Safety Measures After the 2011 Joplin, Missouri, Tornado.

SAFETY MEASURES AFTER THE 2011 JOPLIN, MISSOURI,TORNADO*

BIMAL KANTI PAUL and MITCHEL STIMERS

ABSTRACT. Immediately following the May 2011 tornado, the city of Joplin, Missouri, ini-tiated recovery efforts to rebuild the part of the community devastated by this event. Indoing so, city officials introduced two new safety measures and recommended severalothers. The main objective of this paper is to explore the tornado survivors’ compliancewith the safety features recommended by the Joplin city authorities. Face-to-face inter-views, as well as other methods, were used to survey respondents who were living withinthe tornado path (damage zones) at the time of the tornado. Results showed that 43

percent of all respondents implemented at least one recommended tornado measure inrebuild or repair their destroyed/damaged homes. Further, it was found that three vari-ables considered here— rebuild or repair, damage zone category, and perceived tornadorisk—were all significant contributors to a homeowner’s decision to implement recom-mended tornado safety measures. The paper concludes with a recommendationfor expanding tornado safety education among the respondents. Keywords: May 2011tornado, Joplin, Missouri, tornado safety measures, recovery and reconstruction.

Recovery and reconstruction in Joplin, Missouri, began within one monthafter the city experienced a powerful EF-5 tornado on the evening of May 22,2011. This event directly impacted 13,547 people or 27 percent of the city’s pop-ulation, and nearly 3,500 others were indirectly impacted (Paul and Stimers2014). It caused a record number of deaths, killing 161 people, and damaged ordestroyed 7,500 homes and over 500 businesses, with property damage esti-mated to be $3 billion—the highest ever for a U.S. tornado (Smith and Sutter2013).1 No other single tornado in the United States has ever affected so manypeople, damaged/destroyed so many structures, or covered so large an area of asingle city, yet the community’s speed of recovery was relatively rapid.

By July 2013, the city issued repair and building permits for 6,719 (89 per-cent) of the 7,500 damaged or destroyed homes (City of Joplin 2013). Of those7,500 dwellings, approximately 4,000 homes (53 percent) were destroyed, andthe remaining 3,500 (47 percent) damaged (City of Joplin 2013). In addition,553 businesses were either destroyed or severely damaged. Also by June 2013,500 businesses (90 percent) had reopened or were in the processes of reopening(City of Joplin 2013). Wes Johnson, a USA Today reporter, who visited Joplintwo years after the event wrote: “Block after block, new homes have risen,

*The authors would like to acknowledge the support of the Association of American Geographers (AAG)grant in this project. We also thank members of the survey team for their hard work in collecting the dataused in this research, as well as two anonymous reviewers for providing insightful comments.

k DR. PAUL is professor of geography at Kansas State University, Manhattan, Kansas, 66506;[[email protected]]. DR. STIMERS is the director of Institutional Planning, Research, and Effectiveness,and instructor of geography and geoscience at Cloud Community College, Junction City, Kansas,66441; [[email protected]].

Geographical Review 105 (2): 199–215, April 2015Copyright © 2015 by the American Geographical Society of New York

many on the bare foundations left behind after the May 22, 2011, storm. Newroofs—not twisted debris—mark the path of the storm through the heart ofthe city. A recent tour of the tornado’s path revealed only a handful of struc-tures with torn roofs or damaged walls” (Johnson 2013).

The postdisaster context in the recovery phase of a disaster-managementcycle offers a “window of opportunity” for disaster-risk reduction (DRR) andimproved redevelopment (Mileti 1999; Wisner and others 2004). Emergencyagencies and disaster survivors use this phase to attempt to reduce vulnerabilityand increase safety against potential recurrence of a disaster (Platt 1998). Localgovernment, together with support from emergency agencies at the state andfederal level, generally introduce hazard-safety measures, such as land-use con-trols and new building codes, to strengthen construction and better prepare forfuture disasters (Prater and Lindell 2000; Tierney and others 2001; Paul 2011).The adoption of these safety measures reduces the risk of loss of life, injury,and property damage, as well as increasing disaster resilience for communitiesaffected by an extreme event (Mileti 1999; Haigh n.d.).

In general, disaster survivors are supportive of DRR in the immediateaftermath of an extreme event and comply with new safety measures enforcedand/or recommended by local government agencies. As well, recent disasterexperiences typically generate pledges of greater vigilance and safer behavioramong disaster survivors. Personal awareness of disaster risk is generallyraised, and commitment to future mitigation practices may appear strong(Haigh n.d.).

The Joplin city authorities appropriately took advantage of the “window ofopportunity” and introduced new and better tornado safety measures. Theauthorities enforced two safety measures—hurricane straps and anchor bolts—for new constructions. Although city authorities had not enforced these twomeasures prior to the 2011 tornado, some homeowners of Joplin had installedhurricane straps and anchor bolts on their own initiative when they con-structed their residences. In addition to the two enforced measures, the cityauthorities also recommended installation of either a safe room or storm shel-ter for rebuilt and repaired structures.2

The main objective of this paper is to explore the tornado survivors’ com-pliance with the safety features recommended by the Joplin city authorities.The specific objectives of this paper are to investigate the extent of households’compliance with recommended safety features in the rebuilding/repairing ofresidential structures, and to examine the factors influencing compliance withrecommended safety measures.

