|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 SOCIALLY VULNERABLE POPULATIONS AND THE HURRICANE PREPARATION DECISION PROCESS By MOLLY KAY MOON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Molly Kay Moon
3 To my husband
4 ACKNOWLEDGMENTS I would like to thank my entire committee, Dr. Marilyn (Mickie) Swisher, Dr. Francisco Escobedo, Joan Flocks, Dr. Lionel (Bo) Beaulieu, and Dr. Larry Forthun for all of their help and support in completing this dissertation. A special thanks to Mickie for encouraging me to pursue a doctorate. Francisco and Joan hav e also been with me from the beginning and I thank them for hanging in there. It seems so long ago that I started down this road. Also, thanks to Bo and Larry for coming in at the last minute. Your contributions and insights were greatly appreciated. I also need to thank some friends that have encouraged and commiserated with me along the road. I would like to thank Brad Thompson, Sherith Colverson, and Fernanda Pernambuco. Their support, and sometimes a shoulder to lean on, helped keep me on track. Finally, I need to thank my husband, Michael Droll. During all the emotional ups and downs of the process, he was always there for me. I thank and love him for that.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...................................................................................................... 4 LIST OF TABLES ................................................................................................................ 8 LIST OF FIGURES .............................................................................................................. 9 ABSTRACT ........................................................................................................................ 10 C H APT ER 1 INTRODUCTION ........................................................................................................ 12 Distant Memories ........................................................................................................ 12 Hurricanes and Florida ............................................................................................... 1 4 Current Florida guidelines .......................................................................................... 17 Why and How do Individuals Prepare for Disaster? .................................................. 18 Research Questions ................................................................................................... 20 Summary ..................................................................................................................... 21 2 LITERATURE REVIEW AND THEORETICAL FRAMEWORK ................................. 22 Disasters and Social Vulnerability .............................................................................. 22 What is Disaster? ................................................................................................. 22 Social Vulnerability ............................................................................................... 27 Theoretical Framework ............................................................................................... 31 Social Cognitive Theory ....................................................................................... 31 Transactional Model of Stress and Coping ......................................................... 34 Social Networ k Theory ......................................................................................... 37 A Proposed Social -Cognitive Preparation Model ...................................................... 39 Research Concepts and Hypothesized Model .......................................................... 41 Disaster ................................................................................................................ 41 Social Vulnerability ............................................................................................... 42 Hypothesized Model ............................................................................................ 43 3 METHODOLOGY ....................................................................................................... 45 Conceptual Framework .............................................................................................. 45 Research Design ........................................................................................................ 46 Unit of Analysis ........................................................................................................... 48 Sample and Site Selection ......................................................................................... 49 Instrument Development ............................................................................................ 54 Concepts and Variables ............................................................................................. 58 Predictor Variables ............................................................................................... 58 Motivational Factors ............................................................................................. 58
6 Coping Self Assessment ...................................................................................... 61 Moderator Variables: Response Efficacy, Past Experience, Perceived Responsibility: ................................................................................................... 66 Outcome Variable ................................................................................................ 68 Analyses ...................................................................................................................... 70 Demographics ...................................................................................................... 70 Comparison of Central Tendency ........................................................................ 70 Pearson Correlation ............................................................................................. 70 Structural Equation Modeling .............................................................................. 70 Summary ..................................................................................................................... 71 4 RESULTS .................................................................................................................... 72 Reliability Testing of Instruments ............................................................................... 72 Preliminary Screening of Cases ................................................................................. 74 Frequency of Response Data .................................................................................... 76 SocioDemographic Characteristics .................................................................... 76 Hurricane Experience .......................................................................................... 77 Analyses and HypothesesTesting Results ................................................................ 79 Additional Analyses of Full Sample ..................................................................... 92 Summary ..................................................................................................................... 93 5 DISCUSSION AND CONCLUSION ........................................................................... 95 Discussion of Hypotheses .......................................................................................... 95 Theoretical Retrospective ......................................................................................... 103 Social Cognitive Theory ..................................................................................... 103 Stress and Coping .............................................................................................. 105 Social Network Theory ....................................................................................... 106 Policy Implications .................................................................................................... 107 Limitations of the Study ............................................................................................ 109 Future Research ....................................................................................................... 109 APPENDIX A LIST OF VARIABLES AND SCALAR RESPONSE RANGE ................................... 111 B PERCENT POVERTY BY FIRE DISTRICT ESCAMBIA COUNTY ..................... 113 C 65+ POPULATIONS ESCAMBIA COUNTY ......................................................... 114 D SPECIAL NEEDS REGISTRANTS ESCAMBIA COUNTY .................................. 115 E ZIP CODE COMPARISON ....................................................................................... 116 F HURRICANE PREPARATION QUESTIONNAIRE ................................................. 118
7 LIST OF REFERENCE S ................................................................................................. 124 BIOGRAPHICAL SKETCH .............................................................................................. 133
8 LIST OF TABLES Table page 1 -1 Hurricane strikes 1851 -2006 on the mainland US coastline, and for indivi dual states ...................................................................................................................... 16 3 -1 Frequencies of socioeconomic characteristics for all respondents in study ........ 52 3 -2 Frequencies of hurricane experience factors in study assessing hurricane prepa redness and factors ...................................................................................... 54 4 -1 List of Variables and Scalar Response Range ..................................................... 73 4 -2 Frequency distribution by sample by key demographic and socioeconomic characteristics of respondents ............................................................................... 76 4 -3 Frequencies of hurricane experience factors for all respondents ........................ 78 4 -4 Combined sample correlation coefficients between 11 variables used to assess hurricane preparedness and factors influencing preparedness ............... 81 4 -5 Non vulnerable sample correlation coefficients between 11 variables used to assess hurricane preparedness and factors influencing preparedness ............... 82 4 -6 Vulnerable sample correlation coefficients between 11 variables used to assess hurricane preparedness and factors influencing preparedness ............... 82 4 -7 Results of path analysis in SEM of the hypothesized hurricane preparation decision model........................................................................................................ 90 4 -8 Significance Levels and Unstandardized and Standardized Estimates for the Exercise Model ....................................................................................................... 91 4 -9 Groups statistics and independent samples test for correlation coefficients between 5 variables used to assess hurricane preparedness ............................. 92
9 LIST OF FIGURES Figure page 1 -1 The paths of the 2004 Florida Hurricanes. Source: National Weather Service/National Hurricane Center ........................................................................ 12 1 -2 2005 Category 5 Hurricanes. Source: National Weather Service/National Hurricane Center .................................................................................................... 13 1 -3 Paths followed by 2004 Florida Hurricanes. Source: National Weather Service/National Hurricane Center (Smith & McCarty, 2009) .............................. 15 2 -1 Transactional Model of Stress and Coping (Wenzel et al., 2003) ........................ 35 2 -2 Patons Proposed Social -Cognitive Model (2003) ................................................ 40 2 -3 Proposed hypothesized model. ............................................................................. 44 3 -1 Hypothesized Hurricane Preparation Decision Model .......................................... 46 4 -1 Hypothesized Hurricane Preparation Decision Model .......................................... 85 4 -2 Revised Hurricane Preparation Decision Model ................................................... 90 5 -1 Revised Model. ....................................................................................................... 96
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SOCIALLY VULNERABLE POPULATIONS AND THE HURRICANE PREPARATION DECISION PROCESS By Molly Kay Moon May 2010 Chair: Marilyn E. Swisher Major: Interdisciplinary Ecology Research in disaster has generally focused on whether a household is prepared or not for various types of disaster events, e.g., hurricane, flooding, etc. However, little has been written about why individuals choose to prepare or not prepare. This study fills a ga p in the existing disaster literature by focusing on hurricane preparation intentions by socially vulnerable communities. Grounding the research in social cognitive theories moves the discussion beyond do socially vulnerable communities prepare to one i n which attention is given to why do they choose to prepare or not prepare for hurricanes. This study tests a model developed by Douglas Paton in New Zealand where he examined individual disaster preparedness intention and actions as it related to earthqu akes. The same social cognitive constructs were examined in this research but focused on socially vulnerable populations and hurricane preparation. Using a cross -sectional design, two populations (vulnerable and nonvulnerable) were compared at one p oint in time in terms of differences on the outcome variable of hurricane preparation. Data from 153 households in Escambia County, Florida, were collected using a mail out questionnaire. Correlation analyses and structural equation modeling were used to determine the effect of 11 different variables on the outcome
11 variable of hurricane preparation. Comparisons were made between the combined sample (n=153) and the two subpopulation samples of vulnerable (n = 53) and nonvulnerable (n=100). Results show a much less complex model than Patons. In my model, there are three directional paths to disaster preparation. In the first path individuals go from risk perception directly to individual intentions with the result being hurricane preparations. A second path indicates risk perception leads to perceived control resulting in hurricane preparations. The final path individuals go from risk perception to perceived control to individual intentions through action coping. Due to the small sub -samples, a model comparison between the two groups was not possible. However, correlation analyses indicate tha t the non vulnerable population use s a much more complex decision process in deciding to prep are for hurricanes than the vulnerable population. The vulner able sample appears to react by preparing and not contemplate what would happen if they did not prepare.
12 CHAPTER 1 INTRODUCTION Distant Memories The 2004 hurricane season is slowly becoming a distant memory to many Floridians who experienced one or more hurricanes. During 2004, Florida experienced one of its busiest hurricane seasons ever with four hurricanes occurring between August 13 and Sept ember 25 (Fig. 11) O ne or more of the hurricanes (Charley, Frances, Ivan and Jeanne) affected all 67 Florida counties (Kapucu, 2008). Figure 11: The paths of the 2004 Florida Hurricanes Source: National Weather Service/National Hurricane Center Following on the heels of the 2004 hurricane season, 2005 was the most active on record and produced three hurricanes (Katrina, Rita and Wilma) that reached Category Five (Cutter et al 2006) (Fig. 12) Millions watched the devastating images of New Orl eans and the Gulf Coast struggle through the aftermath of Hurricane Katrina. The inadequate national and local response during the aftermath of Hurricane Katrina
13 resonated across the nation and brought to light the vulnerability that many of our communiti es face (Cutter et al 2006). Figure 12: 2005 Category 5 Hurricanes Source: National Weather Service/National Hurricane Center In the disaster literature, Blaikie, Cannon, Davis and Wisner (1994) argue that vulnerability to natural hazards is not based on the event itself but rather social, economic and political processes that create the different conditions under which individuals have to face hazards. Studies have shown that racial and ethnic populations in the US are more vuln erable to natural hazard events due to language, housing patterns, building construction, community isolation and cultural insensitivities (Fothergill, Maestas and Darlington, 1999, p. 156). Historically, natural events such as floods, earthquakes, and hurricanes have caused millions of deaths. In many cases, it is not the event itself that causes human suffering and deaths, but the follow on effects such as job loss and household displacement that impact socioeconomic processes within the human enviro nment system (Haque and Etkin, 2007). Mileti (1999) argues that we, as humans, must accept
14 responsibility for hazards and disasters. Human beings make the c hoices about where and how development will proceed. For example, development decisions made in N ew Orleans destroyed buffers between the city and the sea resulting in an increased vulnerability to flooding when major storms come inland (Steinberg, 2008). Technology cannot make the world safe from forces of nature. E ffective disaster mitigation can take place only when this is understood Mitigation could involve the modification of the hazard as well as reducing vulnerabilities. Unfortunately, the processes associated with vulnerability (social, economic and political) that determine t he resilienc e of the community are rarely pursued in developing mitigation strategies (Cannon, 1994). Hurricanes and Florida Hurricanes are intense weather phenomena and can cause tremendous amounts of damage either through direct impact on a location or through after effects such as flooding. T he most destructive hurricane seasons in history occurred in 2004 and 2005 and four hurricanes directly impacted Florida in 2004 (Smith and McCarty, 2009). Charley made landfall on the southwest coast and caused extreme damage in Charlotte County. Frances and Ivan struck the southeast coast and caused flooding in Dade County. Ivan made landfall in the panhandle. According to the National Hurricane Center (2005), the storms were responsible fo r at least 47 deaths and $45 billion in
