship.Coping skills: The lowest income quartile and theremainder also contrast strongly in regard to their self-reported competence to cope with lifes problems. Personsin the lowest income quartile are significantly more likelythan the rest of the respondents to report feeling helpless,though a majority of people in both groups report notfeeling helpless.Demographic results confirm the higher concentration ofspecific at-risk populations within the lower incomequartile, signaling an alert to emergency managementagencies. If coupled with reduced levels of preparednessand/or abilities to respond, the situation would thusdemand that emergency management agencies prioritizeat-risk populations for outreach and capacity building.5.2. Haznrd k~toivledgeand cortcerrtsThe mean self-reported residential distances from thechemical storage site are similar for the two groupsevaluated: 8.7 and 8.8 miles (14 and 14.2km), respectively.2If other aspects of their lives were equal, the two groupsmight be expected to have similar levels of concern relatedto a potential major chemical accident. However, the firstquestion in Table 5 shows that persons in the lowestincome quartile are significantly more likely to express highlevels of concern about a chemical weapons accident.Nearly three-quarters of the lowest income quartile reportbeing "concerned" or "very concerned" about a majoraccident. Thisfinding may be associated with the tendencytoward feelings of helplessness shown in Table 4, reflectingthe lowest quartiles greater proportion of persons who areolder, live under more difficult economic conditions, orhave more health impairments.The second question in Table 5 pertains to knowledge ofemergency planning and warning systems, which are partof the CSEPP. Though more concerned about an accident,respondents in the lowest income quartile are significantlymore likely to report being "little informed" or "not at all~ifferenceswere nonsignificant in a I-test of the mean distances for thetwo groups.Table 5Hazard knowledge and concerns: a comparison of the lowest incomequartile with the remainder of the sample-Question and responses Lowest income Remainderquartile (%) (%)Ho~vconcerned ubuut u 1?1ujordrer~~iculuccichnt? ,y2 = 34 @ = 0.00)Very concerned 48.2 39.5Concerned 24.4 19.1Somewhat concerned 10.7 18.1Little concerned 9.5 14.0Not at all concerned 7.2 9.3Holv nvll infornlecl clo yolrfiel crbolrr CSEPP? ,y2 = 38 @ = 0.00)Well informed 3.9 10.6Informed 10.1 13.8Somewhat informed 18.2 20.4Little informed 27.7 24.6Not at all informed 40.1 30.5informed" regarding CSEPP. This may be a result of fewersocial and community contacts (see Table 4), which meansfewer opportunities for receiving information about theenvironmental hazard (Lindell and Perry, 1992). It isnotable, however, that a majority of the remainder of thesample also feels poorly informed (55% in the "little" or"not at all" categories), possibly because, at the time of thesurvey, chemical weapons incineration was not especiallynewsworthy (Nisbett and Ross, 1980). Consistent with thelowest income quartiles perception of being uninformed, asignificantly lower percentage of the lowest incomequartile, 65% versus 71% of the remainder (t = 2.29,p = 0.02), correctly identified the weekday on which theemergency sirens are tested.Responses to selected questions related to emergencypreparedness issues are summarized in Table 6. The firsttwo questions pertain to access to emergency preparednessinformation. In regard to participation in preparednessclasses, about 61% of the lowest income quartile and 68%of the remainder express a willingness to attend classes.Conversely, a slightly higher percentage of the lowestincome quartile (21% versus 15%) indicates that they- w o u l d d e m T i t e l y ~ 1 p ~ o - b a - b l p a r ~ eclasses. This response is consistent with the greatertendency toward feelings of helplessness reported byrespondents in the lowest income quartile (Table 4). Itmay be that the extent to which individuals believe thatthey control their own lives or what happens to them isdirectly related to their perceived ability to undertakeprotective actions (Sims and Bauman, 1972).To provide information regarding recommended emer-gency actions, county emergency management agencieshave mailed annual calendars, brochures, evacuation routemaps, and other explanatory materials to each household;information is also printed in area telephone books and has