SAFETY MEASURES AND WINDOW OF OPPORTUNITY: AN OVERVIEW

There is a temporal distinction between hazard-safety measures introduced inthe recovery phase of a disaster and mitigation measures, although their pur-poses are similar. Hazard-safety measures are enforced and/or recommended by

200 GEOGRAPH ICAL REV IEW

local government agencies in the immediate aftermath of a disaster, while miti-gation measures are generally implemented after the recovery phase is com-pleted. The safety measures are introduced in the early part of the recoveryphase for several reasons. First, disaster survivors’ support of future DRRdecreases over time, and therefore the window of opportunity is limited (Paul2011). While the length of this time is not absolute, according to RobertOlshansky, the window of opportunity lasts at most seven months postdisaster(2006). Among other factors, the decrease of support is caused by a decrease inpublic awareness (Kingdon 1995).

Second, the political will to act in reducing the risks of future disasters ismuch stronger in the immediate aftermath of a disaster than during predisasterperiod. This results in more proactive political leadership. Similarly, interest inDRR either forgotten or sidelined before the most recent disaster is renewed inthe recovery period (Haigh n.d.). Third, external funding is generally availableduring the postdisaster period. This additional resource provides an opportu-nity to reduce hazard vulnerability at both household and community levels(Paul 2011).

Evidence suggests that despite the opportunities presented following a disas-ter, emergency agencies are not always able to take advantage of the window ofopportunity for many reasons (Haigh n.d.). The postdisaster period should beused to reduce not only disaster risks, but other stresses as well. Responses to adisaster are motivated by several principles and policy objectives. Christoplosclaims that, “Even if DRR would seem to have a more powerful stance in pol-icy arenas, those forces promoting other agendas are usually stronger, betterorganized and (above all) more firmly established in key decision-making insti-tutions” (2006, 2). Additionally, effective reconstruction of the built environ-ment in the postdisaster recovery period competes with many other priorities,such as poverty alleviation, income redistribution, gender equity, improvedhealth, sustainable development, and good governance. Many organizationspartake in implementing these goals during reconstruction and recovery, withall the confusion that can engender. Both Christoplos (2006) and Haigh (n.d.)suggest a bevy of obstacles that close the window of opportunity before it canbe used effectively. However, almost all disasters, such as tornadoes and earth-quakes, destroy and/or damage houses that were built to standards that arenow decades old and generally far less rigorous than the strict building codesof today (Paul and Stimers 2012). Thus, a disaster provides an extraordinaryopportunity to rebuild the affected community better and stronger than it wasbefore the disaster (Kavanaugh 2003).

For example, the city manager of Hoisington, Kansas, believes that the2001 F-4 tornado led to the modernization of many buildings and homes thatwere in need of renovation (Brock and Paul 2003). After the tornado, the cityenforced two new zoning ordinances that restricted the construction of homesto fifty-foot lots in the tornado-affected area. This compelled survivors who

JOPLIN TORNADO SAFETY MEASURES 201

wished to rebuild homes to purchase adjacent lots. As a result, lower housingdensity reduced tornado exposure by limiting the density of development (Paul2011).

Another zoning regulation enacted by Hoisington was that any house con-structed west of Alexandria Street could not have a basement. This area wasidentified by FEMA as being in a floodplain zone and, because no one wantedto be without a basement, no one rebuilt his or her home in the area. The citythen bought this land and converted it into a park with playground equipmentand tennis courts. All 2,000 residents of the city welcomed this change in landuse (Paul 2011). The city officials also encouraged residents to upgrade thequality of construction for new residential buildings to better withstand futuretornadoes. An overwhelming majority of the owners of new homes added tor-nado safety features, such as hurricane straps and foundation brackets. Nearly90 percent added a basement and 30 percent of the rebuilt homes included saferooms (Brock and Paul 2003).

On the evening of May 4, 2007, an EF-5 tornado hit Greensburg, a south-western Kansas town of 1,400 people. It killed 11 residents and destroyed 95

percent of this small town, including its downtown area (Paul and others2009). After the tornado, the city officials of Greensburg not only rebuilt theaffected community stronger and better, but also enforced upgraded codes,requiring city-owned buildings to meet Leadership in Energy and Environmen-tal Design (LEED) Platinum level standards, and recommended energy-efficienthousing (Paul and Che 2011). Rebuilding methods and considerations alsoincluded the use of recycled construction materials and highly insulated materi-als, and the use of alternative-energy sources, such as solar panels, solar hot-water systems, and small wind turbines (Paul and Che 2011).