15 damages (Smith and McCarty, 2009; Bl ake, Jarrell, and Rappaport, 2006). Figure 13: Paths followed by 2004 Florida Hurricanes Source: National Weather Service/National Hurricane Center (Smith & McCarty, 2009) T he National Weather Service (NWS) provides the following information for the United States (1) F ourteen out of the fifteen dead liest hurricanes were category three or higher. (2) L arge death tolls were primarily a result of 20 feet or greater storm surg e. (3) I nland floods caused by torrential rains resulted in four of the twenty costliest storms. (4) O ne -third of the deadliest hurricanes were category four or higher. In light of the se facts, Florida is an ideal location to study the phenomena of hurric ane preparation actions. Historically, Florida has experienced more hurricanes than any other state
16 (Table 31 ), 37 major hur ricanes (Category 3 or above) and eight category 4 or 5 hurricanes. Table 1 1 : Hurricane strikes 18512006 on the mainland US coastline, and for individual states, including inland areas if effects were only inland portions of the state, by Saffir/Simpson category (Blake, Rappaport, and Landsea, 2007) The National Weather Service reports that although there has been an increase in overall activity, the number of strong hurri cane landfalls has not increased. During the past 35 years in the United States, only a few category 4 or stronger hurricanes have made landfall. On average, a c ategory 4 or stronger hurricane strike s the United States about once every seven years (Blake et al., 2007). The NWS is quick to note, how ever, that fewer hurricanes do not lessen the threat of disaster. As the NWS noted,
17 Hurricane Andrew the second costliest hurricane in history, occurred in a year with below average hurricane activity (Blake et al., 2007) An analysis of hurricanes and landfall indicate that 40 percent of all US hurricanes and major hurricanes hit Florida and 83 percent of categ ory 4 or higher hurricanes have struck either Florida or Texas. This, of course, can be attributed to the extensive coastlines that eac h state has. Unfortunately, these are the same coastlines that were experienced high population growth until 2008 Als o, low hurricane experience levels are a problem and would likely impact preparation decisions by individuals. Current Florida guidelines Title XVII, Chapter 252 of the 2008 Florida statutes recognized the vulnerability of the state to a wide range of emergencies including natural, technological, and manmade disasters. It is the Legislatures intent to reduce the vulnerability of the people and property of this state (Title XVII, Chap 252.311). This statute established a state emerg ency management agency, the Division of Emergency Management. The statute further requires the development of state and local comprehensive emergency management plans. S pecific elements in this statute are to : (1) establish guidelines for annual exercises for political subdivisions to respond to minor, major and catastrophic disasters; (2) assist political subdivisions in preparing and maintaining emergency management plans; and (3) review periodically political subdivision emergency management plans for consistency with the state comprehensive emergency management plan. The State of Florida Comprehensive Emergency Management Plan 2004 (FL CEMP 2004) outlines specific responsibilities for counties, special districts, state
18 government and federal government. Counties are directed to: (1) coordinate emergency management needs for all municipalities within their counties; (2) implement a broadbased public awareness, education and information program designed to reach all citi zens of the county; and (3) maintain an emergency management program that is designed to avoid, reduce and mitigate the effects of hazards (FL CEMP 2004). The state also has the responsibility to maintain a broad -based public awareness, education and preparedness program designed to reach a majority of the citizens of Florida (FL CEMP 2004, p. 14 ). The guidelines and policies set forth by the state of Florida should serve as impetus for local government and emergency planners to develop comprehensive emergency plans. Local governments are required to submit their emergency management plans for approval (FL CEMP 2004). I review ed emergency management plans from various counties S ome county plans are specif ic with detailed guidelines while others pr ovide only general guidelines. A consistent element in the various plans is that individuals need to be encouraged to take responsibility for their own preparations. The next section explores the literature and research with regard to individual preparat ions. Why and How do Individuals Prepare for D isaster? Individuals living in risk prone ar eas have demonstrated limited knowledge and motivation to prepare for natural hazards (McIvor and Paton, 2007; Johnston, Bebbington, Lai, Houghton, and Paton, 1999; L echliter and Willis, 1996; Rustemli and Karanci, 1999). A national study conducted for the American Red Cross in the immediate aftermath of Hurricane Katr ina and Rita found: 1) P eople were no more prepared after Katrina and Rita than before; 2) few peopl e prepared emergency kits; 3)
19 few had communication plans; 4) preparedness was approached wit h a detached sense of reality; 5) and individuals still lacked knowledge of preparation activities (American Red Cross, 2005). T his was a national study and not f ocused on any specific type of disaster (natural or manmade) However, the researchers isolated re sults from areas that are hurricane prone, such as th e south. Although the results from these areas reflected a slightly higher preparation awareness and a ctivity level than those in other parts of the country the Red Cross study described the Ameri can public as ill prepared (American Red Cross, 2005, p. 3). Kapucu (2008) conducted a study to determine how prepared Central Florida households were for an emergency. His results reflected similar numbers to the earlier 2005 American Red Cross study and he concluded that households in Central Florida (Orange, Osceola and Seminole County) were ill prepared for emergencies (Kapucu, 2008). There are many studies that have examined the level of preparedness of indivi duals and households, but only a few studies have examined why individuals prepare or choose not to prepare for natural hazards. The hazard literature lacks in depth study of socially vulnerable individuals and social cognitive constructs as they relate to individual motivations, individual intention, networks and preparedness ac tions. H azard information does not determine action Rather, how individuals perceive this information in the context of their own life experiences determines action Researchers have argued that risk perception is culturally and socially constructed, and social groups construct different meanings for potentially hazardous situations (McIvor and Paton, 2007, p. 80).
20 State and local leaders continue to focus on dissemination of information as a major form of disaster preparation. It is often assumed that the mere dissemination of disaster preparation information will encourage individuals to take preparedness actions (Paton, 2003; Smith, 1993). However, researchers have found that despite the increase in disaster hazard education and information dissemination, preparation levels remain l ow (Paton, 2003; Ballantyne, Paton, Johnston, Kozuch and Daly, 2000; Duva l and M ulilis, 1999 ). In fact, Ballantyne et al (2000) found that increased public education reduced preparation levels as individuals tended to transfer responsibility from self to others supporting their decision to not prepare. A more thorough examination of why and how individuals prepare for disaster could provide relevant and useful information to state and local leaders as they develop and implement community disaster preparedness and mitigation plans. Research Questions This cur rent study will examine hurricane preparation actions by socially vulnerable populations in Florida A thorough discussion of social vulnerability and disaster is provided in Chapter 2. However, in context of my research, I have decided a priori to use specific indicators (age inc ome and disability) to designate households as vulnerable or non vulnerable. Also, instead of focusing on natural disasters in general, my research focuses on hurricane preparations specifically. I use two theoretical frameworks in my research. The first is the transactional model of stress and coping (Wenzel, Glanz and Lerman, 2002) provides the theoretical basis for Patons social cognitive hazard preparation model (Paton, 2003). I also incorporated constructs from social network theory to ex amine the role of social networks in preparedness decisions. My research will focus on hurricane preparedness
21 actions by disadvantaged individuals and their motivations to prepare or not prepare for hurricanes. I pose four questions. 1 How do socially vulnerable populations perceive risk as it relates to hurricane preparation? 2 Do socially vulnerable populations prepare for hu rricanes? Of those who prepare, what motivates them to do so? 3 What coping resources (if any) do vulnerable populations use? 4 To what e xtent is intention to act linked to preparedness ? Summary This study seeks to fill a gap in the existing disaster literature by focusing on disaster preparation intentions by socially vulnerable communities. Whereas research conducted by Douglas Paton in New Zealand examined individual disaster preparedness intention and actions as it related to earthquakes, this study will focus on the same social cognitive constructs but will focus on socially vulnerable communities and hurricanes in Florida Grounding the research in social cognitive theories moves the discussion beyond do socially vulnerable communities prepare to one in which attention is given to why do they choose to prepare or not prepare for hurricanes. This dissertation is organized around fi ve chapters This introductory section serves as Chapter 1. Chapter 2 provides the literature review and theoretical constructs Chapter 3 describes the research design and method ology employed in this study. Chapter 4 analyzes the information as a product of our primary data collection activities and examines the data in the context of our key research questions. Finally, Chapter 5 presents conclusions, final remarks and direction for future research.
22 CHAPTER 2 LITERATURE REVIEW AN D THEORET ICAL FRAMEWORK This chapter presents background information for this research. The fi rst section reviews th e concepts of disaster and social vulnerability The second section describes the theoretical framework s used in this research. Thorough disc ussi ons of s ocial cognitive theory, transactional model of stress and coping and social network theory are presented. I conclude with a synthesis of the con cepts and theories and my research hypotheses. Disasters and Social Vulnerability What is Disaster? Numerous scholars have tried to answer the questionwhat is disaster? Quarantelli (1998) challenged several disaster researchers to define this concept A review of the responses Quarantelli received indi cates the answer is linked to the academic background of the scholars. Cutter (2005) argues that risk, hazards and disaster research communities do not understand each others science S he questions how social scie nce perspectives can be strengthened if researchers are unaware of the totality of the s oc ial science perspectives. O ne truism that is evident throughout the literature is that no one definition applies in every situation. Quarantelli (1987) argues there is little hope in trying to devise a useful definition that is universally accepted (Pe rry, 2006). As a result there is some concern about the intellectual health of the field (Quarantelli, 1985, 1995; Oliver Smith, 1999). However, Olive r-Smith (1999) argues that disaster research does not necessarily have to have definitional consensus There are many academic fields that encounter the same issue. A
23 definitional debate is fruitful because it encourages exploration of new dimensions of disaster (Oliver -Smith, 1999). Q uarantelli (2005) believes that the audience wh o will be exploring t he concept is key to the definition of disaster. Sociologists focus on sociological context and tradition, attending in particular to delimiting the phenomenon to become a focus for the processes of social science (P erry, 2006, p. 6). Research about the impact of bombing in Europe and Japan during World War II focused on social disruption caused by the event rather than the event itself. Three interrelated definitions emerged: (1) pattern of interrupted stability; (2) adaptation to the interruption; and (3) resumption of behavior in a stable period (Perry, 2006). T he system/d escriptive approach to disaster is s imilar to the interrelated definitions provided by Perry Porf iriev (1998) says a disaster is an event that destabilizes and disrupts the co nnections of a social system, results in destruction, and overloads physical and psychological systems, making it necessary for extraordinary response measures. Gilber t (1998) approach ed Quarantellis question theoretically and outlined three main p aradigms: (1) disaster as a duplication of war in that an external agent disrupts the social system; (2) disaster as vulnerabilities in the social order; and (3) disaster as uncertainty disruption of systems in society. Disaster as duplication of war exp lores the linkages between external ag ents (arms and enemies) and communities are through conflict (Gilbert, 1998). The exploration of this concept emerged in the United States during the Cold War when United States government institutions funded research exploring reactions of people to possible air raids. The more contemporary paradigm eliminates t he notion of agent and ex amines disaster linkages to local communities
24 through social vulnerabilities (Gilbert, 1998). This ch ange in paradigm shows how di saster research is moving from studying disaster as an effect to disaster as a result of the underlying logic of the community (Gilbert, 1998, p. 14). Gilberts (1998) final paradigm of disaster as uncert ainty considers three points: U ncertainty result s from a lack of defined causes and effects of disasters; modern communities are growing in complexit y that may increase uncertainty; and modern communities are finding it difficult to define disaster situations. Dombrowsky (1998) argues that defining the social processes of disaster is different than defining disaster. He uses the example of how the German Red Cross defines disaster as an extraordinary situation in which the everyday lives of people are suddenly interrupted and thus protection, nutrition, clothing, housing, medical and social aid or other vital necessities are requested (Katastrophen-Vorschrift, 1988, p. 2 as cited by Dombrowsky, 1998). Similar to this, the United States has also established a set of constructs that formally define di saster. Codified in the Robert T. Stafford Disaster Relief and Emergency Assistance Act, a major disaster : means any natural catastrophe (including hurricane, tornado, storm, high water, wind driven water, tidal wave, tsunami, earthquake, volcanic erupt ion, landslide, mudslide, snowstorm, or drought), or, regardless of cause, any fire, flood, or explosion, in any part of the United States, which in the determination of the President causes damage of sufficient severity and magnitude to warrant major disa ster assistance under this chapter to supplement the efforts and available re sources of States, local govern ments, and disaster relief organizations in alleviating the damage, loss, hardship or suffering caused thereby (FEMA 2003). The hazards -disaster tr adition comes from geographers and geophysical scientists. This perspective focused on the hazard earthquakes, tornadoes, floods and so forth and understanding the hazard (Perry, 2006). Quarantelli (2005) argues that focus on the event itself may lim it t he analysis of disasters because there are some
25 disasters such as famine or computer system failures that have no identifiable originating agent. Carr (1932) su ggests that the collapse of cultural protections constitutes a disaster. Not every event is catastrophic. If a hurricane hits and the community survives intact, there is no disaster (Carr, 1932). In the broad context, Mileti (1999) emphasizes that disasters flow from overlaps of the physical, built, and social environments but that they are s ocial in nature. In defining and studying disasters, one should look first at social systems, since they (not the agent) are the real source of vulnerability (Quarantelli, 2005). The issue is not disaster as events but instead human vulnerability (and resilience) to environmental threats and extreme events (Cutter, 2005, p. 39). Based on the idea that disaster is a social phenomenon, Quarantelli (2005) emphasizes that disasters are based on vulnerabilities caused by changes in the social structure or system and not on the event itself. Oliver Smith (1998) also espouses a social defi nition of disaster, placing disruption and vulnerability within the social structure. Oliver -Smith (1998) also takes an ecological perspective of disaster. He suggests the need for an understanding of the relationship between society and environment. The focus is on how society adapts to the total environment to include the natural, modified and constructed contexts of which the community is a part (Oliver Smith, 1998). Oliver -Smith (1998) defines disaster as: a process/event involving the combination of potentially destructive agent(s) from the natural, modified and/or constructed environment and a population in a socially and economically produced condition of vulnerability, resulting in a perceived disruption of the customary relative satisfactions of individual and social needs for physical survival, social order and meaning (p. 186).