128 B.D. Plrillips el 01. / EnuironmenrTable 6Emergency preparedness: a comparison of the lowest income quartile withthe remainder of the sampleTable 7intent to take protective actions: a comparison of the lowest incomequartile with the remainder of the sample"Question and responses Lowest income Remainderquartile (%) (X)Question and Responses Lowest lncorne RemainderQuartile (%)Woulcl you purricipctre in prepureclness clusses? ,y2= 20 @ = 0.00)Definitely yes 29.5 32.6Probably yes 31.0 35.0Maybe 18.8 17.3Probably not 11.0 10.3Definitely not 9.7 4.8Receiuecl u CSEPP cctlendur? Yes. r = -3.68 @ = 0.00)33.5 42.5Able ro lrerir ~t~rtrrrinysiren from insicle lrorire? ,y2= 6 @ = 0.05)Yes, all the time 80.7 76.0Sometimes 14.5 16.7No, never 4.9 7.3Huue un en~eryencyplurr for futi~ily?Yes. I = -5.27 @ = 0.00)11.9 22.3Huue an enlergency prepctreclness kil? Yes. r = -4.70 @ = 0.00)7.4 15.3Tinres n~irlrno cur uuuiluble? ,y2= 27 (p = 0.00)Always 2.9 1.7Frequently 5.1 2.6Sometimes 9.5 4.6Rarely 5.8 6.3Never 76.8 84.7Neiylrbor ~vou/r/lrelp Vcur is rmauctilcible? Yes. I = -3.44 @ =0.00)37.0 53.9Tin~es1v11encl~ilrlren11omectlone? z2= 3.6 @ = 0.30)Frequently 7.4 5.9Sometimes 12.8 13.5Rarely 8.7 14.2Never 71.1 66.4Huue urrunyenrenrs i/ ctccirlet~coccitrs 11dri1eclrilclren ure ulone? Yes.1 = -0.85 @ = 0.40)30.2 37.1I/ I lectrn of un ucciclen!, I ~vo~rlcl... X2 = 9.8 @ = 0.00)Act at once 50.7 58.5Seek more information 49.3 41.5To obruin riiore inforniulion. I 11~oir1cl...Turn on local TV, r = 2.98 89.8@ = 0.00)See what neighbors are doing, 47.7r = 3.26 @ = 0.00)Call relative or friends, r = 2.64 59.5@ = 0.00)I/ucciclen! 11~11ileclrilclren U I sclrool, I ~vortlcl... ,y2= 7.7 @ = 0.02)Try to pick them up myself 76.8 70.0Leave them in care of school 23.2 29.8In rlre even! of un uccirlmr, I n~iNlrelp ny neiylrbors. ,y2= I I (p = 0.02)Strongly agree 26.6 32.6Agree 68.6 61.7Disagree 3.1 3.3Strongly disagree 0.2 1.1I/urtrl~oririesucluisecl euctnturion, 11~ortlt1yolt.. . ,y2= 6 @ = 0.18)Definitely leave 85.1 85.3Probably leave 12.1 13.0Probably not leave 2.1 0.8Definitely not leave 0.4 0.5In cuse of euucuurion, ~souMyort ... ,y2 = 72 @ = 0.00)Stay in a motel/hotel 16.3 35.9Stay with friendstrelatives 59.3 50.1Stay in a public shelter 24.4 14.1I/ctrrrlrori!ies ud;iserl Itsing u coti~nr~miryslreller, 11~ort1clyou (lo? z2= 26@ = 0.00)Definitely go 60.0 48.4Probably go 29.9 34.0Probably not go 5.7 9.6Definitely not go 4.1 7.2been presented in area newspapers. In spite of this effort todisseminate information, less than half of either groupreported receiving the calendar, as shown in the secondquestion in Table 6. Comparing the.twogroups, the lowestincome quartile is significantly less likely to report havingreceived a calendar. Responses to questions regardingreceipt oftheotheraterlals are simllFiF€F€he rates o rreceipt reported for the calendars. Thus, since people in thelowest income quartile are less willing or less able to attendclasses that explain preparedness action and report lessrecall of mailed materials, it may take concerted, focusedand/or innovative efforts to reach them with emergencypreparedness information.Hearing warnings: With regard to their ability to hear theemergency warning sirens, the lowest income quartile isslightly more likely to hear the siren from inside their homeand the differences are statistically significant. This couldbe due to the concentration of those in the lowest income"For some questions, responses of small numbers of undecidedrespondents have been omitted from the table, though not from thestatistical analysis. IIquartile in more urbanized areas with the densest coverageof outdoor warning sirens. The effectiveness of the sirens isevidenced by the fact that over three-quarters of bothgroups report hearing ihem.Rowever, once th-recelvethe warning, the patterns of warning confirmation differbetween the two groups, as discussed below in relation toTable 7.Readiness: Two questions addressed actions alreadytaken to prepare for protective action. The remainder ofthe sample was twice as likely as the lowest income quartileto answer that they have a plan for family protection incase of a major accident or that they have prepared a kit ofnecessities for such an event. However, only a smallminority of the sample has made these preparations; mostpeople in both groups had neither a plan nor a kit of