DATA SOURCES

The major data set used in this study came from a questionnaire survey admin-istered by four researchers to residents of the 2011 tornado-affected areas ofJoplin, Missouri. Surveys were administered face-to-face during June and July2013. Due largely to the unavailability of tornado survivors and unwillingnessto participate in the survey, a combination of convenience and snowball sam-pling was used to identify residents located within the damaged zones for inter-views. On average, one in three residents contacted for an interview agreed toparticipate in the survey. The overwhelming majority of the nonparticipantsmentioned that they did not want to talk about the tornado. By contrast, somerespondents participated in the survey as a way to verbalize and better copewith their losses.

A prestructured interview schedule was used to collect relevant informationfrom the eligible respondents. The questionnaire consisted of twenty-nine ques-tions divided into three sections: structure damage and rebuilding; risk andrecovery perception; and relevant demographic and socioeconomic questions to

202 GEOGRAPHICAL REV I EW

collect information on gender, age, marital status, educational attainment,employment status, and annual household income.

The data collected contained no identifiers, resident participation in thestudy interview was voluntary and noncoercive, and participant confidentialitywas respected throughout the study. Four trained personnel conducted face-to-face interviews among 102 respondents. Interviews were conducted at therespondents’ residences. In addition, nineteen interviews were completed usinga combination of telephone and Facebook, for a total of 121 responses. It isworthwhile to mention that renter residents of the tornado-affected areas werenot included as respondents in this study. According to the City of Joplin,some renter landlords took insurance money and did not rebuild houses inJoplin (2013). The city further claims that between 6 and 10 percent of theaffected households built their houses outside the Joplin city limit (personalcommunication, Steve Cope, City of Joplin building-code supervisor).

In addition to houses built by owners, at least two-dozen homes were builtby national charities, faith-based organizations, and the volunteer sector. Forexample, the Home Edition program of ABC-TV’s Extreme Makeover builtseven homes in seven days in October 2011. Habitat for Humanity built tenhomes by November of the same year. Nonprofit organizations were planningon building at least 100 more homes as of early 2012 (McKinney 2012; Smithand Sutter 2013). Occupants of these houses were also excluded from thisstudy.

Relevant data was also collected from local emergency-management person-nel, Joplin city officials, community leaders, and residents affected by the tor-nado through informal interviews. The Joplin Globe, the local daily newspaper,and other print media, including Joplin city reports, provided valuable infor-mation for this study.

ANALYTICAL FRAMEWORK AND EMPIRICAL MODELING

This section outlines an expected utility framework for household behavior tomitigate the effects of environmental stresses, such as poor water quality (Jan-maat 2007) or to reduce health risks, such as skin cancer (Dickie and Gerking1996). The discussion of this framework is kept to a minimum given that simi-lar theoretical models of aversion behavior that have been presented elsewhere(see Dickie and Gerking 1996; Janmut 2007). However, no one has yet appliedthis framework to studies of tornado safety measures. Considering this, it isused here to examine the factors associated with the adoption of city or self-recommended tornado safety measures in Joplin, Missouri.

In this theoretical model, households derive utility from consumption ofcomposite good (X): all goods a household needs to consume daily, monthly,or yearly. It is assumed that perceived tornado risks (PR) have a negativeimpact on household utility, because of potential property losses, deaths, inju-ries, and psychological stress. Perceived tornado risks are directly associated

JOPLIN TORNADO SAFETY MEASURES 203

with the actual probability of being affected by a tornado (T) and inverselyrelated to adopting aversion measures to cope with tornados (A). Householdschoose the levels of composite good and aversion measures that maximize theirutility subject to their budget constraint as follows:

MaxX;A ¼ UfX;PRðT;AÞgs:t:PAAþ PXX ¼ I

ð1Þ

where PA = cost of implementing recommended tornado safety measures,PX = the price of composite good, and I = household income.

The solution to this household problem implies that the optimal choice ofaversion-risk behaviors (A*) by a household depends on tornado risk, house-hold income, costs of implementing tornado safety measures, and the price ofthe composite good:

A� ¼ AðT; I; PA; PXÞ ð2Þ

Given that perceived risks increase with the probability of being affected bya tornado, it is expected that households will implement tornado safety behav-ior to reduce perceived risks (i.e. dA/dT < 0) and thus maximize their utility.Consistent with standard utility theory, the adoption of a given safety measuredecreases with its costs (dA/dPA < 0), and increases with household income(dA/ dI > 0). The adoption of safety measures will also increase if householdincome is higher than the price of a composite good (I > PX).

Adoption/implementation of recommended tornado safety measures ismodelled following a logit specification:

ln½P=ð1� PÞ� ¼ aþ bX ð3Þ

Here P = probability of implementing a recommended tornado safety mea-sure, which is expressed in binary form: yes (1) or no (0); and X = matrix ofcovariates. Based on the theoretical framework, relevant literature survey, andour experience regarding the Joplin tornado event, we selected seven indepen-dent variables: construction type (rebuilding or repairing), household locationin terms of four damage zones, household income, household size, level of edu-cation and employment status of head of the household, and perceived tornadorisk. The last variable is trichotomized and derived from a question aimed atcollecting data regarding the respondent’s perceived fear towards tornados afterthe May 2011 tornado. After the tornado, the Federal Emergency ManagementAgency (FEMA) divided the width of the tornado path in descending order ofdamage into four zones: catastrophic, extensive, limited, and moderate (Pauland Stimers 2014). These are considered damage zones in this study.