26 In line with the social construction of disaster, Dynes (1998) defines di saster as events when norms fail and communities utilize extraordinary efforts to protect their social resources. Of note, Dynes (1998) uses community as the social unit that provides an initial conceptualization of disaster. His reasoning is that the community is a social unit that has cross -national and cross -cultural applicability, and it has the capacity and resources to respond to a disaster. He argues that the community is a multi organizational system, and, as such, the location of social action i s the community (Dynes, 1998). Dynes (1998) synthesizes his concept of disaster as being a normatively defined occasion in a community when extraordinary efforts are taken to protect and benefit some social resource whose existence is perceived as threat ened (p. 113). Kreps (1985) describes the social constructs of disaster as alternative structural forms that occur before, during and after events. He argues that the basic sociological constructs such as collective behavior, formal organizing, and soci al networking can be used to describe disasters as social constructions (Kreps, 1998). Quarantelli (2005, p. 339) argues there are two fundamental ideas in defining disasters: (1) disasters are inherently social phenomena. It isnt the hurricane or stor m surge those are the source of damage; (2) disaster is rooted in the social structure and reflects the processes of social change. Rosenthal (1998) compares the traditional world of disasters to the contemporary world of disasters. He argues that t he traditional world of disasters is a world of unness. Disaster characteristics are negative. Disasters are unexpected, unprecedented and unmanaged phenomena d erived from natural processes or events
27 that are uncertain. Human victims are unaware and unready. Simply put, disasters in North America are unscheduled events (Hewitt, 1998). The contemporary world of disasters involves linkages, chains and processes. Compared to the traditional view, the dominant view about contemporary disasters is that disasters have consequences. It is the disaster after the disaster that provides a new dimension to defining disaster (Rosenthal, 1998). Rosenthal (1998) indicates the process oriented approach is defined by the interrelationship between characterist ics, conditions and consequences of disasters. Social Vulnerability The one common theme in all of the definitions and models of disaster has been the focus on vulnerable groups in society. The social science community agrees that many factors influence social vulnerability. Vulnerable communities have been identified based on nominal variables such as gender, race, or socioeconomic status that stratify wealth, power and status (Mileti, 1999). Class, caste, and immigration status have also been included as factors (Bolin, 2006). Special needs categories such as physically or mentally challenged, homeless, transients, non-English speaking immigrants and seasonal tourists have also been included in the definition (Cutter, Brouff, and Shirley, 2003). Some researchers have expanded the definition to include factors that are not as easily defined. Cutter et al., (2003) go further and include lack of access to resources, limited political power and representation, social capital, beliefs and customs, age o f built environment, and infrastructure and lifelines. There is disagreement, however, about the specific variables used to represent these concepts (Cutter et al 2003). Cutter (2005) submits that disaster studies have spent too much time defining th e phenomena under study rather than researching the
28 vulnerabilities (and resiliency) to environmental threats and extreme events. The linkages and interdependencies between human systems, natural (or environmental) and technological systems and the built environment define vulnerability (Cutter, 2005). While each can be studied independently, it is the interaction that is important in understanding vulnerability. Cutter (2005) states, the whole (vulnerability) is greater than the sum of its parts (huma n systems, the built environment, technological systems, natural systems) (p. 40). Cannon (1994) argues that vulnerability is a complex set of characteristics based on an individuals place in society and it is a combination of these factors that cause ha zards to have varying degrees of impact. Wisner et al. (2004, p 11) defines vulnerability as the characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist, and recover from the impact of a natural hazard (Italics in original). Researchers are moving past the nominal variables as measures of vulnerability and are researching vulnerability as a process.1 Oliver -Smith (1996) indicates that current research emphasizes political and economic inequalities as well as processes of racial and ethnic marginalization. Social vulnerability should be understood in terms of social, economic, political and cultural processes that make the society vulnerable (Haque and Etkin, 2007). Cannon (1994) states the processes a re not necessarily those that may generate differences in wealth, resources or power. The vulnerability concept is a means of translating known everyday processes for the economic and political separation of people into a more spe cific identification of those who may be at 1 Nominal variables refer to age, race, ethnicity, gender, disability, housing type, etc.
29 risk in hazardous environments (Cannon, 1994, p. 17). Bolin (2006) argues that factors in social processes such as race, class and ethnicity affect a communitys legal and political rights and access to resour ces and livelihoods. Further, a process model of vulnerability proposed by Wisner et al (2004) linked three elements, root causes, dynamic pressures and unsafe conditions. Haque and Etkin (2007) suggest there is a link between social and political vulnera bility and the extent to which a community treats hazards through prevention, mitigation, preparedness, response and recovery. Some scholars have made a distinction between passive and active vulnerability. Passive vulnerability is based on the amoun t of resources available to community members and active vulnerability is based on the ability of community members to change their situation (Bradshaw, 2004). Viewing vulnerability through the social capital lens, Murphy (2007) argues that vulnerability may be increased or decreased due to placebased or interest/kinship based communities. Ethnographic data from Hurricane Andrew informed a discussion of race, class, gender and poverty. Data showed that politi cal and economic processes created vulnerable groups especially among Caribbean immigrant and African American communities (Bolin, 2006). The Chicago heat wave of 1995 serves as an example of social vulnerability F atalities increased because community residents were elderly T heir fear of cri me influenced their decision about responding to community workers who came to help them and they would not leave their apartments to go to shelters or cooler areas. In researching Hurricane Mitch and its impact on Central American republics, Comfort et a l (1999, p.41) found that many of the residents were aware of
30 their vulnerability to natural hazards but none were aware of the degree to which cumulative economic and environmental changes had set the stage for a major disaster. Gilbert (1998) sugges ts that social and political vulnerability is based on a communitys lack of social and political boundaries (p. 15). Vulnerability can be defined as a type of political ecology in that disasters center on the relationship between the human populatio n and socially generated and politically enforced productive and allocative patterns, and its physical environment (Oli ver Smith, 1998 in Tierney, Lindell, and Perry 2001, p 21). In At Risk (1994) Blaikie and his co authors develop a framework that c haracterizes disasters as involving the convergence of socially produced vulnerability and exposure to hazards. Vulnerability to disasters results ultimately from political, economic, and ideological/cultural processes that put individuals and groups at r isk by institutions that fail to provide adequate protection. C ontribu tions over the past 20 years have examined the importance of socioeconomic factors in household preparedness decisions. All things being equal, households with higher socioeconomic st atus are better prepared for disasters than financ ially less well off households and ethnic minorities show a low propensity to engage in emergency pr eparedness activities (Tierney et al 2001). Smith (1996) implies that vulnerability is based on a measure of risk combined with the social and economic abilities of a community to respond to a hazard event. Further, resilience and reliability are factors in determining vulnerability. Resilience refers to a communitys capacity to absorb and recover from a hazard event and r eliability describes how effective a comm unitys preparation efforts are (Smith, 1996).
31 Tierney et al (2001) state that hazard vulnerability, along with the inability to prepare or respond, may be related to social and economic inequali ty. However, Hewitt (1998) argues that the vulnerability paradigm involves the idea of voluntary and involuntary risks. Choices are made by groups or persons to share some or all of the risk (Hewitt, 1998). Theoretical Framework This section presents the theories that were used in the development of the hypothesized research mod el. Specifically, this section explores three theories: social cognitive, transactional model of stress and coping and social network. The section further discusses Patons (2003) proposed social -cognitive preparation mod el and explains how the model was used to develop the hypothesized research model employed in this study Social Cognitive Theory Social cognitive theory (SCT) addresses many of the psychosocial construct s that explain why people make the choices they do. The theory date s back to 1962 when it was develop ed to understand social learning and personali ty development (Baranowski Perry, and Parcel, 2002). The theory has evolved over time and now includes co nstructs that explain the humandecision making process. Social cognitive theorys premise is that h uman consciousness is the critical factor of mental life and is what makes life manageable. This involves accessing information, processing information, and constructing and determining courses of action (Bandura, 2001). Social cognitive theory is used extensively in the development of health behavior modification programs. The same approach is applicable to di saster preparedness research because the SCT used in health research examins the same decisionmaking
32 processes used by individuals be they health behavior decisions or disaster preparedness decisions. Therefore d isaster researchers have used the same concepts found in the original SCT. However, they have included additional variables such as problem -focused coping, self efficacy, and sense of community to construct a more comprehensive social -cognitive model of natural hazard preparedness (Paton, 2003). Since its conception in the 1960s by Al bert Bandura, SCT has evolved from a theory of pre determined human behavior to one that recogni zes that the individual has control over his/her own life (Baranowski et al., 2002). Bandura (2001) further describes social cognitive theory as a model of int eractive agency where people are more than onlookers within the environment. People are agents of experiences. The sensory, motor, and cerebral systems are tools people use to accomplish the tasks and goals that give meaning, direction and satisfaction to their lives (Bandura, 2001, p 4). Human agency is comprised of core features such as intentionality, forethought, self reactiveness and self -reflectiveness concepts that are explained more fully below These concepts directly relate to disaster prep arednes s. Human agency is a critical factor in an individuals disaster preparedness decisionmaking process. According to Bandura (2001), intention is a representation of future courses of action. It is not merely the expectation that something will happen. It is a commitment to brin g the intention to action. Intention is a representation of an action that is to be performed. Intentions and actions are separated by time. Therefore, intentions may be viewed as motivators that affect future actions (Bandura, 2001). Similar ly, Paton (2003) describes the link between motivators and preparation as intentions. He uses social cognitive concepts of outcome expectancy, self efficacy, problem -focused coping
33 and response efficacy as the i ntention formation phase in his model. Intentions are linked to preparation through a number of concepts (see Figure 2-2) (Paton, 2003). Forethought is another extension of the decision -making process. Forethought is what motivates individuals and will guide their actio ns in regards to future events. As Bandura (2001) explains, futur e events do not exist and therefore cannot be causes of current motivation. However, if future events are represented cognitively, they can be transformed into motivators (Bandura, 2001). As Bandura (1986) states, P eople construct outcome expectations from observed conditional relations between environmental events in the world around them, and the outcomes given actions produce (p. 7). Banduras (2001) self -react iveness concept links i ntentions to disaster preparations in Patons (2003) model. Not only must individuals make decisions to prepare and take courses of action, they must have the ability to act on thos e decisions. S elf -reflectiveness allows individuals to evaluate their mo t ivations and actions in regard to their decisions. Band ura (2003) states that the factors that guide and motivate individuals are rooted in the individuals belief that he/she has the power to produce effects by their actions. Efficacy plays a central r ole in human agency. People choose what they want to pursue, decide how much effort they want to put into their actions and decide how long to continue with their actions (Bandura, 2003). In summary, the SCT model provides a framework to examine the decision making processes individuals use in decidi ng to prepare or not for a natural hazard. Do individuals rely on their beliefs and experiences when making disaster preparation
34 decisions? Do individuals consider their abilities and capacity to take actions prior to making preparation decisions? Transactional Model of Stress and Coping Contemporary research on stress and copin g has evolved f rom a concept of de fense and unconscious processes. C oping is now looked at through a cogn itive behavioral lens (Folkman and Moskowitz, 2004). Lazarus (1966) seminal work emphasized cognitive and behavioral responses to how people manage problems and stress found in everyday life. In the 1970s Folkman and Lazarus (Folkman and Lazarus 1980; Lazarus and Folkman 1984) further defined coping as thoughts and behaviors that people use to manage the internal and external demands of situations that are appraised as stressful (Folkman and Moskowitz, 2004, p. 746). Numerous measures have been created to analyze coping behaviors A s a result many studies have relied on the prevailing cognitive approach. An examination of numerous studies indicates that stress and coping are social c ognitive constructs that applicable to numerous situations. In fact, Lazarus and Folkman (1984) argue t hat coping is a process directly linked to a situation or condition and the i ndividuals resources (Folkman and Moskowitz 2004). Lazarus and his colleagues developed a theory of psychological stress and coping. The Transactional Model of Stress and Coping is a framework for examining the process of coping with stressful events (Wenzel et al 2002) (Fig. 21) The theory identifies cognitive appraisal and coping as factors that influence personenvironment relations and their near term and long-range outcomes (F olkman et al, 1986). The process starts with an initial appraisal of events or stressors that could caus e harm or loss. Folkman and Moskowitz (2004) indicate that the emotions associated
35 with this initial appraisal are oft en negative and intense. The negative, intense emotions may often regulate or interfere with forms of coping. Thus, reappraisal is conducted as one works through control over outcomes and emotions (Folkman and Moskowitz, 2004). The twoappraisal process is illustrated in the model provided by Wenzel et al (2002) (Fig. 21 ). Figure 2 1 : Transactional Model of Stress and Coping (Wenzel et al 2003) Stressors are defined as internal or external environmental demands that may impact physical or psychological well being (Wenzel et al 2 002). As explained by Wenzel et al (2002), when an individual is faced with a stressful situation, the individual evaluates the threat (primary appraisal) as well as their ability to cope with the situation (secondary appraisal). Coping strategies mediate primary and secondary appraisals. Lazarus and Folkman (1984) define coping as a process that focuses on what an individual thinks or does in response to stressful situations. Additionally, they argue that coping is Mediating Processes Moderators Primary Appraisal Perceived responsibility Perceived severity Motivational relevance Causal Focus Stressors Outcomes Dispositional coping style Social support Secondary Appraisal Perceived control over outcomes Perceived control over emotions Self efficacy Coping Effort Problem management Emotional regulations Adaptation Emotional wellbeing Functional status Health behaviors Meaning Based Coping Positive reappraisal Revised goals Spiritual beliefs Positive events