B.D. Phillips el 01. / Environmentnecessities. This situation exists despite relatively highlevels of concern about an accident at the chemicalweapons site. It is an alarming situation consideringresearch that shows that if one wants to evacuate, theabsence of a plan for doing so is sufficient to hinder anyadaptive response (Perry, 1985). Clearly, an effort to assisthouseholds with such preparedness seems advisable.Transportation: With regard to the availability oftransportation to evacuate the area or to reach acommunity shelter, the lowest income quartile wassignificantly less likely to have a car readily available.Researchers elsewhere have found that those who do notown vehicles are most likely to obtain assistance inevacuating from the same people who routinely help themon a daily basis (Tierney et al., 2001). In the survey, a seriesof questions attempted to identify sources of help insituations when a car is unavailable. The two groupsshowed no difference in their expectations of aid fromrelatives (50%) and friends (30%); however, previousdisaster studies have shown that evacuation in carsbelonging to relatives and friends occurs infrequently(Lindell and Perry, 1992), a problem noted with poorfemale-headed families in particular (Enarson and Mor-row, 1997). As to expectations of aid from neighbors, thelowest income quartile had significantly lower (37%)expectations of such aid than the remainder of the sample(54%). Evacuation planners should not automaticallyassume that people without vehicles will receive informalassistance (Tierney et al., 2001).Children in sey-care: When respondents in householdswith children were asked about the possibility that childrenmight be left alone, about 20% of households in eachgroup indicated that children are frequently or sometimesleft alone (no significant difference). Responding to aquestion regarding emergency planning for children, themajority of households in each group lacked plans forcoping with an accident if it were to occur while childrenare alone. The percentage of those having such plans wasslightly lower for the lowest income quartile but thedifference was nonsignificant.,5.4. Emergency decision making and likeliltood of takingprotective actions.- Tabld-summarizesresponsesby-the-lowestincome--quartile and the remainder of the sample regardingemergency decision-making and their likelihood of takingvarious protective actions. Though these responses indicateintended actions, the assumption that intended actions willactually occur or even become possibilities for under-resourced households is problematic. However, usefulinformation for planning purposes can be gleaned fromthese questions. For example, more than 70% of allrespondents indicated that they would attempt to pick uptheir children at schools. Should the majority of parentsattempt to retrieve their children, such behaviors couldseriously gridlock school bus evacuation routes or preventsuccessfulsealing of school buildings against contaminatedair. Given the intention of parents to try to pick up theirchildren, either adapting school planning or conductingpublic education campaigns would be advisable. In termsof statistical tests, the greatest differences between thegroups were found in their information-seeking actions andpotential destinations in the event of an evacuation.Warning response and information-seeking: The surveyinquired about how respondents would likely seek infor-mation before taking protective action in the event of anaccident. As shown in Table 7, the lowest income quartilehouseholds suggest they would be significantly more likelythan the remainder of the sample to seek more informationbefore taking action, a finding consistent with otherresearch on evacuation decision-making (Mileti, 1999).Proximity to the perceived threat may also positivelyincrease the number of sources used as well as the form ofcommunication preferred (Diggory, 1956).Respondents who indicated that they would seek moreinformation were then asked what information sourcesthey expect to use. The second question in Table 7addresses alternative ways of confirming the warningmessage, such as gathering further information, andestablishing a warning belief (Perry et al., 1981). Oneresearch initiative found that more than 80% of warningrecipients tried to confirm the first warning with at leastone additional information source (Perry and Greene,1983).In this survey, the lowest income quartile and theremainder were about equally likely to indicate that theywould (1) turn on radio news (nearly 70%) and (2) callpolice or 9-1-1 (nearly 50%). However, the lowest incomequartile appears significantly more likely to turn on localTV channels or to check with neighbors, relatives, andfriends before taking action, a finding that has previouslybeen associated with gender and racelethnicity variables(Lindell and Perry, 2004; Legates and Biddle, 1999).Although most people in both groups (more than 80%)indicate that they would turn to radio or TV news forinformation and instructions, there