204 GEOGRAPH ICAL REV IEW

The variable prices of composite goods were not included in this study,because the data was not collected. The cost of implementing safety measureswas also excluded, because almost all households who had implemented themopted for high-priced measures.

Before applying the logit regression, differences in frequency distribution ofthe dependent and each of the selected independent variables were tested bythe chi-square test. Independent variables included in the logit model were allstatistically associated (p < 0.05) with the dependent variable in the bivariateanalysis. The statistical software STRATA was used for analyzing the quantita-tive data, and the analysis results are presented in both tabular and graphicalform.

RESULTS

The analysis of survey data indicates that the median respondent age was 49.5,and the average household size was 2.63 (median of 2). The majority of respon-dents were married (51 percent), with single (28 percent) being the secondmost-populated category in marital status. A high school diploma was thehighest reported level of education (62 percent), followed by a bachelor’sdegree (35 percent). Close to half of the respondents reported working full-time(49 percent), about a quarter reported themselves as retired (25 percent), and alittle over 13 percent reported they were employed part time. Concerning sectorof employment, the business sector was the most frequent response, followedby a tie between service and medical sectors. An annual household income of$20,000 to $39,999 was reported by 42 percent of the respondents.

MANDATED SAFETY MEASURES

Joplin city authorities enforced only two (modest) safety measures for new resi-dential constructions: hurricane straps (also referred to as hurricane clips ortornado straps) and anchor bolts (also referred to as anchor rods or J-bolts,depending on the specific type used). Hurricane straps anchor the top plate ofthe home to the rafters, keeping the roof more secure in high winds. Generally,a roof is set on top of the walls and nailed into place, but the nails merely keepthe roof from subtlety shifting; they do not really anchor the two sections ofthe structure together. The strapping connects the roof to the wall along twoplanes of contact, which in turn helps maintain structural integrity when theroof is stressed. Anchor bolts, as the name implies, anchor the bottom plate—the sill plate—of a house to the stem wall. Sinking anchor bolts into wet con-crete during the construction of stem walls allows them to be secured andlocked in place as the concrete cures, with the sill plating then attached to theconcrete footing or wall via the bolt. Joplin authorities mandated a reductionin the spacing between anchor bolts from eight inches to six inches. Both mea-sures are relatively inexpensive—hurricane straps sell for as low as $0.50 apiece,while each anchor bolt costs slightly over $1. This means that installations of

JOPLIN TORNADO SAFETY MEASURES 205

these two safety measures for a medium-sized new home would not cost morethan $1,000.

There are many other ways to protect homes against tornadoes, such asreinforced concrete, steel layered between plywood, and concrete-filled masonryblocks. But Joplin officials deliberately chose not to mandate too many safetymeasures to avoid placing added expenses on an economically vulnerable popu-lation. However, the phrase “too many” was not operationally defined by anycity official, so it remained unclear to the researchers exactly what, to the cityof Joplin, constituted the proper or adequate number of new measures tointroduce in the wake of a disaster. Additionally, to the best of the researchers’knowledge, the city of Joplin performed no cost-benefit analysis to determinethe extent to which homeowners would be willing to add new measures. Thecity’s elected leaders enforced measures that appeared to be affordable to allhomeowners, In doing so, they hoped to achieve recovery in the shortest possi-ble time while increasing the likelihood that displaced households would returnand rebuild in Joplin. Joplin officials did not mandate the installation of saferooms, which would increase the cost of rebuilding by at least $10,000 perhome. Because of the high water table, rocky ground, and concerns about oldmining tunnels, the city also did not recommend construction of basements asa tornado safety measures (Paul and Stimers 2012).

Nor did the city enforce restrictive zoning rules. In fact, city officials unoffi-cially waived building regulations, procedures, and local zoning laws in theimmediate aftermath of the tornado (Smith and Sutter 2013). The city council’sactions indicated preferences for minimal intervention in the recovery process,including lifting the moratorium—the council implemented a moratorium onnew residential construction within the disaster zone for 60 days (Smith andSutter 2013)—as portions of the debris removal zone were cleared.

RECOMMENDED SAFETY MEASURES

As indicated, Joplin city authorities mandated two tornado safety measures tothe owners of both destroyed and damaged homes. However, there was a largepush by the city to incorporate other recommended safety measures. Of the 121

homeowners, 52 (43 percent) implemented at least one of the following city- orself-recommended safety measures: safe room, storm shelter, double-pane glasswindow, reinforced closet door, and walls of concrete (Figure 1). Four respon-dents adopted more than one recommended safety measure. Although double-pane windows reduce energy consumption, three respondents installed this typeof window because they are thought to be stronger and better able to provideprotection from strong winds.