36 contextual be cause it is influenced by a persons appraisal of the situation and available resources for m anaging the situation (Lazarus and Folkman, 1984). Wenzel et al (2002) describe problem man agement strategies for coping as active coping, problem solving and information seeking. In contrast, emoti on-focused coping is directed at changing an individuals feelings or attitude about the stressful situation (Wenzel et al 2002). Generally, the model predicts that problem -focused strategies are better when the str essor is changeable and that emotion-focused strategies may be used when the stressor is unchangeable or all problem -focused strategies have been exhausted (Wenzel et al 2002). Most coping research has focused on how people cope with events that have ha ppened in the past or are occurring A new development in coping research is what is referred to as futureoriented proactive coping (Folkman and Moskowitz, 2004). Aspin wall and Taylor (1997) state, P roactive coping consists of efforts undertaken in adv ance of a potentially stressful event to prevent it or modify its form before it occurs (p. 417). They argue that it can be differentiated from coping and anticipatory coping in three ways (Aspinwall and Taylor, 1997). First, it involves acquiring resources to prepare in general --not prepare for a specific stressor. Second, proactive coping requires different skills such as the skill to identify a stressor before it occurs. Third, skills that may be succ essful for proactive coping may not be adequate f or coping or anticipatory coping. Aspinwall and Taylor (1997) examined ways in which people cope in advance to negate or alleviate the stressors of future events such as a pending lay off or results of a medical test. There are five components to the proactive coping process: (1) B uilding
37 resources that can be used to offset projected losses; (2) rec ognition of potential stressors; (3) initial appraisal of stressors; (4) pr eliminary coping efforts; and (5) use of feedback about ones efforts (Aspinwall, 2003). Folkman and Moskowitz (2004) argue that futureoriented or proactive coping deserves additional attention. However, they also state that we need measures that can tap into futureori ented coping methods so that we can examine how individuals cope with adverse impacts of future events. In summary, stress and coping can be researched and examined in many different ways. However, approaches to understanding stress and coping treat coping as a complex, multidimensional process that is sensitive to t he environment and its demands (Folkman and Moskowitz, 2004). Social Network Theory Social network theory is the study of relationships between individuals and formal or informal groups and how those relationships affect behavi or. For example, an individ uals relationships with a formal organization such as a church or community organization, may influence how he/she copes with a stressful event. A n informal network of family and friends might also influence the individual s coping Granovetter (1973, 1983) articulates the differences between strong and weak ties in his examin ation of network theory. He argues that individual behavior and action d epend on whether relationships are based on strong ties or weak ties (Granovetter, 1973, 1983). In his an alysis, Granovetter (1973) refers to the strength of weak or strong ties. Strength is defined as a linear model combining the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the t ie (Granovetter, 1973, p. 1361). T he more one interacts with someone, the stronger the tie will be (Homans, 1950; Granovetter, 1973). It is also
38 hypothesized that the more homogenous the individuals, the stronger the tie will be. Therefore, family and close friends may be considered strong ties. Weak ties, on the other hand, are those that may be nothing more than an acknowledgement of an individual such as an acquaintance at work. Granovetter (1973, 1983) argues that weak ties can have a tremen dous impact on individual behavior because there are a large number of them. Research has shown that the fewest people were reached through strong ties (family and friends) whereas the most people were reached through weak ties (acquaintances) (Rapoport and Horvath, 1961; Granovetter, 1973). The weak tie network is generally presumed to be larger than the strong tie network thus providing additional resources. How would weak or strong networks influence disaster preparedness decisions? Weak ties prov ide people with access to information and resources beyond those available in their own social circle; but strong ties have greater motivation to be of assistance and are typically more easily available (Granovetter, 1983, p. 209). Let us now consider ho w strong ties and weak ties may affect a vulnerable community. Peter Blau (1974) argues that class structure lends itself to creating groups of individuals based on homogeneity. That is, i ndividuals will choose to interact with like individuals. Blau (1974) argues that the l ower ones class status the greater likelihood they depend on strong t ies. In a study conducted in Philadelphia, Ericksen and Yancey (1977) concluded that the structure of modern society is such that some people typically find it advantageous to maintain strong networks and we have shown that these people are more likely to be young, les s well educated, and black (p. 23). They found strong networks to be linked to economic insecurity and lack of social services
39 (Ericksen and Yancey 1977). Other studies have demonstrated the same point (see Stack, 1974 and Lomnitz, 1977). Economic pressures present particular challenges to a vu lnerable population. Strong ties are evident because the vulnerable population depends on the reciprocit y of close friends and families for support during challenging times (Granovetter, 1983). Weak ties are an important resource. As Granovetter (1973) explains, the increased number of weak ties allows for more information dissemination. Vulnerable commu nities may have to depend on a large, weak tie d network to ensure they receive the necessary information to make disaster preparation decisions. As mentioned above, the strong tie network is based on homogeneity. Therefore, the information flow may not b e as extensive. A larger weak tie network may result in increased resources and information available for disaster preparation decisions. A Proposed Social -Cognitive Preparation Model Douglas Paton of the University of Tasmania in Australia recognized the need to understand the reasoning and judgment that underpin decisions regarding disaster preparedness (Paton, 2003, p. 210). Using a psychological perspective, he has developed and proposed a disaster preparedness social cognitive preparation model (Pa ton, 2003) (Fig. 2 -2 ) Patons (2003) model draws from social cognitive theory and stress and coping theory.
40 Figure 2 2 : Patons Proposed Social -Cognitive Model (2003) S everal models of protective behavior from the research about health behaviors influenced Paton (2003). His model comprises three phases (see Figure 22) The first phase contains the precursors or what motivates individuals. The second phase links the initial motivation to intention formation. The third phase links intentions to actual preparations (Paton, 2003). The health protective behavior literature indicates that risk perception is a valid precursor variable. Paton (2003) argues that other attributes may also influence intention. Thus, the variables of criti cal aw areness and hazard anxiety are included in his framework (Paton, 2003). Critic al awareness is the extent to which people discuss and think about a particular hazard. O nly when an individual perceive s a hazard as critical will he/she be motivated to take action (Paton, 2003). Paton (2003) further Critical Awareness of Hazards Risk Perception Hazard Anxiety Outcome Expectancy Self Efficacy Response Efficacy INTENTIONS ADJUSTMENT ADOPTION/ PREPARATION Motivators or precursors Intention formation Linking intentions and preparedness Problem focused coping Sense of community Normative Factors Trust Empowerment Perceived responsibility Timing of hazard activity Response efficacy
41 states that a cer tain amount of anxiety is associated with a hazard because natural hazards are unpredictable and may cause extensive destruction. Consistent with social -cognitive approach es, Patons (2003) model shows that individuals will decide if their actions can, in fact, mitigate th e hazard once they are motivated to think about a threat Thus, linking outcome expectancy to intentions is a function of self efficacy (Paton, 2003). R esponse effi cacy describes perceptions as they relate to availability of resources. Problem -focused coping might encourage individuals to confront the problem However, if they do not believe they have the resources to cope they may not act on the probl em (Mulilis and Duval, 1995; Paton, 2003). Finally, Paton (2003) includes additional variables that may influenc e the link between intentions and preparations. V ariables like sense of community and perceived responsibility are moderating factors that may contribute to or affect an individuals decision to prepare or not prepare. Research Concepts and Hypothesized Model Disaster Although researchers tend to define disaster in many different w ays, there is a common theme in the ir discussions. R esearchers agree that vulnerability to the specific type of event and resilience to the event are factors that may define an event as a disaster whether it is a natural or manmade disaster (Cutter, 2005; Rosenthal, 1998; Gilbert, 1998). Vulnerability and resiliency are not the same for every type of disaster, but most studies focus on one type of event such as an earthquake, wildfire, hurricane, or terrorism
42 M y research examines hurricane preparation decisions. Hurricanes are one of several natural l y occurring weather phenomena in Florida. Hur ricanes come with warnings and warrant specific preparation actions. I chose to examine hurricane preparation actions for this reason. Individuals make decisions about the extent of their preparation actions and I intend to examine the decision processes that they use Social Vulnerability The discussion about social vulnerability has focused on examining political, economic and cultural processes that create socially vulnerable populations (Tierney et al., 2 001; Blaikie, 1994; Ol iver -Smith, 1996). Although many researchers discuss vulnerability as a process, the ir discussion s often lead back to demographic and socioeconomic factors as precursor s to vulnerability. The challenge is to determine the specific v ariables that represent these processes. Measuring social vulnerability as a proces s is difficult. T herefore, researchers tend to categorize vulnerability based on specific characteristics (Cutter et al., 2003). Cutter et al. (2003) attempted to measure social vulnerability by creating a social vulnerability index. The index uses 11 independent factors derived from demographic, socioeconomic, and built environment information for an area. Cutter et al.s (2003) index has been used to identify areas (st ate, county, census block) as vulnerable. However, I have not found any research where it has been used to measure vulnerability at the household level. My research examines the household decision process as it relates to hurricane preparations. Theref ore, I chose to use the characteristics of age, income and disability to define a vulnerable or non vulnerable household. These are characteristics commonly used in disaster studies to define vulnerability. Additionally, Florida statutes
43 identify the eld erly and disabled as populations requiring specific preparation actions in regards to emergency management (FL CEMP 2004). A more extensive discussion of age, disability and income is provided in Chapter 3. Hypothesized Model I use a hybrid of Patons proposed social -cognitive preparation model is being adapted to examine my own hypotheses Similar to Patons model, the proposed model includes social cognitive variables as well as variables drawn from the transactional model of s tress and coping. The m odel examines similar moderators to those proposed by Pa ton (2003). Paton used his model to study earthquake preparations, and I will test Patons model with regard to hurricane preparations. A primary departure from Patons model is the inclusion of soci al networks as a possible link between intentions and preparation in the hypothesized model. The goal of this research is not to analyze the formal and informal networks that individuals may be part of but the goal is to determine whether formal or infor mal networks affect preparation decision processes. The addition of strong and weak social networks as mediating variables in Patons model may improve the explanatory power of the model. Finally, Patons researc h was conducted with g eneral populations in different communities and did not make a distinction among various subpopulations, e.g., vulnerable or non vulnerable. It has been argued that there is a difference in preparation levels and preparation decisions based on vulnerability. My research w ill test the model on the two subpopulations of vulnerable or nonvulnerable households to determine if, in fact, there is a difference in their decision making process and hurricane preparations
44 Figure 2-3: Proposed hypothesized model. Therefore, my research will examine the hurricane preparation decision process by testing Patons proposed social -cognitive model on vulnerable and nonvulnerable populations. T he fol lowing hypotheses are presented : H1: The causal pathways found in Patons social -cognitive model for earthquake preparat ion decisions will be the same for the hurricane preparation decision process H2: Strong and weak social networks added as a link between intentions and hurricane preparations increase the explanatory power of the model. H3: Vulnerable and non vulnerable populations differ in their hurricane preparation decision process Risk Perception Hazard Awareness Motivational Relevance Perceived Control over Outcomes Self Efficacy Action Coping Individual Intentions Strong Network Weak Network Hurricane Preparation Stressors Coping Self Assessment Intention to Preparation Link Moderator Response Efficacy Past Experience Personal Responsibility
45 CHAPTER 3 METHODOLOGY This chapter presents the methods and scientific reasoning used in this research. This study measures attitudes toward disaster preparation. Included is a discussion of the conceptual framework linking variables, units of analysis, site selection, sampling framework, data collection, and the overall research design used in this study. Conceptual Framework This study focuses on factors that influence an ind ividual households decision to prepare or not prepare for hurricanes. Using Patons (2003) proposed social -cognitive preparation model and other disaster preparation research, a conceptual framework was developed to assess relationships am ong v ariables. I used variables similar to Paton et al.s (2003) model (Fig. 2 2) because predictive validity had been established in his research. The conceptual framework for this research posits that social cognitive factors link stressors to disaster preparation through coping and intention actions (Fig. 3 -1)
46 Figure 3 1 Hypothesized Hurricane Preparation Decision Model Research Design A cross-sectional design was used for this research. T wo populations at one point in time were compared in terms of differences on the outcome variable (hurricane preparation). Analysis relied on the differences in the samples at that point in time. Unlike other research designs, a cross -sectional design has no time dimension, no intervention and groups a re based on existing differences between groups (DeVaus, 2001). Cross -sect ional studies are widely used because they can be ac complished quickly and are cost effective. However, t he cross-sectional design is weak in showing direct causality. Lindell and Perry (2000) reviewed 23 different cross sectional studies that looked at earthquake hazard perceptions and earthquake hazard adjustments. They found the correlations in the various studies made it difficult to determine if hazard Risk Perception Hazard Awareness Motivational Relevance Perceived Control over Outcomes Self Efficacy Action Coping Individual Intentions Strong Network Weak Network Hurricane Preparation Stressors Coping Self Assessment Intention to Preparation Link Moderator Response Efficacy Past Experience Personal Responsibility