are also strong indica-tions that both groups have a propensity to call others,thus delaying evacuation compliance, not an unexpectedfinding (Tierney et al., 2001). As noted before, differencesbetween stated behavioral intent and likely behaviors could- b e a d d r e s ~ r r t i a l l y ~ e d d t ~ h m - e d u c a -tional outreach effort targeting these at-risk populations. Adesired outcome of such a program would be to reducetime spent in confirmation behavior and expedite takingprotective action.Cltildren: Respondents with children in the home wereasked about their likely actions if an accident were to occurwhile the children were in school. One countys annualcalendar provides residents with the following instructions:"Do not attempt to pick up children from school. Youmight cause traffic congestion. Schools will be evacuatedby zone, if necessary". Less than 30% of each groupindicated that they would leave their children in the care of
the school. Most reported that they would try to pick upthe children themselves or try to get someone else to pickthem up; while the difference in the percentages preferringthis action is relatively small, it is significantly higher forthe lowest income quartile. Research studies have indicatedthat under actual circumstances, separated families becomeanxious and attempt to reunite before evacuating (Perry,1985).Helping behavior: A series of questions concerned therespondents commitment to helping various categories ofothers in the case of an accident. Most people indicated awillingness to help neighbors (fourth question in Table 7), afinding consistent with numerous studies reporting highlevels of pro-social behavior (Tierney et al., 2001).However, the rate of commitment for the lowest incomequartile is significantly lower. This may reflect a realisticview of their capabilities given their typical characteristicsof advanced age, poor health, and lack of transportation.Anticipated evacuation rates: Most members of bothincome groups indicated that they would follow autho-rities advice to evacuate the area, and statistical analysisshowed no significant differences between the groups.Annual calendars distributed to people living near thedepot advised residents that they should "not waste timegathering personal items such as photo albums" beforeevacuating. This advice on the part of authorities isintended to encourage people to move swiftly. For bothgroups, the self-estimated mean time needed to prepare toleave their homes after hearing of an accident is aboutI0 min. Considering both groups together, 25% estimatedthat they could leave in 3min, over 50% in 5min, and 75%in IOmin. These estimates represent a more rapid potentialmobilization than has been found in other situations-onethat has not been tested in a local public drill. In a study ofevacuations caused by hazardous materials releases, 6O0/0of one communitys residents evacuated within IOmin,while 85% of another community evacuated in less than30 min after receiving the warning (Sorensen, 1991).Another researcher estimated that the distribution ofevacuation times for a population would be 15% atIOmin, 50% at 20min, 83% at 30min, and 100% at40min (Lindell and Perry, 1992). These estimates vary byhazard and context. Communities with prior disasterexperience, for example, evacuate at faster rates (Baker,1991; Phillips, 1992).--Intenderl-evamatiurdestinatio~Significa~~en~es-between the income groups were found regarding potentialdestinations in an evacuation. Responses of the remainderof the sample indicated that these persons are more thantwice as likely to stay in a hotel or motel (36% compared to16% for the lowest income quartile) and relatively unlikelyto stay in a public shelter (14%). In contrast, persons in thelowest income quartile reported being substantially morelikely to stay with friends or relatives (59%) or in a publicshelter (24%). CSEPP Planning Guidance (US Departmentof the Army and Federal Emergency Management Agency,1994) estimates that approximately 15% of an evacuatingpopulation is likely to use a public shelter, a finding slightlyin excess of comparable contexts (e.g., see Phillips, 1992).Higher rates of use would be expected if evacuees are olderor poorer than the general population, a finding consistentwith other shelter use studies (e.g., see Phillips, 1998;Yelvington, 1997). This estimate is substantiated by thesurvey data.The greater propensity of the lowest income quartile touse public shelter facilities is also shown in the last questionin Table 7. If using a community shelter is the recom-mended protective action, persons in