As noted, 43 percent of all respondents implemented at least one recom-mended tornado safety measure, either in constructing or repairing theirdestroyed/damaged homes. This is commendable, because 92 percent of theabove respondents spent at least $5,000 on implementing the measure. Figure 1

206 GEOGRAPH ICAL REV I EW

clearly shows that adding a safe room was the most popular measure among allfive non-mandatory safety measures adopted by the respondents of this study.Thirty one (60 percent) of the 52 respondents installed a safe room eitherwithin their homes or outside the home, but on the premises.

Of the 31 safe rooms, 29 (95 percent) of these were built in newly con-structed houses. It is reported that no more than 3 percent of houses across theUnited States have a safe room. Younker and Kabel reported that 15 percent ofall new houses built in the Midwest have safe rooms (2011). This implies thatJoplin city authorities were successful in persuading tornado survivors toimplement this costly safety measure. In the case of storm shelters, only newhomes included this precaution, with no reported instances of retrofitting ahome to include a shelter.

The relative popularity of the safe room among all recommended safetymeasures was because respondents considered a safe room an alternative to abasement. As noted, the rocky soil of Joplin is not suitable for building base-ments, so many turned to safe room. Additionally, after the tornado, severalsafe-room manufacturers—such as the Springfield, Missouri-based, MissouriStorm Shelter—exhibited their products in the local mall and persuaded Joplinresidents located in the damaged zone of their importance. One local firminformed us that after the tornado, its number of employees increased fromfour to twenty and the firm sold at least four times as many of the rooms asthey did the year before.

The survey data reveals that of all the respondents who installed a saferoom, the overwhelming majority of them built it within their homes. Onlyfive respondents installed a safe room outside their dwelling (Figure 2).Although prices varied according to the size, the cost of installing a safe roomoutside a home was about $5,000 higher than installing within the home. Thisprice difference may explain why more respondents opted for an in-house saferoom over an external safe room.

FIG. 1—Number of respondents who implemented city- or self-recommended tornado safetymeasures (N = 121).

JOPLIN TORNADO SAFETY MEASURES 207

Figure 1 further shows that seventeen of the fifty-two respondents installeda storm/tornado shelter or storm cellar, a type of underground bunkerdesigned to protect the occupants from tornadoes (Figure 3). An average stormshelter for a single family built close to the home costs around $5,000. Such ashelter has only one door, which is mounted at an angle rather than perpendic-ular to the ground. An angled door allows for debris to blow up and over thedoor, or soil to slide off without blocking the door. The angle also reduces theforce necessary to open the door if rubble has piled on top. Floor area is gener-ally eight-by-twelve feet with an arched roof, but entirely underground. In mostcases, the entire structure is built of cement-faced blocks with reinforced bar(rebar) threaded through the bricks for stability, making it nearly impossiblefor the bricks to tumble.

Because storm shelters are built underground, this safety measure appearsmore common in high-elevation areas of Joplin. Joplin city contains many lowhills, with elevation ranges between 968 and 1,116 feet; this is another reason forthe popularity of safe rooms in Joplin. Several respondents built a safe roomunder their garage. Yet, as a safety measure, no one installed stormproof garagedoors that can withstand high winds. High wind may buckle standard garagedoors, allowing wind to easily penetrate the garage, rip apart a roof, anddestroy the structure. Thus, protecting the garage means protecting the home.

FACTORS INFLUENCING COMPLIANCE WITH RECOMMENDED TORNADO

SAFETY MEASURES

Table 1, which presents the results of bivariate analysis, shows that 74 (61 per-cent) of the 121 respondents built new houses, while the remaining 47 (39 per-cent) respondents repaired their homes at the time of field survey. The tablefurther shows that 43 percent of all respondents implemented one recom-

FIG. 2—Building a safe room outside a dwelling in Joplin (Photograph by Mitchel Stimers,October 2013)

208 GEOGRAPHICAL REV I EW

mended tornado safety measure. As expected, the respondents who constructednew houses were more in favor of implementing such measures: 64 percent ofthese respondents implemented one recommended tornado safety measure. Thecorresponding percentage was only 11 among respondents who repaired theirdamages homes. This difference is statistically highly significant because the cal-culated chi-square value is significant at the <0.000 level.

Information presented in Table 1 shows that implementation of a safetymeasure also significantly differs by tornado damage zone. Based on the extentof damage, the Federal Emergency Management Agency (FEMA) divided thewidth of the Joplin tornado path into four zones: catastrophic, extensive, lim-ited, and moderate (Paul and Stimers 2014). Within the overall damage zone(tornado path), damage consistently decreased on both sides with distanceaway from the central (catastrophic) zone. Because the overwhelming majorityof the respondents were from two zones (catastrophic and extensive), the dam-age zones were dichotomized in Table 1. Consistent with expectation, imple-mentation of a tornado safety measure was much higher among catastrophiczone respondents compared to the respondents of other zones (62 percent ver-sus 17 percent).