47 perception caused haz ard adjustment or if hazard adjustment caused hazard perception. Due to the weakness in establishing direct causality, Lindell and Perry (2000) argue that cross sectional studies of factors associated with the adoption of seismic adjustments would be wel l advised to supplement reports of current adjustment adoption with the collection of either retrospective reports of past adoption or behavioral intentions for future adoption (p. 498). They believe that although a quasi -longitudinal design, or repeat ed points in time cross -sectional design, is inferior to a true longitudinal design, it is better than a single point in time cross -sectional design (Lindell and Perry, 2000). Despite this limitation, the cross -s ectional design is appropriate for underst anding relationships between a number of variables. It allows the researcher to include many more variables than the experimental group of designs, enhancing explanatory power. Further, the design can be greatly strengthened when multiple comparison groups are used, the case here, and when the selection of variables for the study are theoretically based, which is also the case here. Marsh (1982) argues that by using models and theories as guides, the researcher is able to draw causal conclusions. Based on previous th eories and research, variables that have been shown to influence an individuals decision making process in regards to earthquake preparations (Paton et al., (2003) were selected and will be applied to the hurricane preparation decision proces s in my research. The objective is to understand how these variables interact in two populations, the socially vulnerable and th e nonvulnerable. T he conceptual model presented in the previous section was developed based on previous research and was used to guide decisions about which data to select. Furthermore, cross -sectional
48 models are very effective in demonstrating where causal relationships do not exist. Eliminating relationships is as important as showing causal relationships (DeVaus, 2001). Cr oss -sectional designs require statistically representative samples to enhance external validity. External validity refers to the degree to which the conclusions reached, based on the evidence from one study, can be generalized to other conditions. That m ay be the theoretical population but also commonly includes other t imes, places, or phenomena. This research compared two samples which should represent two theoretical populations in Escambia County, Florida and their individual household decision proces s in making hurricane preparations. To enhance external validity, it was necessary to obtain a statistically representative sample from each group. The question may be asked if the specific results of a locale -focused study can be general ized to the wide r population. I dont believe external validity can be ensured. However, the intent of this study is to examine the hypothesized model. I am more concerned with the theoretical generalizations associated with the constructs in the model which are derived from social cognitive theory, stress and coping theory and social network theory. Unit of Analysis Individual households are the unit of analysis for this research because hurricane disaster preparation generally focuses on household preparation e.g., disaster kits that include enough supplies for the household, an evacuation plan for the household, etc. The US Census Bureau defines a household a unit that consists of all the people who occupy a housing unitincludes the related family member s and all the unrelated people, if any, such as lodgers, foster children, wards, or employees who share the unit. A person
49 living alone in a housing unit, or a group of unrelated people sharing a housing unit such as partners or roomers, is also counted a s a household. (US Census Bureau, n.d.) I made t he assumption that one individual in the household generally makes hurricane preparation decisions for everyone that resides in that residence. Therefore, the questionnaire asked that the person responsible for making hurricane decisions for the household complete the questionnaire. Sample and Site Selection This study will examine preparation decisions by socially vulnerable populations, the theoret ical population of interest. The Federal Emergency Management Agency (FEMA) defines socially vulnerable or disadvantaged populations as individuals or households that are more likely to suffer from a hazard because of social or economic marginalization (i.e., minority, low -income, nonEnglish speaking, etc.). T hese individuals may be disadvantaged by a lack of resources, services and/or capability to take care of themselves (FEMA EPD Project, 2009). Additionally, the theoretical population is defined as at -risk, meaning the household is physically located in a n area th at is vulnerable to h azards like hurricanes, flooding, etc (FEMA EPD Project, 2009). Escambia County, Florida, was selected as the research site for this study. It is the westernmost county in the stat e and shares a north and west border with A labama. This county is vulnerable because it is located directly on the Gulf of Mexico. Escambia County experienced numerous hurricanes in the past (NWS, nd) The population is diverse in relation to socioeconomic factors (US Census Bureau, 2000) Wealthier residents tend to live close to the coast, while disadvantaged residents reside throughout the county. Of particular interest, the county has mapped vulnerable populations and residents for disaster planning purposes since Hurricane Ivan struck in
50 2004. I obtained copies of these maps from the Escambia County Public Safety Office to locate areas where low -income, elderly and registered special needs residents are concentrated (Appendix 1, 2 and 3). The ability to identify a disadvantaged population is essential to ensuring the appropriate households are being sampled for this study. The hurricane preparation decision process is the primary focus of this research. Participants in the research should have been exposed to hurricane preparat ion information and programs and have had the opportunity to decide whether or not they should prepare for hurricanes and to what extent. Escambia County initiated a robust emergency preparedness program after Hurricane Ivan in 2004. They created the Be Ready Alliance Coordinating for Emergencies organization, an umbrella agency that provides hurricane preparation information to all c ounty residents. The BRACE organization has identified vulnerable residents and BRACE partners extensively with support an d service organizations to ensure information reaches the vulnerable populat ion (BRACE, 2009). I t is probable that residents in Escambia County have adequate knowledge level of hurricane preparation actions and have decided to prepare or not prepare based on that knowledge. I want to determine whether disadvantaged populations differ in their decision making process from other populations, not the effect of knowledge on the decisionmaking process. Therefore, it is important that I conduct the research i n a place where all residents, including vulnerable residents, receive similar information. Escambia County provides these conditions. This study measures the extent to which certain social -cognitive factors influence hurricane preparation decisions for socially vulnerable and non vulnerable populations. I am particularly interested in understanding if there are significant differences in the
51 role that the theoretical factors play in hurricane preparedness for this population, compared to a population that is not socially vulnerable or disadvantaged. Theoretical considerations guided my selection of research participants in that I needed a large enough sample from the disadvantaged population to establish external validity. Based on the variance of th e outcome variable (39.097), my set precision level (.625) and confidence interval (95 percent), I determined that a sample size of 239 was needed. A desirable sample size for structural equation modeling is a 20:1 ratio of cases to the number of parameters. A 10:1 ratio may be more realistic (Kline, 2005). Using the 10:1 ratio, a minimum sample size of 130 would be required. Kline (2005) notes that the more complicated the model, the more cases needed. In order to identify the accessible population, s everal commercial listing services were contacted. Based on their information, there are approximately 128,000 residential mail listings for the county. Using 2000 US Census data, I conducted a demographic analysis by county zip code to define and locate the accessible population (Appendix 4). Additionally, I compared the US census data information and overlayed the county zip code map (Appendix 5) with the vulnerable population maps mentioned previously to identify areas with high concentrations of vuln erable households. Based on this analysis, zip codes 32501, 32505, 32535 and 32568 were targeted. I used the same process to identify zip codes for non vulnerable households. Zip codes 32508, 32514, and 32533 were targeted. A commercial listing service was used to obtain a list of 5,000 residences to include names from these two sets of zip codes. Using a random number generator, a random sample of 1,000 residences was selected for the first mailing. I decided to send 70 percent of my surveys to the z ip codes that were
52 selected for its number of vulnerable residents and 30 percent to the nonvulnerable selected zip codes. Assuming that the vulnerable population may be less inclined or able to complete the survey, I decided a 70/30 split might provid e the best opportunity to obtain statistically representative sample from each group. The first mailing resulted in approximately 160 r eturned surveys. An other round of surveys was mailed to an additional 500 residences Of those, approximately 85 were returned, for a total sample size of 245. Descriptive statisti cs for the sample are in Table 3 1 Table 3 1. Frequencies of socioeconomic characteristics for all respondents in study assessing hurricane preparedness and factors influencing preparedness in 245 households in Escambia County, Florida, 2009 Demographic Characteristics (n=245) Frequency Relative Frequency Cumulative Frequency AGE (Years) 18 3 9 40 5 9 60 67 68+ GENDER 37 98 44 63 15.2% 40.0% 18.2% 25.7% 15.3 % 55.8 % 74.0% 100.0% Male Female 127 114 52.3% 46.9% 52.3% 100.0% ETHNICITY African American Asian Native American Caucasian Other 39 4 7 175 12 16.4% 1.7% 2.9% 73.5% 5.0% 16.4% 18.5% 21.4% 95.0% 100.0% EDUCATION Did not complete HS Completed HS Some college or 2 yr college deg Completed 4 yr deg 16 41 95 89 6.6% 17.0% 38.8% 37.0% 6.6% 23.7% 63.1% 100.0% HOUSEHOLD INCOME (Mean annual income) Under 34,999 $35,000 $49,999 $50,000 $74,999 $75,000 or more 83 36 37 60 38.4% 16.7% 17.1% 24.5 % 38.4% 55.1% 72.2% 100.0% DISABLED No 190 79.8% 79.8%
53 Yes 48 20.2% 100.0% HS High School; Yr Year Since I had no way of determining the vulnerability status of a household a priori I assigned households to comparison groups post hoc I used the following factors to determine vulnerability: age (born 1941 or before); income (less than $35,000); or disabled or a disabled individual in the household. Disaster literature refers to elderly as being vulnerable. However, no specific ages are equated with the term. Therefore, I selected age 68 as this is one year past the age that all individuals would be considered full retirement age as defined by the Social Security Administration (Social Security Administration, 2010). I used income of less than $35,000 as the income threshold to examine the questionnaires more closely to determine vulnerability. There were cases where a respondent might fall into the vulnerability category based on income, but other factors or notes on the survey may have indicated the individual was nonvulnerable. An example is a respondent falling into the vulnerable status based on income level, but notes on the survey indicated that he/she was a navy pilot. I would not consider this a member of a soci ally vulnerable population. Another example might be where an individual makes less than $35,000 and there is only one person in the household. Finally, I used disabled as that is a category defined in Florida statutes as vulnerable citi zens in local emergency management plans (CEMP 2004). As a result of my decision to use a 70/30 split in the mail outs and criteria used to determine vulnerability, 124 surveys from vulnerable households and 121 surveys from the nonvulnerable households were received. A t test for independent samples was conducted to evaluate differences between the vulnerable and non vulnerable samples with regard to the outcome variable.
54 Distribution of scores for the outcome variable was normal for both groups and variance was equal (Levenes test of homogeneity, p = .111). Therefore, I concluded the two groups to be homogen ous. The sample size of 245 is adequate for analysis. Additional questions were asked regarding hurricane experience. Ninety -five percent of all respondents had experienced a hurricane within the last ten years. Over 70 percent of the respondents experienced some damage or loss from a hurricane. Finally, close to 70 percent think there is a good chance a hurricane will affect them within the next three years. Table 3 2 Frequencies of hurricane experience factors in study assessing hurricane preparedness and factors influencing preparedness in 245 households in Escambia County, Florida, 2009 Hurricane Experience (n=242) Frequency Relative Frequency Cumulative Frequency PAST EXPERIENCE No Yes DAMAGE No Yes WILL AFFECT This Year Next Year Within Next 3 Years Within Next 5 Years 10 232 57 176 38 11 115 71 4.0% 96.0% 24.4% 75.6% 16.2% 4.5% 48.9% 30.2% 4.0% 100.0% 24.4% 100.0% 16.2% 20.9% 69.8% 100.0% Instrument Development I based the questionnaire used for this research on Paton et al.s (2003) instrument used in earthquake preparation research. This questionnaire addresses the same social -cognitive factors used in the earthquake research. I added questions to address soc ial network factors that may influence preparation decisions, and, where necessary, changed questions to reflect hurricane preparation versus earthquake
55 p reparation. Dr. Douglas Paton granted permission to adapt his instrument for this research. Paton et al.s (2003) reported that a factor analysis, using maximum likelihood estimation, was applied to each set of indicators on their phase one questionnaires. The confirmatory factor analysis in their study confirmed that the properties of the scales wer e sufficient for further analyses (Paton et al 2003). However, I did not presume that the same instrument would be reliable and valid in my study. Therefore, I assessed reliability, validity and precision for my instrument. Cross -sectional designs rely on individuals (participants) and reliability varies between individuals (Wikman, 2006). This theoretical population may be educationally challenged or have other impediments to understanding the questionnaire. Therefore, I tested instruments to ens ure that a similar population would be able to understand what was being asked. I have experience working with vulnerable populations and used my own contacts to locate individuals to test the instrument. Seven people tested the instrument. Wikman (2006) argues that reliability of surveys is dependent on language used and the clarity of the questions. Therefore, I conducted a cognitive evaluation of the questionnaire with each test participant to evaluate their understanding of what I asked and the wor ds I used. The use of the cognitive approach greatly reduces the number of test cases needed to establish validity and reliability because of the insights the procedure provides to the researcher. A large test group is therefore not necessary. Changes t o the questionnaire were made based on the analysis to improve reliability and validity. Reliability tests for each instrument were conducted post -hoc Based on the reliability results, indicators were eliminated for each subsequent analysis. Reliabilit y test results are noted in Chapter 5.
56 The next consideration is validity. Adcock and Collier (2001) argue that measurement validity is a synonym for construct validity. They define measurement validity as, whether operationalization and the scorin g of cases adequately reflect the concept the researcher seeks to measure (Adcock and Collier, 2001, p. 259). In my research, I operationalized several concepts such as risk perception, self -efficacy, indivi dual intentions and others, that I believe resu lt in hurricane disaster preparedness actions. How can I determine if the concept, the items used to measure the concept and the scoring of the cases are valid? Concepts originating from theory provide a stable foundation for research (Adcock and Collier 2001). A good theory should provide better concepts that, in turn, result in improved theory (Kaplan, 1964; Adcock and Collier, 2001). This study originates from theories as discussed in Chapter 2; social cognitive theory, stress and coping theory and s ocial network theory. I chose the concepts found in these theories because they have been researched extensively and there is a body of literature that supports the use of the concepts. After deciding what concepts to use, consideration must be given to the proposed items to measure the concept. A concern with s electing the proposed items is what Adcock and Collier (2001) refer to as contextual specificity. This occurs when a score on one item may have a different meaning in a different context (Adcoc k and Collier, 2001). For example, one concept I am exploring is hazard awareness. Hazard awareness as it relates to hurricane preparedness may be different than hazard awareness as it relates to living next to a nuclear power plant. To address this challenge of contextual specificity, I used items to measure hazard awareness, as well as most of the other concepts, that were hurricane specific.