the lowest incomequartile indicate that they are significantly more likely tocomply. Within the lowest income quartile, differences inthe proportions of the frail elderly and of households withindividuals with special needs who intend to use publicshelters are not statistically significant. However, otherstudies find that age, when influenced by income, genderand race, will increase the propensity of these populationsto use shelter (Tierney et al., 2001).6. Summary and conclusionsDocumentation of vulnerability within specific contextsprior to a disaster is rare in the scientific literature. Below,we review the findings vis-l-vis the existing literature andmake suggestions for future disaster research and thepractice of emergency management.6.1. Implications for researchThe lowest income quartile reported substantially moreconcern about a major chemical accident, a findingcomparable to studies on other low-income groups as well(Fothergill and Peek, 2004; Lindell and Perry, 2004).Despite their elevated concern, the lowest income groupreported fewer preparedness behaviors, which is alsoconsistent with the literature (Tierney et al., 2001). Lowerincome groups also reported they would be likely to solicitadditional information before evacuating, a behavior thatmay be influenced by multiple variables such as the greaterconcentration of women, elderly, and minorities in thisquartile (Heinz Center, 2002; Mileti, 1999; Russell et al.,1995).Lower income households in this study lack transporta-tion resources, a dilemma observed during evacuation for~ w r ~ c a n c K ~ t h e - U ~ o w e v e r ~ - c ~ t - th r -researchers need to study transportation decision makingfurther as "the rationale for the household decisions is notwell-documented" (Dow and Cutter, 2002, p. 14). Com-pounding the evacuation transportation challenge, lowerincome households in this study did not expect as much aidfrom family and friends although they did expect to staywith them after evacuating. As expected, lower incomehouseholds reported they would be more likely than higherincome households to use public shelters, also consistentwith prior research (Dawson, 1993; Yelvington, 1997;Phillips, 1993). Thus, a relatively high percentage of the
B.D. Phillips el (11. / Environmenta1 Hazards 6 (2005) 123-133 131vulnerable population might relocate to public shelters inan evacuation for a sustained period of residence (Morrow,1999). The frightening reality of a deadly chemical agentrelease (however, unlikely) could prompt residents of areassubject to evacuation to travel further than anticipated iftransportation resources become available to them.A number of research questions remain. To start,behavioral intent may not necessarily coincide with actualbehavioral response during a chemical accident or anyother disaster. Researchers could employ comparablescenarios, such as hazardous materials transportationaccidents, to assess behavioral response among vulnerablepopulations. Respondents in the lowest income quartilewere less likely to want to attend classes on creating a homeshelter environment and to have a family plan orpreparedness kit; additional research is needed to learnwhy lower income groups did not want to do so.Identifying strategies that motivate and support prepared-ness efforts among those most vulnerable is advisable. Thelow-income households in this study expect to engage inconfirmatory behavior if a protective action or evacuationorder is issued. Warning messages that personalize the risktend to increase compliance rates (Lindell and Perry, 2004).Consequently, additional research should investigate deli-vering tailored warning messages designed to reach themost vulnerable populations. Finally, more systematicefforts to understand the cumulative and interactive effectsof multiple variables would be beneficial, as the incomevariable alone remains insufficient to understand behavior-al intent and actual response.6.2. Emergency management practiceIn the community described here, emergency managersprovided hazard information to residents through maildelivery of annual calendars that described emergencyplanning, emergency alert and notification procedures,family emergency planning procedures, county emergencyplanning efforts, protective action options, and CSEPP. Inaddition, presentations were made at community eventsand to special audiences; other public information itemswere also distributed. Local newspapers often featuredstories and editorials about ANCA (Bragg and Wilson,2002; Landers, 2002). Despite these extensive efforts, locallow-income-housetrolds-still-exper:ted~-gage-irrc~)n~firmatory