FIG. 3—A storm cellar in Joplin (Photograph by Mitchel Stimers, October 2013)

JOPLIN TORNADO SAFETY MEASURES 209

Considering that the annual household income was between $20,000 and$39,999 for nearly half of the respondents, their income is dichotomized using$39,999 as a cutoff point (Table 1).3 Surprisingly, the table shows that the adop-tion of a recommended tornado safety measures was 16 percent higher amongrespondents with lower income than those with higher income; however, thisdifference is not statistically significant. Field observation and further analysisof household income reveal that relatively more respondents who built homeswere from the lower-income category. Houses of respondents belonging to thehigher-income category were newer and stronger, and thus their houses experi-enced relatively more partial damage than total damage. As noted, the compli-

TABLE 1—IMPLEMENTATION OF RECOMMENDED TORNADO SAFETY MEASURES BY SELECTED VARIABLES

VARIABLE

IMPLEMENTATION STATUS

YES NO TOTAL

Type of construction

Rebuilding 47 27 74

Repairing 5 42 47

Chi-square = 32.792; d.f. = 1; p = <0.000Damage zone

Catastrophic (central) zone 43 26 69

Noncatastrophic zone 9 43 52

Chi-square = 24.51; d.f. = 1; p = <0.000Income

Up to $39,999 30 30 60

>$39,999 22 39 61

Chi-square = 1.86; d.f. = 1; p = 0.173

Household size

1-2 persons 29 41 70

>2 persons 23 28 51

Chi-square = 0.16; d.f. = 1; p = 0.689

Education

Up to high school 40 47 87

Above high school 12 22 34

Chi-square = 1.138; d.f. = 1; p = 0.286

Occupation

Employed (full-time) 22 34 56

Employed (part-time) 8 9 17

Retired 15 19 34

Others 7 7 14

Chi-square = 0.73; d.f. = 3; p = 0.866

Perceived risk

Very high 36 23 59

High 9 24 33

Medium-to-low 7 22 29

Chi-square = 10.21; d.f. = 2; p = 0.006

210 GEOGRAPHICAL REV IEW

ance with recommended tornado measures was higher among respondents whoexperienced total damage compared to their counterparts. David Roueche andDavid Prevatt reported that in 2011, 64 percent of the houses in Joplin wereover thirty years old, and many of them suffered from the added effects ofaging and deferred maintenance (2013).

Among the four other variables included in Table 1, only “fear after the2011 event”—perceived risk—appears as a statistically significant variable influ-encing a respondent’s decision to adopt a tornado safety measure. This variableis trichotomized as: high-perceived risk (“much more afraid of tornados nowthan was before the 2011 event”), medium-perceived risk (“somewhat moreafraid of tornados now than was before the 2011 event”), and low-perceived risk(“somewhat less afraid of tornado now than was before the 2011 event”). Con-versations with the respondents revealed that some believed that once a tor-nado hit a community, the likelihood of that community being struck again inthe near future was very low. For this reason, their perceived tornado risk waslow.

It appears that fear after the 2011 tornado event was high across the incomelevel of respondents, their household size, educational level, and occupation.However, fear was higher among respondents of catastrophic damage zone.Thus, the bivariate analysis of data clearly suggests that socioeconomic anddemographic factors are less important in the case of compliance with tornadosafety measures. Of the seven factors in Table 1, only three—rebuild or repair,damage zone, and perceived risk—are statistically significant. These three inde-pendent variables were included in the logistic regression analysis.

The results of logistic regression analysis are presented in Table 2. Althoughwe experimented with various interactions of the independent variables, theadditive model with three independent variables recorded the lowest AkaikeInformation Criterion (AIC) value and thus we selected the latter model. Theselected model is statistically significant: LR chi-square = 54.40, p < 0.000.Although, as opposed to ordinary least square (OLS) regression, there is notrue R2 in logit regression, Table 2 shows that pseudo R2 value as 0.33, whichmeans that the three independent variables explained about 33 percent of thetotal variation in deciding to implement tornado safety measures.

Table 2 further shows that one independent variable, the damage zone, isnot statistically significant. A careful review of data reveals that the type of con-struction (rebuild or repair) and the damage zone are correlated. The calcula-tion of phi coefficient confirmed that the two variables are significant at the0.073 level. We retained both variables because to drop one of the two offendedvariables increases the AIC value. As indicated, three categories are used for theperceived-risk variable. The category designated as “Medium” is also not statis-tically significant (Table 2). All other variables are significant at the 0.003 orlower level, including the intercept.

JOPLIN TORNADO SAFETY MEASURES 211

In addition to regression coefficients, Odds Ratios (OR) are included inTable 2. For all variables, the lowest category is used as a reference category.The table shows that the adoption of recommended tornado safety measuresincreases by a factor of 11.28 for those who constructed new houses comparedto those who repaired their homes. Similarly, for homeowners living in the cat-astrophic damage zone, their adoption of safety measures increases by a factorof 2.39. Such an increase is a factor of 6.21 for respondents whose risk percep-tion was reported as high compared to those who rated their risk perception aslow. The logistic regression results clearly suggest that all independent variablesare important determinants for the implementation of tornado safety measuresin Joplin. However, the most important determinant is shown to be type ofconstruction, followed by perceived risk, and, finally, damage zone.