57 Convergent/discriminant va lidation examines each item and its relationship to the concepts and the focus is on shared and nonshared variance among items (Adcock and Collier, 2001). Finally, Adcock and Collier (2001) refer to nomological/construct validation as a process to tease out factors not identified in convergent/discriminant validation. This process serves as confirmation that the item scores fit the concepts. An association of concepts validly measured in a hypothesized model is positive evidence for validity (Adcock and Collier, 2001). Structural equation modeling is a common tool used for convergent/ discrimin ant validation. My research hypothesizes a decision making model for hurricane preparation. Results of the structural equation model analysis (further discussi on in Chap 5) for my research shows a causal relationship between risk perception and individual intentions resulting in dis aster preparation. T his is positive evidence of validity. Finally, Cohen (1988 p. 6) defines the precision of a sample statistic as the closeness with which it can be expected to approximate the relevant population value. It is necessarily an estimated value in practice, since the population value is generally unknown. The outcome variable of hurricane preparation is a cumulative score of yes responses to 31 different items. I determined a value of 5 as the a cceptable error in themean, resulting in a precision of .625 that was used in determining required sample size. The construction and distribution of the questionnaire followed the tailored design method (Dillman, 2007). The questionnaire was constructed for ease of reading and answering questions quickly. The questionnaire was six pages long and was designed to take less than 30 minutes to complete. It was mailed with a cover letter addressed by
58 name to the resident. The cover letter included informed consent information. Additionally, a self addressed, stamped envelope was included for ease of response. I a l so included a Florida Division of Emergency Management hurricane preparation pamphlet checklist in the mail out as a token of appreciation for participation. A follow up postcard was mailed approximately one week after the questionnaires were mailed out. This process was used for all 1,5 00 mail outs Concepts and Variables This study focused on three primary constructs as they relate to prepar edness, motivational factors, coping self assessment and intentions to prepare. Motivational factors are defined as risk perception, hazard awareness and motivational relevance. Motivations are linked to intention by coping self assessment that includes perceived control over outcomes, self efficacy and action coping. Finally, intentions are linked to preparation by individual actions or actions taken working through a social network. A table of variables and items are provided at Appendix A. Predictor Variables The following explains each vari able and provides the items used to measure that variable. Scoring for all variables except the outcome variable of hurricane preparation was mean cumulative (total divided by number of items). The scoring for the outcome variable was cumulative (number of yes responses to 31 items). Motivational F actors Risk perception. Perceived risk has been examined in many different ways. Risk perception probably includes at least three dimensions real risk, risk experience and risk target (Sjoberg, 2000). Real risk may be one determinant of perceived risks in some contexts (Sjoberg, 2000). For example, Lichtenstein et al., 1978, estimated
59 mortality rates for illnesses and accidents were related to statistical data and as a result, risks were perceived, by the average person, in a rather verit ical manner (Sjoberg, 2000, p. 2). Risk experience may also influence risk perception (Thompson and Mingay, 1991). An individual participating in a FEM A roundtable discussion regarding hurricanes shared an example of how her experience influenced her risk perception. The individual stated that she had never experienced a hurricane and was therefore unaware of the associated risk. After her personal exp erience with Hurricane Ivan, where she suffered serious damage to her home, her perception has changed and she now recognizes the risks associated with hurricanes. Finally, risk target is important in risk perception. Individuals will make different risk perception estimates based on whether the risk is to themselves, their family, or people in general (Sjoberg, 2000). Numerous studies conducted by Paton and others in examining volcanic hazard perceptions linked perceived risk to proximity to the hazard, likelihood of future disasters and past experience (Paton et al., 2003; Johnston et al., 1999; Lindell and Whitney, 2000). Johnson et al. (1999) found, in a study of two communities, that direct experience affected risk perception, but did not necessaril y influence preparedness (Johnston et al, 1999). In this research, participants were asked to express their opinions and perceptions from strongly agree to strongly disagree on hurricanes as a threat to their person and property. The following statemen ts were used: A hurricane could pose a threat to your personal safety A hurricane could pose a threat to your daily life A hurricane could pose a threat to your property Additionally, a question was asked concerning the timing and impact of future hurrica nes: The most likely time within which a damaging hurricane could affect me is:
60 This year ______ Next year ______ Within next 3 years ______ Within next 5 years ______ Hazard awareness. Paton (2003) refers to awareness as how often someone thinks or talks about hazards. Deriving from the community psychology literature, Dalton et al. (2001) refer to this as critical awareness. This is a process that involves conscious reasoning about i ssues people perceive as critical or salient (Paton et al., 2003, p. 6). Paton (2003) argues that critical awareness is necessary because natural hazard events are rare. Hurricane preparations in Florida take place in the spring prior to the start of th e hurricane season in June. Hurricanes compete with other problems for attention. The attention paid to hurricane preparation will be evident in how much people think or talk about it (Paton, 2003). Bagozzi and Dabholar (2000) state that how and what p eople communicate with one another affects their decision -making processes. Thus, the frequency that individuals discuss hurricanes would represent the importance they place on that natural hazard (Paton, 2003). To address this factor in this research, r espondents were asked to reply on a scale of 1 being not at all to 5 being a great deal to the following: How much do you think about hurricanes? How much do you talk about hurricanes? Motivational Relevance The transactional model of stress and coping defines stressors in two categories, primary and secondary appraisal. Wenzel et al. (2002) define primary appraisal as a persons judgment about the significance of an event. An individual will determine if the event is stressful, positive, control lable, challenging, benign, or irrelevant (Wenzel et al., 2002). Primary appraisal consists of several
61 factors. I examined two of these factors, motivational relevance and causal focus as precursors to hazard preparation decisionmaking in this resear ch. Motivational relevance refers to a stressor that will have a major impact on an individuals well -being (Wenzel et al., 2002; Folkman et al., 1986). In the case of high motivational relevance, an individual may experience distress or anxiety. Hurric anes are unpredictable and uncontrollable. Therefore, hurricanes can cause a certain amount of anxiety or fear for individuals often threatened by them. In studying earthquake response, it has been argued that earthquake anxiety may cause individuals to prepare less for earthquakes (Duval and Mulilis, 1999; Lamontaigne and LaRochelle, 2000; Paton, 2003). In order to lessen their anxiety levels, individuals may choose to ignore information about what is causing their anxiety (Paton, 2003). In reviewing t he disaster literature, I was unable to locate a scale that addressed hurricane anxiety. However, Paton et al. (2003) addressed earthquake anxiety in his research. Adopting similar statements to reflect hurricane anxiety, the respondents were asked to reply on a scale from 1 being not at all to 5 being a great deal to the following: I get nervous when there is discussion about approaching hurricanes When hurricane ads come on TV, I change the channel or dont pay attention I avoid things that remind m e of hurricanes If I believe a hurricane is approaching, I make sure I know the evacuation route If I believe a hurricane is approaching, I make sure I know where my shelter is I avoid thinking about hurricanes Coping Self Assessment Perceived control over outcomes. Assessing an individuals coping resources and options falls into the secondary appraisal category of the transactional model of stress and coping (Wenzel et al., 2002). Whereas primary appraisal focuses on the situation, secondary appraisal focuses on what one can do about the situation.
62 Perceived control over outcomes is similar to what Paton et al. (2003) labels outcome expectancy. Once individuals are aware of a particular hazard, such as a hurricane, they then make i ndividual judgments with regard to what actions to take and whether or not those actions constitute successful preparation actions for the hurricane (Paton, 2003). Paton et al. (2003) argues that outcome expectancy precedes self -efficacy. Individuals wil l make assumptions about whether preparatory actions will be successful or not before deciding to take a particular preparation action (Paton et al., 2003). They are more likely to take those actions that they believe will be successful. The same outcome expectancy measures used by Paton et al. (2003) were used in this research. On a scale of strongly agree to strongly disagree, respondents were asked to respond to the following statements. Hurricanes are too destructive to bother preparing for A serio us hurricane is unlikely to occur during your lifetime Preparing for hurricanes will reduce damage to my home should a hurricane occur Preparing for hurricanes will improve my everyday living conditions Preparing for hurricanes will improve the values of my house/property Preparing for hurricanes will reduce the disruption to family/community life following a hurricane. Preparing for hurricanes is a hassle for me Self efficacy. Health behavior research has shown that self efficacy plays a central role in ones health decisions (Wenzel et al, 2002). Bennett and Murphy (1997) argue that efficacy beliefs may be generalized (I can cope with events in my life), or may be behavior -specific, such as making decisions about using drugs or even preparing for a n atural hazard such as a hurricane. Paton (2003, p. 212) argues that
63 if a person forms a favorable outcome expectancy, whether or not they progress towards the formation of preparedness intentions is a function of the level of their self efficacy beliefs The person-relative-to event (PrE) theory postulates that self efficacy refers to self assessments of ones knowledge, skill, ability, energy, and financial resources in relation to the hazard event (Lindell and Whitney, 2000, p. 14). Paton et als (2003) analysis of self efficacy tends to emphasize the generalized version, whereas his statements focus on issues and problems that one might deal with in everyday life. On a scale of strongly agree to strongly disagree, respondents were asked to respond to the following statements: I have considerable control over what happens in my life I can solve most of the problems I have by myself What happens to me in the future mostly depends on me I can do a lot to change many of the important things in my lif e I can do just about anything if I really set my mind on it I rarely feel helpless in dealing with the problems of my life Action Coping. Carver et al. (1989) describe a number of dimensions of coping. Active coping is the process of taking active steps to prevent the stressor or taking action to improve the stressful situation (Carver et al., 1989). Active coping is similar to what Lazarus and Folkman (1984) and others label problem -focused coping. Paton et al. s (2001) research on volcanic hazard preparedness concluded that problem -focused coping should be included as a factor in determining preparation for natural hazards. The problem -focused coping predictor is also a factor found in the person-relativeto event (PrE) model (Duval and Mulilis, 19 99; Lindell and Whitney, 2000; and Patonet al., 2003). Paton et al. (2003) used Carver et al.s (1989) measure of action coping in his disaster research, and I used the same measure in this research. On a scale of one to four (I usually dont do this at all to I usually do this a lot), respondents were asked to
64 respond to the following statements in regard to dealing with their everyday life problems. I try to come up with a strategy about what to do I make a plan of action I think hard about what steps to take I think about how I might best handle the problem Individual Intentions Health behavior models have shown that intentions are an indicator of whether an individual adopts preventative behavior with regards to health threats (Paton et al. 2003). Paton et al. (2003, p. 10) define intentions as a precursor to adjustment adoption. Paton et al. (2003) incorporated measures cited by Bennett and Murphy (1987) in his survey. Responding on a scale of no, possibly or definitely, I asked respondents to respond to the following with regards to their hurricane preparation actions over the next month or so: Check your level of preparedness for hurricanes Increase your level of preparedness for hurricanes Become involved with a local group to disc uss how to reduce hurricane damage or loss Seek information on hurricane risk Seek information on things to do to prepare Strong and Weak Soc ial Networks and Social Support. Social networks are the links between people and the relationships they share (H eaney and Israel, 2002). Individuals have various forms of relationships with others that may impact their decision making processes about everyday life events or specific actions such as hurricane preparedness. Granovetter (1973) distinguishes between s trong and weak ties in networks. The strength of a tie is probably linear and has characteristics of reciprocity, intensity, complexity, density, homogeneity and geographic dispersion
65 (Granovetter, 1973; Heaney and Israel, 2002). Granovetter (1973) state s that weak ties link members of small groups and strong ties are concentrated within particular groups. Using this definition, one might consider linkages w ith family and close friends a strong tie. A weak tie may be a relationship that one has with an acquaintance at church or school. Granovetter (1983, p. 209) states that, Weak ties provide people with access to information and resources beyond those available in their own social circle; but strong ties have greater motivation to be of assistance and are typically more easily available (p. 209). During a recent FEMA project, a roundtable discussion with disadvantaged community members in a rural county reflected how vulnerable groups are dependent on family and friends (strong ties) for assistance wi th disaster preparedness One individual stated that her son helped her with obtaining filled sandbags in preparation for possible flooding. Another individual stated that the local sheriff called to inform of possible flooding problems and where they could go for assistance. Whereas a sheriff in a community may be considered a weak tie, this was a rural community where everyone knows each other quite well. In this case, the sheriff may be considered a strong tie (FEMA Emergency Preparedness Demonstration Project, 2009). My research explores the strong and weak ties that a disadvantaged community member may or may not use in executing disaster preparedness actions. Review of the research has shown that disadvantaged people tend to rely on strong ties more than others (Ericksen and Yancey, 1977; Granovetter, 1983). Responding to the question, Where do you look to get information about how to prepare for hurricanes? participants were asked to rate on a scale of never to always their family, friends,
66 neighbors, churches, local government, co workers, social clubs, county cooperative extension, TV, radio and other. The respondents were also asked: Who do you turn to when you need to prepare for a hurricane? Based on Granovetters (1983) definition of strong and weak ties, I classified strong ties as family, friends, neighbors and coworkers and weak ties as local government, social clubs, TV, and radio. Moderator Variables: Response Efficacy, Past Experience, Perceived Responsibilit y: Disaster research has included additional variables like personal responsibility, past e xperience and response efficacy that might be considered in predicting preparedness (Bishop et al., 2000; Duval and Mulilis, 1999; Lindell and Whitney, 2000; Paton et al., 2000). Paton et al. (2003) refers to these variables as moderating variables. Paton et al. (2003) argue that these variables influence the link between intention and preparation. I have included the same variables in this research. Response efficacy describes the resources and capabilities an individual may or may not have in regard to preparation (Paton et al, 2003eqc). Examples are time, skill, financial, and physical resources. Even if people intend to prepare for a hurricane, they may no t have the resources to do so. Participants were asked to answer the question, To what extent might each of the following prevent you preparing for hurricanes? They responded on a scale from 1 being not at all to 5 being a great deal to the follow ing: The cost The skill or knowledge required Time to do them Other things to think about Need for co operation with others
67 Personal responsibility is an important consideration in hurricane preparations. Researchers have found a relationship between an individuals acceptance of personal responsibility for their own safety and preparation actions (Ballantyne et al., 2000; Duval and Mulilis, 1999; Lindell and Whitney, 2000; Mulilis and Duvall, 1995; Paton et al., 2000). Ballantyne et al. (2000) found t hat the availability of hazard information lessened preparedness action by individuals. People are less likely to prepare if they perceived that others (local government agencies) are re sponsible for the safety of citizens. Disadvantaged residents made n umerous comments about individuals taking personal responsibility for their safety during the FEMA project (FEMA Emergency Preparedness Demonstration Project, 2009) Several individuals stated others should not depend on local government for the ir safety. Participants in my research were asked to describe the extent to which they agree or disagree (scale of 1-5) with each of the following statements: I feel responsible for preparing for a hurricane The local government is responsible for making sure that I am prepared for a hurricane Past experience assesses whether an individual has experienced a hurricane and whether or not they have incurred damage or loss due to that hurricane. Participants were asked the following questions: Have you been in a hurr icane in the last 10 years? If yes, what year(s) did this occur? If yes, did you experience damage or loss (i.e., requiring repairs/insurance claims)?