behavior before taking action. The implicationsfor emergency management are clear: preparedness, out-reach and educational efforts must target those at highestrisk repeatedly and thoroughly. Research literature sug-gests that such efforts should be done through credibleauthorities, social networks, and in ways that engage thoseat risk (NHRAIC, 2002).Researchers have advocated that empowering those at-risk to become co-involved in emergency preparedness andresponse planning can make a difference. Practitionersmight employ "participatory processes to help buildinclusiveness of gender, age, and race-based groups, andthose with disabilities into consensus-building publicmeetings, discussions, and workshops related to mitigationand disaster response" (Heinz Center, 2002, p. 110;see alsoNHRAIC, 2002). Including marginalized groups maypositively influence the perceived efficacy of those at risk,thus motivating higher compliance rates (Feldman, 1990).At the very least, involving those at risk with localauthoritiesshould serve to increase trust and credibility,both of which are necessary to decrease confirmation-seeking behaviors and motivate compliance.However, getting persons with low incomes to partici-pate in planning may be problematic considering thatrespondents indicated a general lack of transportationresources. Emergency management agencies might addressthis problem by developing partnerships with agenciespossessing transportation resources or by promoting "helpnetworks" of nearby neighbors, friends, and relatives whowould provide transportation, thus increasing protectiveactions for the lowest quartile.Finally, pre-planning special-needs shelters to accom-modate persons with disabilities and the frail elderly isessential. Given the presumably cumulative effects ofincome levels plus rural location plus impaired health/disability levels, the percentage seeking public shelteringunder the circumstances reported here is expected to behigh. Special-needs shelters might need to be located atscattered locations with varying distances from theevacuation zone to accommodate demand due to the lackof transportation resources and to reduce trauma amongthe elderly by remaining in proximity to their community.This study of the lowest income quartile providesinsights to guide emergency preparedness and responseplanning for a very vulnerable segment of the community.Public confidence in a communitys emergency plans mayincrease if residents are assured that everyone, even themost vulnerable, is being included in a credible planningprocess and the resulting plans correspond to the commu-nitys specific needs. Actively involving those at risk shouldalso address the theoretical origins of this problem, namely,a "lack of personal control over potential outcomes7among lower-income households (Fothergill and Peek,~ 0 0 4 ) ~ I n V o ~ 1 ~ t ; f ~ p o p u l a t i o n sIn detining, learningabout and testing appropriate resources for preparednessand response would presumably heighten compliance aswell.As noted elsewhere (e.g., Cutter et a]., 2003a, p. 243),social vulnerability is "most often described using theindividual characteristics of people", the strategy that wasused here to identify those at greatest risk for specialassistance and training by emergency management autho-rities. Such research documents the presence of individualsat-risk and identifies potential points of intervention forpractitioners. However, the significantly higher concentra-tion of at-risk groups within the lowest income quartile
132 B.D.Phillips er crl. / Etluironn~enfclearly demonstrates that vulnerability does not occur duesimply to chance. Understanding why inequalities existdemands further investigation. Social inequality research,such as that based on sociological conflict theory, suggeststhat some of the findings reported here may be influencedby the larger socio-economic structure (Tierney et al., 2001;Peacock and Ragsdale, 1997; Cutter, 2005). Social pro-blems, such as poverty, inherently serve as the drivingforces behind vulnerability. Merely developing programs tomotivate preparedness and protective actions then isinsufficient. Continually addressing the conditions thatrender people vulnerable is part of the solution.AcknowledgementsThe submitted manuscript has been created by theUniversity of Chicago as Operator of Argonne NationalLaboratory ("Argonne") under Contract no. W-31-109-ENG-38 with the US Department of Energy. The USGovernment retains for itself, and others acting on itsbehalf, a paid-up, nonexclusive, irrevocable worldwidelicense in said article to reproduce, prepare derivativeworks, distribute copies to the public, and perform publiclyand display publicly, by or on behalf of the Government.This research was funded by the Alabama EmergencyManagement Agency under the direction of CharlesWilliams, Preparedness Division Chief. We would like tothank him for his energetic support and valuable guidancein this endeavor.ReferencesAuf der Heide, E., 1989. Disaster response: principles of preparation andcoordination. Available at http://www.coe-dmha.org/. Accessed 20March 2006.Baker, E.J., 1991. Hurricane evacuation behavior. International Journalof Mass Emergencies 9 (2). 