DISCUSSION AND CONCLUSION

The 2011 EF-5 tornado provided an opportunity for Joplin city authorities toenforce and recommend safety measures for newly constructed, as well asrepaired, houses. The evidence produced here suggests that the city has seen arapid rebuilding despite the extensive damage. The city enforced only two rela-tively inexpensive tornado safety measures, but recommended several others.City officials and elected leaders opted for just modest improvement to build-ing codes, primarily the use of metal reinforcing straps on roof rafters andanchor bolts to provide more stability during high winds. This seems to be aprudent idea in light of the overall damage that occurred within the tornadopath, because most of the houses damaged or destroyed within the damagezones either lacked fasteners to the foundation or the fasteners they had failed(personal communication, Steve Cope, City of Joplin building-code supervisor;see also Younker and Kabel 2011). In order to encourage totally and partiallydisplaced homeowners to rebuild and relatively quickly, no other restrictions,such as land-use change, were imposed. The city even resisted the temptationto require “safe rooms” in rebuilding (Beito and Smith 2012). Given the fact

TABLE 2—RESULTS OF LOGISTIC REGRESSION

VARIABLE COEFFICIENT ODDS RATIO

95% CONFIDENCE

INTERVAL

Construction type 2.423** 11.282 3.056-41.649

Damage zone 0.870 2.387 0.751-7.590

Perceived risk

Medium 0.023 1.023 0.279-3.747

High 1.826* 6.209 1.851-20.831

Constant -3.447** 0.032 0.007-0.140

LRv2 = 54.40

Log likelihood = 55.472

Pseudo R2 = 0.33

**p < 0.01; *p < 0.05.

212 GEOGRAPHICAL REV IEW

that 90 percent of tornados are EF-2s or lower, the city did not feel the need toenforce such an expensive safety measure.

Through conversations with city officials, it appears they wanted homeown-ers to rebuild rapidly and they suspected that delay in building due to imposi-tion of too many restrictive and expensive measures would lead some tornadosurvivors to relocate to neighboring communities. Faster rebuilding and recov-ery seemed more important to the policy of the city than did stronger and saferrebuilding. Some hazard researchers criticized this approach, although the citywas successful in pursuing its rebuilding policy (for example, Meyer 2012).These researchers maintain that enforcing few building codes is a risky proposi-tion that could bring long-term costs to homeowners.

In accordance with its rebuilding priority, Joplin city authorities did notrecommend many tornado safety measures. The field data shows that 43 per-cent of all respondents implemented one or more recommended tornado safetymeasures as they repaired or reconstructed their old damaged or destroyedhouses; each homeowner spent at least $5,000 on those endeavors. The percent-age of tornado surviving homeowners complying with recommended safetymeasures exceeded our expectations. Such expectations were grounded in thefact that people who experienced a recent disaster generally assume that theevent will not occur again in the near or distant future and thus they are reluc-tant to adopt any new safety measures during the postdisaster period. Thisstudy, however, clearly suggests that many survivor households in Joplin didnot discount the likelihood of future tornados. Few implemented tornadosafety measures “only for peace of mind.”

During our conversations, at least 50 percent of homeowners in the tornadodamage zones expressed their satisfaction toward the city for not burdeningthem with too many regulations. Some respondents who did not comply withthe recommended safety measures indicated that financial constraint was themain reason. Only a few of these respondents were from the central damagezone.

A considerable number of respondents did not respond positively becausetheir damaged houses needed only minor repair. Consistent with our expecta-tions, nearly one dozen respondents did not comply with the city’s recom-mended tornado safety measures, because they believed that once people haveexperienced a tornado, they do not expect another tornado for another forty tosixty years. Conversations with these respondents revealed that they were notlong-time residents of Joplin, and, therefore, not familiar with the area’s histor-ical tornado activity: 199 percent greater than the overall U.S. average. Severeweather is common in this area during spring.

Statistical analyses of field data clearly suggest that of the seven variablesconsidered in this study, three variables—rebuild or repair, damage zone, andperceived tornado risk—strongly influenced a respondent’s decision to complywith recommended tornado safety measures. This implies that socioeconomic

JOPLIN TORNADO SAFETY MEASURES 213

and demographic factors are less important in complying with such measures.The study further shows that the implementation of safety measures is lesslikely among the respondents who repaired their damaged houses, lived innon-catastrophic zones, and had low-perceived risk compared to their counter-parts. This is an important finding because it points to the need for expansionof tornado-safety education among the groups less responsive to the recom-mended tornado safety measures.

NOTES

1 No single tornado from 1980 to 2010 killed more than forty people in the United States(Simmons and Sutter 2012). Over the last thirty years, tornado fatalities in the country have aver-aged around fifty-five per year; the Joplin tornado alone caused nearly three times that average(Simmons and Sutter 2011).

2 Safe rooms are made of thick concrete walls and ceilings, reinforced with steel bars. A saferoom can be installed in a basement, in the center of the ground level of a home without a base-ment, or under a garage. Storm shelters serve the same purpose as safe rooms, but they are gen-erally constructed underground either next to or inside a house. Unlike an indoor safe room, thestorm shelter inside a home is not built with concrete walls. Simply fortifying a closet door canqualify as an indoor storm shelter. Underground storm shelters are usually safe from flying deb-ris and high winds.