68 Outcome Variable Hurricane Preparation Preparation is assessed using a preparedness scale created from the Ready.gov website. The website provides a stepby -step plan to prepare for hurricanes. Research participants were asked to respond to the following items by checking yes or no: I ha ve an emergency kit containing: Water, one gallon of water per person per day for at least three days, for drinking and sanitation Food, at least a threeday supply of non-perishable food Battery -powered radio and extra batteries NOAA Weather Radio with tone alert and extra batteries Flashlight and extra batteries First aid kit Whistle to signal for help Dust mask, to help filter contaminated air and plastic sheeting and duct tape to shelter in place Moist towelettes, garbage bags and pl astic ties for personal sanitation Wrench or pliers to turn off utilities Can opener for food (if kit contains canned food) Local maps Cell phone with charger Medications and glasses Important family documents such as copies of insurance polici es, identification and bank account records in a waterproof, portable container Cash or traveler's checks and change
69 Emergency reference material such as a first aid book or information from www.ready.gov Sleeping bag or warm blanket for each person Complete change of clothing including a long sleeved shirt, long pants and sturdy shoes. Household chlorine bleach and medicine dropper Fire Extinguisher Matches in a waterproof container *Personal hygiene items Mess kits, paper cups, plates and plastic utensils, paper towels Paper and pencil Do you have a family emergency plan? Do you know where your family will meet, both within and outside of your immediate neighborhood? Do you have an out of -town contact to communicate with? Do yo u have a plan for evacuation? Do you know the evacuation route? Do you know your designated shelter? Also, to test respondents hurricane knowledge, three true/false questions were asked about the meaning of a hurricane watch, a hurricane warning and hurricane categories: A hurricane watch means a hurricane could hit within 24 hours? A hurricane warning means a hurricane could hit within 36 hours? Hurricane categories (for example, Category 1, 2, 3, 4 or 5) are based on wind speed?
70 Analyses Demographic s Demographic variables included age, ethnicity, average annual household income, gender, residence, transportation resources, household characteristics, e.g., education and disability. As stated previously, age, income level and disability were the only factors used to determine vulnerability. Comparison of Central Tendency Several of the research hypotheses compare the vulnerable population to the nonvulnerable population and the impact of the predictor variables on the outcome variable. The t test is a parametric test that provides reliable information about the sampling distribution (Sheskin, 2007). Therefore, t tests for two independent samples were ran on all the variables to determine equal variances. Pearson Correlation Measures of correla tion are inferential statistical m easures that reflect the strength of the relationship between variables. The hypothesis that particular variables in the model are correlated with the outcome variable requires an analysis that can compare the correlations for t he two different populations. Structural Equation Modeling Although regression analysis provides information about relationshi ps, the researcher cannot draw conclusions regarding directionality of causal relationships (Sheskin, 2007). H owever, researchers use past studies to support their argument that there is causality. Hair et al. (1995, p. 622) explain structural equation modeling as a series of separate but interdependent multiple regression equations. Struc tural
71 equation model ing allows analysis of multiple relationships and multiple independent and dependent variables (Sheskin, 2007). Pat o n and his colleagues used structural equation modeling in their 2003 earthquake research (Paton et al, 2003). The model I am using in my research is similar to the Paton model. Therefore, it is appropriate to analyze the results in much the same way. Summary This chapter presented the methods and scientific reasoning used in this research. Included was a discussion of the concept ual framework linking variables, units of analysis, site selection, sampling framework, data collection, and the overall research design and analysis used in this study.
72 CHAPTER 4 RESULTS I conducted several statistical analyses to explore social cognitiv e factors that impact hurricane preparation decisions by vulnerable and nonvulnerable populations. First, I conducted a reliability analysis on all instr uments and eliminated items as necessary. I then examined the data to identify useable cases. Afte r discarding incomplete responses, useable responses remained. The sample size limited the types of analyses I could use to compare the two groups. I used correlation to estimate the relationships between variables for each group and binary logistic regr ession to analyze the hypothesized model, as a precursor to the structural equation model (SEM) analysis. Finally, the SEM analysis was conducted for the hypothesized model. Due to sample size limitations, I was unable to conduct an SEM on the two groups However, the SEM for the entire group resulted in a revised model. Reliability Testing of Instruments A reliability coefficient test was conducted for each variable. Cronbach alpha scores were used to determine the extent to which item responses correlate with each other. Alpha scores were used to determine if and what items should be eliminated from each variable. I used the cut off of .70 for alpha as this is the widely accepted reliability coefficient in the social sciences for a set of item s to be considered a scale (Nunnaly, 1978). The results of this analysis resulted in indicators being eliminated from risk perception, motivational relevance and perceived control (see Table 4 -1)
73 Table 4 1. List of Variables and Scalar Response Range Variables #Questions /Items Scalar Response Range Cronbach alpha Mediators: Risk Perception 3 1 5 Strongly disagree to strongly agree .750* Hazard Awareness 2 1 5 Not at all to a great deal .731 Motivational Rele 2 1 5 Not at all to a great deal .740* Perceived Control 4 1 5 Strongly disagree to strongly agree .757* Self Efficacy 6 1 5 Strongly disagree to strongly agree .766 Action Coping 4 1 4 I usually dont do this at all to I usually do this a lot .878 Individual Intention 5 1 3 No, possibly, definitely .842 Strong Social Ntwk 8 1 5 Never to always .862 Weak Social Ntwk 14 1 5 Never to always .829 Moderators: Response Efficacy 5 1 5 Not at all to a great deal .795 Personal Responsibility 1 1 5 Strongly disagree to strongly agree .840* Past Experience 3 -Nominal responses .098** Outcome: Hurricane Prep 31 0, 1 No or Yes .881 Items deleted. ** Removed from further analyses due to low reliability scores Items removed: Risk Per ception one item removed The most likely time within which a damaging hurricane could affect me is: This year ____ Next year ____ Within next 3 years ____ Within next 5 years ____ Motivati onal Relevance four items removed I get nervous when there is discussion about approaching hurricanes When hurricane ads come on TV, I change the channel or dont pay attention I avoid things that remind me of hurricanes I avoid thinking about hurricanes Perc eived Control three items removed Hurricanes are too destructive to bother preparing for A serious hurricane is unlikely to occur during your lifetime Preparing for hurricanes is a hassle for me
74 Personal Responsbility one item removed The local government is responsible for making sure that I am prepared f or a hurricane There were many data missing in the strong social network and weak social network variables. Although respondents were asked to fill in one circle for each item, it appeared that many of them chose to select only those items that applied to them. Therefore, I decided to recode missing data on surveys where some but not all items had been checked in questions 14 and 15 to never. The recoded responses were not used in the structural equation model analysis. Preliminary Screening of Cases A total of 72 participants were missing large portions of data in a non random patte rn. Therefore, these responses were discarded. Of the remaining 170 participants, less than one percent of data points were missing. When the amount of missing data are few (i.e., <5%), the patterns of missing data are nonconsequential (Kline, 2005). An expectation -maximization procedure available in SPSS 17 was utilized to impute missing values. This procedure has been shown to yield more accurate parameter estimates and standard errors than more traditional methods of handling missing data (e.g., list wise or pairwise deletion; Schafer, 1997). An alternative approach, full information maximum likelihood (FIML) estimation was considered. However, this approach would have required estimation of the model using latent (unobserved) constructions instead o f examining a path model with observed variables (i.e., total scores). In the present study, the latter approach is preferred because analysis of latent variables introduces a larger number of parameters, and thus a larger sample size would be required.
75 Items other than those that comprise the individual intention and hurricane preparation scales were screened for univariate outliers, defined as responses greater than 3.29 standard deviations from the mean. The intention and preparation items were not con tinuous, and therefore they do not need to conform to the assumption of univariate normality. A total of 16 univariate outliers were identified and deleted from the sample, resulting in a final sample of 154 participants. After deletion of outliers, all it ems met standard criteria for univariate normality (i.e., skew between 2 and 2; kurtosis between 7 and 7; Kline, 2005). Mean scores for all scales were then computed. It should be noted that the following items were reverse scored so that directionality was consistent within scales: Hurricanes are too destructive to bother preparing for A serious hurricane is unlikely to occur during your lifetime Preparing for hurricanes is a hassle for me The local government is responsible for making sure that I am pr epared for a Hurricane All tot al scores met criteria for univariate outliers and normality. One dichotomous variable (past hurricane experience) was created by recoding past experiences into either Yes or No. All variables included in the path analysis were also screened for multivariate outliers using a regression procedure outlined by Tabachnick and Fidell (2007). With 2 = 26.1. Thus, multivariate outl iers were operationalized as cases with Mahalanobis Distance Values greater than 26.1. Using this method, one multivariate outlier was detected and deleted. The subsample of participants retained for all further analyses consisted of the remaining
76 153 case s. Means, standard deviations, and intercorrelations among variables are presented in Table 4-4. Frequency of Response Data Socio D emographic Characteristics Analysis began with a review of the socio-demographic characteristics of the remaining 153 cases to assign vulnerability status. An individual was coded as vulnerable if they were 68 years of age or older, and/or had a household income of less than $35,0002, and/or indicated there was a disabled person in the home. The heading on the questionnair e requested that the person responsible for making hurricane decisions for the household complete the questionnaire. Table 4 2. Frequency distribution by sample by key demographic and socioeconomic characteristics of respondents in 153 households in Es cambia County, Florida, 2009 Demographic Characteristics (n=153) Frequency Relative Frequency Cumulative Frequency AGE 18 3 9 40 5 9 60 67 68+ GENDER 31 73 24 24 20.4% 47.7% 15.7% 15.7% 20.4 % 68.4 % 84.2% 100.0% Male Female 79 72 51.6% 47.1% 52.6% 100.0% ETHNICITY African American Asian Native American Caucasian Other 16 2 2 121 7 10.5% 1.3% 1.3% 79.1% 4.6% 11.4% 12.8% 14.1% 95.3% 100.0% HOUSEHOLD INCOME Under $34 ,999 $35,000 $49,999 37 28 24.1% 18.3% 27.4% 48.1% 2 $35,000 was used as the threshold to examine surveys for vulnerability. If someone listed household income as less than $35,000, it is possible that other socioeconomic characteristics (e.g., number in household) may have determined individual was not vulnerable.
77 $50,000 $74,999 $75,000 or more 27 43 17.6% 29.8% 68.1% 100.0% DISABLED No Yes 134 18 87.6% 11.8% 88.2% 100.0% Data in Table 4 2 are categorical and are presented as frequencies, relative frequencies and cumulative frequencies. Six age categories were used. The majority of the respondents were under the age of 59 (68%). The number of respondents who met the age criteria for vulnerability determination was 24 (~16%). There were 79 (52%) male and 72 female (48%) respondents and the majority of respondents were Caucasian (79%). The majority of respondents (84%) had completed some college or training past high school. Only 18% had a high school diploma or less. Income was a characteristic for determining vulnerability; 27% of the respondents fell into the $35,000 or less household income category. Finally, 11% of the respondents were disabled or had a disabled person in the hom e. A frequency analysis and review of all respondents demographic data and socioeconomic data used to measure vulnerability shows that 100 (65%) of the respondents fall into the nonvulnerable category and 53 (35%) fall into the vulnerable category. I n a few cases, I changed the respondents category based on a review of all of his/her responses. For example, one respondent fell into the vulner able income level and notes on the survey that indicated they were part of an active duty military household. I changed this respondents category to not vulnerable. Hurricane Experience Additional questions were asked in regards to h urricane experience specifically if respondents had been in a hurricane in the last ten years and, if so, whether they
78 suffered any damage or loss. I also asked if they think a hurricane will affect them this year, next year, within the next three years or the next five years. The majority (96%) of respondents have experienced at least one hurricane and approximately 74% of those experienced some sort of damage or loss that required repairs or insurance claims. Respondents were also asked when they thought a hurricane might affect them in the future. The majority (79%) indicated that it would happen in the next three to five years.3 Table 4 3. Frequencies of hurricane experience factors in study assessing hurricane preparedness and factors influencing preparedness in 153 households in Escambia County, Florida, 2009 Hurricane Experience (n=153) Frequency Relative Frequency Cumulative Frequency PAST EXPERIENCE No Yes DAMAGE No Yes WILL AFFECT This Year Next Year Within Next 3 Years Within Next 5 Years 6 145 37 107 27 4 73 45 4.0% 96.0% 25.6% 74.4% 18.1% 2.6% 49.0% 30.3% 4.0% 100.0% 25.6% 100.0% 18.1% 20.7% 69.7% 100.0% Additionally, questions were also asked in regards to the difference between a hurricane watch and warning and how hurricanes are classified. A majority (75%) of the respondents knew that a hurricane watch meant a hurricane could strike within 48 hours, b ut only 43% of respondents knew that a hurricane warning meant a hurricane could strike within 36 hours. Almost all (99%) of respondents knew that hurricane categorie s, e.g., Category 1, 2, 3, 4, and 5, are based on wind speed. 3 Questionnaire was administered during the months of September and October. This was mid hurricane season.