287-310.Bragg, R., Wilson, G., 2002. Burning of Chemical Arms Puts Fear inWind. The New York Times (September 15).Cutter, S.L., 1991. Fleeing from harm: international trends in evacuationsfrom chemical accidents. International Journal of Mass Emergenciesand Disasters 9 (2), 267-285.Cutter, S.L., Boruff, B.J., Shirley, W.L., 2003a. Social vulnerability toenvironmental hazards. Social Science Quarterly 84 (I), 242-261.Cutter, S., 2005. The Geography of Social Vulnerability: Race, Class andCatastrophe. http://understandingkatrina.ssrc.org/Cutter/. Accessed20 March 2006.Cutter, S., et al., 2003b. American Hazardscapes. Joseph Henry Press,Washington, DC.Dawson, G., 1993.A comparison of research and practice: a practitionersview. International Journal of Mass Emergencies and Disasters I I (I),55-62.Diggory, J.C., 1956. Some consequences of proximity to a disease threat.Sociometry 19 (1,), 47-53.Dow, K., Cutter, S.L., 2002. Emerging hurricane evacuation issues:Hurricane Floyd and South Carolina. Natural Hazards Review 3 (I),12-18.Drabek, T.E., 1986. Human Systems Response to Disasters: an Inventoryof Sociological Findings. Springer, New York.Drabek, T.E., 1996. The social dimensions of disaster. EmergencyManagement Institute Instructor Guide. NTIS: PB97-115372. Em-mitsburg, MD.Dynes, R., 1994. Community emergency planning: false assumptions andinappropriate an;ilogies. International Journal of Mass Emergenciesand Disasters 12 (2), 141-158.Enarson, E., Morrow, B., 1997. A gendered perspective: the voices ofwomen. In: Peacock, W.G., Morrow, B.H., Gladwin, H. (Eds.).Hurricane Andrew: Ethnicity, Gender and the Sociology of Disasters.Routledge, London, pp. 116-140.Enarson, E., Morrow, B. (Eds.), 1998. The Gendered Terrain of Disaster:Through Womens Eyes. International Hurricane Center, Miami, FL.Enarson, E., Morrow, B., Childers, C., Wisner, B., Thomas, D., 2003.Social Vulnerability to Disasters. Federal Emergency Mi~nage~nentAgency Higher Education Project. FEMA. Washington DC.Feldman, D.L., 1990.Sara Title I1 and state response to chemical hazt~rds:gauging the effectiveness of public participation for emergencyplanning. The Environmental Professional 12, 67-76.Fothergill, A,, Peek, L.A., 2004. Poverty and disasters in the UnitedStates: a review of recent sociological findings. Natural Hi~zards32,89-1 10.Fothergill, A., Maestas, E.G.M., Darlington, J.D.R., 1999. Race, ethnicityand disasters in the United States: a review of the literature. Disasters23 (2). 156-173.Heinz Center, 2002. Human Links to Coastal Disasters. The Heinz Center,Washington, DC.Landers, J.. 2002. In the shadow of the incinerator: EMA works toidentify residents with special needs. Anniston Star, July 13.Legates, D., Biddle, M.D., 1999. Warning response and risk behavior inthe Oak Grove-Birmingham, Alabama Tornado of 08 April 1998.Quick Response Report #116, Natural Hazards Research endApplications Information Center. http://www.colorado.edu/hi~zards~Accessed 20 March 2006.Lindell, M., Perry, R., 1992. Behavioral Foundations of CommunityEmergency Management. Hemisphere Publishing Company. Washing-ton, DC.Lindell, M.K., Perry, R.W., 2004. Communicating Environmental Risk inMultiethnic Communities. Sage Publications. Newbury Park. CA.Metz. W., Hewett. P.. Muzz;~relli. J., Tanzman. E.. 2002. ldentifyingspecial-needs households that need assistance for emergency planning.International Journal of Mass Emergencies and Disaster 20 (2),255-28 1. Metz, W.C., Tanzman, E.A., Nieves, L.A., Holt, V., 2005. Preparingpeople with special needs for emergencies: the Alabama ChemicalStockpile Emergency Preparedness Program (CSEPP) Model. Journalof Emergency Management 3 (4), 36-48.Mileti, D., 1999. Disasters by Design. Joseph Henry Press, NationalAcademies, Washington DC.Morrow, B., 1999. Identifying and mapping community vulnerability.Disasters 23 (I), 1-18.Morrow, B.H., Phillips, B.D., 1999. Whats gender got to do with it?International Journal of Mass Emergencies and Disasters 17 (I), 5-15.Mulilis, J.P., 1999. Gender and earthquake preparedness. AustralianJournal of Emergency Management Autumn, 4 1-50.Natural Hazards Research and Applications Information Center, 2002.Holistic Disaster Recovery. NHRAIC and Public Entity Risk-Institute, Boulder, CO. Available at http://www.colorado.edu/ha-..