3 Considering the sample size, the variable was not categorized in more than two groups.

REFERENCES

Beito, D. T., and D. J. Smith. 2012. Tornado Recovery: How Joplin is Beating Tuscaloosa. WallStreet Journal, 13 April.

Brock, T. V., and B. K. Paul. 2003. Public Response to a Tornado Disaster: The Case ofHoisington, KS. Papers and Proceedings of the Applied Geography Conferences 26: 343–351.

City of Joplin. 2013. Joplin, Missouri Hit by EF-5 Tornado on May 22, 2011: General Information.1 July.

Christoplos, I. 2006. The Elusive ‘Window of Opportunity’ for Risk Reduction in Post-DisasterRecovery. Discussion Paper. Bangkok, Thailand: Prevention Consortium Forum 2006.

Dickie, M., and S. Gerking. 1996. Formation of Risk Beliefs, Joint Production and Willingness toPay to Avoid Skin Cancer. Review of Economics and Statistics 78 (3): 451–463.

Haigh, R. n.d. Disaster Management Lifecycle. Manchester: Centre for Disaster Resilience.Manchester, U.K.: University of Salford.

Janmaat, J. 2007. A Little Knowledge: Household Water Quality in Investment in the AnnapolisValley. Canadian Journal of Agricultural Economics 55 (2): 233–253.

Johnson, W. 2013. Debris is Gone and New Projects Mark the City’s Ongoing Recovery. USAToday, 21 May.

Kavanaugh, L. H. 2003. Town Battle But Not Broken. The Kansas City Star, 18 May, B1 and B4.Kingdon, J. W. 1995. Agenda, Alternatives and Public Policy. New York: Harper Collins.McKinney, R. 2012. More than 100 Houses in Works through Nonprofits, Other Groups. Joplin

Globe, 1 February.Meyer, J. 2012. Wall Street Journal: Joplin and Tuscaloosa’s Approaches to Disaster Recovery.

Uncategorized, 22 April.Mileti, D. S. 1999. Disaster by Design: A Reassessment of Natural Hazards in the United States.

Washington, D.C.: Joseph Henry Press.Olshansky, R. B. 2006. Planning after Hurricane Katrina. Journal of American Planning

Association 72 (2): 147–153.Paul, B. K. 2011. Environmental Hazards and Disasters: Contexts, Perspectives and Management.

Hoboken, N.J.: Wiley-Blackwell.-., and D. Che. 2011. Opportunity and Challenges in Rebuilding Tornado-Impacted

Greensburg, Kansas as “Stronger, Better, and Greener”. GeoJournal 76 (1): 93–108.

214 GEOGRAPHICAL REV I EW

-., and M. Stimers. 2012. Exploring Probable Reasons for Record Fatalities: The Case of2011 Joplin, Missouri, Tornado. Natural Hazards 64: 1511–1526.-., and-. 2014. Spatial Analyses of the 2011 Joplin Tornado Mortality: Deaths by

Interpolated Damage Zones and Location of Victims. Weather, Climate and Society 6:161–174.-., D. Che, M. Stimers, and S. Dutt. 2009. Disaster in Kansas: The Tornado in Greensburg.

Quick Response Report #196. Boulder, Colo.: The Natural Hazards Center, University ofColorado.

Platt, Rutherford. 1998. Planning and Land Use Adjustments in Historical Perspective. InCooperating with Nature, edited by Raymond Burby, 29–56. Washington, D.C.: Joseph HenryPress.

Prater, C. S., and M. K. Lindell. 2000. Politics of Hazard Mitigation. Natural Hazards Review 1(2): 73–82.

Roueche, D. B., and D. O. Prevatt. 2013. Residential Damage Patterns Following the 2011Tuscaloosa, AL and Joplin, MO Tornadoes. Journal of Disaster Research 8 (6): 1061–1067.

Simmons, K. M., and D. Sutter. 2011. Economic and Societal Impacts of Tornadoes. Boston:American Metrological Society.-., and-. 2012. Deadly Season: Analysis of the 2011 Tornado Outbreaks. Boston:

American Metrological Society.Smith, D. J., and D. Sutter. 2013. Response and Recovery after the Joplin Tornado: Lessons

Applied and Lessons Learned. Independent Review 18 (2): 165–188.Tierney, K., M. K. Lindell, and R. W. Perry. 2001. Facing the Unexpected: Disaster Preparedness

and Response in the United States. Washington, D.C.: Joseph Henry Press.Wisner, B., P. Blaikie, T. Cannon, and I. Davis. 2004. At Risk: Natural Hazards, People’s

Vulnerability and Disasters. London: Routledge.Younker, E., and M. Kabel. 2011. Joplin Homebuilder Celebrates Tax-Credit Project. Joplin Globe,

16 June.

JOPLIN TORNADO SAFETY MEASURES 215