--z a r d s / h o l i s t i c _ d i ~ c o v e r y .Accessed 13 March 2006.Nisbett, R., Ross, L., 1980. Awareness and behavior: public response toprediction awareness. In: Saarinen, T.F. (Ed.), Perspectives onincreasing Hazard Awareness. University of Colorado Institute ofBehavioral Science, Boulder.Peacock, W., Ragsdale, A.K., 1997. Social systems, ecological networksand disasters: toward a socio-political ecology of disasters. In:Peacock, W.G., Morrow, B.H., Gladwin, H. (Eds.), HurricaneAndrew: Ethnicity, Gender and the Sociology of Disasters. Routledge,London, pp. 20-35.Perry, R., 1985. Comprehensive Emergency Management: EvacuatingThreatened Populations. JAI Press, Inc, Greenwich, CT.Perry, R., Greene, R., 1983. Citizen Response to Volcanic Eruptions: TheCase of Mt. St. Helens. lrvington Publishers, New York.
B.D. Pllillips er ul. / Environmenrul Huzuriis 6 (2005) 123-133 133Perry, R., Mushkatel, D., 1986.Minority Citizens in Disasters. Universityof Georgia Press, Athens.Perry, R., Lindell, M., Greene, R., 1981. Evacuation Planning inEmergency Management. Lexington Books, Lexington, KY.Phillips, B.D., 1992. Planning for the expected: evacuation in a chemicalexplosion. Disaster Management 4, 103-108.Phillips, B.D., 1993. Cultural diversity in disasters: sheltering, housing,and long term recovery. International Journal of Mass Emergencies I I(I), 99-1 10.Phillips, B.D., 1998. Housing Low Income and Minority Groups afterLoma Prieta: Some Policy Considerations. In the National Report toCongress on the Loma Prieta Earthquake, US Geological Survey.Pine, J.C., Marx, B.D., Lakshmanan, A., 2002. An examination oraccidental-release scenarios from chemical-processing sites: the rela-tion of race to distance. Social Science Qui~rterly83 (I), 1-15.Quarantelli, E.L., 1984. Sociobehavioral responses to chemical hazards:preparation for and responses to acute chemical emergencies at thelocal community level. University of Delaware, Disaster ResearchCenter, Newark, DE.Quarantelli, E.L., 1985. Organizational behavior in disasters andimplications for disaster planning. Report series 18. University ofDelaware. Disaster Research Center, Newark, DE.Quarantelli, E.L., 1993. Converting disaster scholarship into effectivedisaster planning and managing: possibilities and li~niti~tions.Interna-tional Journal of Mass Emergencies and Disasters I I (I), 1541.Russell, L., Goltz, J.D., Bourque, L., 1995. Preparedness and hazardmitigation activities before and after two earthquakes. Environmentand Behavior 27, 77&774.Sattler, D.N., Kaiser, C.F.. Hittner, J.B., 2000. Disaster preparedness:relationships among prior experience, personal characteristics, anddistress. Journal of Applied Social Psychology 30 (7), 1396-1420.Sims, J., Bauman, D., 1972.The tornado threat: coping styles of the northand south. Science 176, 1386-1392.Sorensen, J., 1991. When shall we leave? Factors affecting the timing ofevacuation departures. International Journal of Mass Emergencies andDisasters 9 (2). 153-166.Tierney, K., Lindell, M., Perry, R., 2001. Facing the Unexpected.Altamira Press, London.US Departrnent of the Army and Federal Emergency ManagementAgency, 1996. Planning Guidance for the Chemical StockpileEmergency Preparedness Program. Washington, DC.US Departrnent of the Army and Federal Emergency ManagementAgency, 2000. Chemical Stockpile Emergency Preparedness Programjoint memorandum: Alabama Chemical Stockpile Emergency Programrisk criterion policy for protective action decisions, March 31.Washington, DC.Van Willigen, M., Edwards, T., Edwards, B., Hessee. S., 2002. Riding outthe storm: experiences of the physically disabled during HurricanesBonnie, Dennis and Floyd. Natural Hazards Review 3 (3). 98-106.Wisner, B., Blaikie. P.. Cannon, T.. Davis. 1.. 2004. At Risk: NaturalHazards, Peoples Vulnerability and Disasters, 2nd edn. Routledge.London.Wood, V.T., Weisman. R.A., 2003. A hole in the weather warning system:improving access to hazardous weather information for deaf and hardof hearing people. Bulletin of American Meteorological Society 84,187-194.Yelvington, K.. 1997. Coping in a temporary way: the tent cities. In:Peacock, W.G., Morrow, B.H., Gladwin, H. (Eds.), HurricaneAndrew: Ethnicity. Gender and the Sociology of Disi~sters.Interna-tional Hurricane Center. Miami. FL. pp. 92-1 15.