79 These data indicate that th e sample is heavily weighted to the non vulnerable population (100 nonvulnerable vs 53 vulnerable). Independent samples t -tests were ran on the outcome and all mediator/moderator variables. Levenes statistic for all variables including the outcome vari able had a pvalue for F > .05. This would indicate that there are not two separate populations (nonvulnerable and vulnerable), but rather one population. Discussion of the sample size and how it impacts structural equation modeling will be discussed later. This research is based on comparing nonvulnerable and vulnerable populations. As such, I analyzed the cases that did not meet the useable sample criteria. It was found that the majority (76%) of the non-useable cases met the vulnerability criteria. Forty -two (42) percent were over the age of 68 and 55 percent met the income criteria of $35,000 or less. Additionally 31 percent were disabled. This would indicate that the elderly population may have had difficulty completing the survey. A cursory review of the completed surveys show that some respondents missed entire pages and some just did not complete different sections. This may have been due to the booklet type survey instrument and the length (six pages) of the instrument. Although analysis supports treating the sample as one population, my research is based on comparing two populations. Therefore, I conducted additional analyses to examine differences between the two groups for predictor and outcome variables. These findings do permit me t o draw conclusions between the two samples, but may not be generalized beyond this study. Analyses and Hypotheses Testing Results My first hypothesis states that the same causal pathways found in Paton et al.s (2003) social cognitive model for earthquake preparation decisions will be the same for
80 the hurricane preparation decision process. Additionally, I hypothesized that the variables of strong and weak social networks added as a link between intentions and hurricane preparations in the model would increase the explanatory power of the model. Finally, I hypothesized that vulnerable and nonvulnerable populations would differ in their hurricane preparation decision process. The hypotheses are similar in that they examine variations on the same model and compare two sub -populations using the model. H1: The causal pathways found in Patons social -cognitive model for earthquake preparation decisions will be the same for the hurricane preparation decision process. I conducted a multivariate correlati on analysis to exami ne relationships between the nine mediator variables and the outc ome variable. Additionally, two moderator variable s were analyzed to determine their relationship to the outcome variable of hurricane preparation and another mediator variable, individual intentions. I first examined the correlat ion coefficients between all nine predictor variables and the outcome variable for the total s ample of 153 households (Table 4-4). Then I treated the vulnerable and non vulnerable samples separ ately (Table 45 and 46). My model (Fig 41) posits a relationship between risk perception, hazard awareness, motivational relevance and perceived control over outcomes. F or the combined samples (Table 44) risk perception is significantly correlated to perceived control (corr. = 0.29, p < 0.01) and motivational relevance is significantly correlated to perceived co ntrol (corr. = 0.19, p < 0.05), but, hazard awareness is not significantly correlated to perceived control. A similar pattern appears for th e non vulnerable sample (Table 45), although the correlation between motivational relevance and perceived
81 control is higher (corr. = 0.30, p < 0.01). None of these relationships are significant fo r the vulnerable sample (Table 4-6). Mean scores were simi lar for perceived control for both the vulnerable and nonvulnerable samples. The model indicates that perceived control affects both self -efficacy and action coping (Fig 4-1). These results show that perceived control is positively and significantly cor related to action coping for the combined samples (corr. = 0.30, p < 0.01), the nonvulnerable sample (corr. = .29, p < 0.01), and the vulnerable sample (corr. = 0.33, p < 0.05). Perceived control is significantly correlated with self efficacy for the nonvulnerable sample (corr. = 0.25, p < 0.05) but not for the combin ed or vulnerable samples. Mean scores were similar for action coping and self -efficacy for all samples. Table 4 4. Combined sample cor relation coefficients between 12 variables used to assess hurricane preparedness and factors influencing preparedness in 153 households in Escambia County, Florida, 20094 Variables M SD 1 2 3 4 5 6 7 8 9 10 11 1. Hurricane Prep 22.00 6.07 -2. Risk Perception 4.42 .72 .06 -3. Hazard Awareness 3.03 .88 .27** < .00 1 .22** < .00 1 -4. Motiv Relevance 3.62 1.25 .29** < .001 .32** < .001 .05 -5. Perceived Control 3.87 .76 .17* .04 .29** < .00 1 .09 .19* .02 -6. Self Efficacy 4.13 .65 .17* .04 .11 .00 .05 .16 -7. Action Coping 3.54 .57 .09 .16 .19* .01 .18* .02 .30** < .001 .15 -8. Individual Intent 1.82 .56 .25** < .00 1 .25** < .00 1 .28** < .00 1 .29** < .00 1 .18* .03 .01 .21** < .00 1 -9. Strong Soc Net 2.84 .84 .15 .20* .01 .15 .18* .03 .16* .05 .07 .08 .14 -10. Weak Soc Net 2.52 .69 .10 .22** < .001 .01 .28** < .001 .08 .01 .22** < .001 .36** < .001 .42** < .001 -11. Respons Eff 2.06 .87 .35** < .00 1 .07 .01 .10 .16 .27** < .00 1 .07 .16 .14 .04 -12.Personal Resp 4.5 .48 .19* .02 .32** <.001 .25** <.001 .19* .02 .36** <.001 .16 .36** <.001 .15 .02 .17 .10 ** Correlation is significant at the 0.01 level ( 2 tailed). Correlation is significant at the 0.05 level (2tailed). 4 This correlation analysis was also used in the structural equation model analysis
82 Table 4 5. Nonvulnerable sample cor relation coefficients between 12 variables used to assess hurricane preparedness and factors influencing preparedness in 100 households in Escambia County, Flori da, 2009 Variables M SD 1 2 3 4 5 6 7 8 9 10 11 1. Hurricane Prep 22.57 5.44 -2. Risk Perception 4.39 .73 .23* .02 -3. Hazard Awareness 3.01 .94 .27** .01 .29** < .00 1 -4. Motiv Relevance 3.49 1.25 .45** < .00 1 .35** < .00 1 .04 -5. Perceived Control 3.83 .67 .32** < .001 .31** < .001 .17 .30** < .001 -6. Self Efficacy 4.20 .56 .16 .01 .06 .17 .25* .02 -7. Action Coping 3.57 .53 .18 .16 .31** .15 .29** < .001 .21* .04 -8. Individual Intent 1.8 .58 .31** < .00 1 .31** < .00 1 .30** < .00 1 .27** .01 .27** .01 .03 .16 -9. Strong Soc Net 2.95 .83 .19 .27** .01 .22*.03 .25* .02 .15 .08 .02 .27** .01 -10. Weak Soc Net 2.47 .72 .16 .17 .15 .26** .01 .11 .05 .21* .04 .37** < .001 .53** < .001 -11. Respons Eff 1.98 .84 .23* .02 .03 .05 .10 .12 .14 .05 .15 .16 .03 -12.Personal Resp 4.46 .66 .02 .13 .02 .06 .26 .25 .37** .007 .04 .05 .10 .20 ** Correlation is significant at the 0.01 level (2tai led). Correlation is significant at the 0.05 level (2tailed). Table 4 6. Vulnerable sample cor relation coefficients between 12 variables used to assess hurricane preparedness and factors influencing preparedness in 53 households in Escambia County, Florida, 2009 Variables M SD 1 2 3 4 5 6 7 8 9 10 11 1. Hurricane Prep 20.91 7.05 -2. Risk Perception 4.48 .71 .20 -3. Hazard Awareness 3.07 .75 .30* .03 .04 -4. Motiv Relevance 3.85 1.23 .16 .25 .07 -5. Perceived Control 3.93 .91 .02 .27 .03 .03 -6. Self Efficacy 3.98 .77 .15 .27 .08 .05 .09 -7. Action Coping 3.46 .63 .04 .18 .02 .27 .33* .02 .06 -8. Individual Intent 1.92 .51 .21 .10 .21 .31* .03 .02 .00 .36** .01 -9. Strong Soc Net 2.63 .83 .03 .09 .02 .14 .23 .15 .15 .05 -10. Weak Soc Net 2.60 .63 .01 .00 .18 .29* .04 .03 .09 .28 .31* .03 .24 -11. Respons Eff 2.23 .92 .48** < .00 1 .14 .17 .17 .09 .45** < .00 1 .21 .25 .21 .23 -12.Personal Resp 4.58 .69 .37** <.001 39** <.001 36** <.001 25* .02 43** <.001 13 .37** <.001 .23* .02 .09 .08 05 ** Correlation is significant at the 0.01 level (2tailed). *Correlation is significant at the 0.05 level (2tailed.)
83 The model (Fig. 41) further supposes a relationship between self efficacy, action coping and individual intentions. The results show no significant relationship between self -efficacy and individual intentions or self efficacy and action coping in the combined, nonvulnerable and vul nerable samples. Means scores for self efficacy and action coping were slightly lower for the vulnerable samples as compared to the combined and the nonvulnerable samples. The results indicated additional relationships with individual intentions not dia grammed in the model. Hazard awareness was significantly correlated with individual intentions in the combined sample (corr. = 0.28, p < 0.01) and the nonvulnerable sample (corr. = 0.31, p < 0.01), but not the vulnerable sample. Motivational relevance c orrelated significantly with individual intentions in the combined sample (corr. = 0.25, p < 0.01), non vulnerable sample (corr. = 0.31, p < 0.01), and the vulnerable sample (corr. = 0.31, p < 0.05). Finally, the results showed a relationship between indi vidual intentions and perceived control only in the combined sample (corr. = 0.18, p < 0.05). The nonvulnerable and vulnerable samples showed no relationship. My model indicates a direct relationship between individual intentions and the outcome variable, hurricane preparation, or an indirect relationship with the outcome variable through strong or weak social networks. The results reflect a positive and significant correlation between individual intentions and hurricane preparation in the combin ed sample (corr. = 0.25, p< 0.01) and the nonvulnerable sample (corr. = 0.31, p < 0.01). This relationship was not significant in the vulnerable sample. Mean scores were similar for all samples. The non vulnerable sample showed a relationship between i ndividual intentions and strong social networks (corr. = 0.27, p < 0.01). Individual intentions were strongly correlated with weak social networks in the combined
84 sample (corr. = 0.36, p < 0.01), non vulnerable sample (corr. = 0.37, p < 0.01), and the vul nerable sample (corr. = 0.31, p < 0.01). Means scores were similar for all samples. Finally, my model reflects a relationship between the moderator variables of response efficacy, past experience, and personal responsibility with individual intentions an d the outcome variable, hurricane preparation. As previously noted, past experience was eliminated from the analyses due to low reliability scores. The results reflect a positive and significant correlation between response efficacy and hurricane preparation in the combined sample (corr. = 0.35, p < 0.01), nonvulnerable sample (corr. = 0.23, p < 0.05), and the vulnerable sample (corr. = 0.48, p < 0.01). The results also reflect a significant negative correlation between response efficacy and self effica cy in the combined sample (corr. = 0.27, p < 0.01) and the vulnerable sample (corr. = -0.45, p < 0.01). Mean scores for response efficacy were higher for the vulnerable sample as compared to the combined and nonvulnerable samples. The results also show ed a significant correlation between personal responsibility and hurricane preparation in the combined sample (corr. = 0.19, p < .05) and the vulnerable sample (corr. = .37, p < .001). Only the vulnerable sample reflected a significant correlation between personal responsibility and individual intentions (corr. = 0.23, p < .05). Mean scores for personal responsibility were higher for the vulnerable sample as compared to the combined and non vulnerable samples. Analysis for this research also included stru ctural equation modeling. The hypothesize d model is presented in Figure 41. Multiple fit indices in addition to the chi square statistic were used to evaluate model fit. This approach is recommended because the chi -square statistic is influenced by sample size (Hu & Bentler, 1999).
85 Additional fit indices that were examined were the root mean square error of approximation (RMSEA), the comparative fit index (CFI), Tucker Lewis index (TLI, also known as the non normed fit index), and the standardized root m ean square residual (SRMR). Values greater than .90 and .95 for the CFI and TLI indicate acceptable and good fit, respectively (Hu & Bentler, 1999) Values less than 0 .08 for the RMSEA indicate a reasonable fit, and RMSEA values less than 0 .05 indicate good fit (Hu & Bentler, 1999). Values less than 0 .05 SRMR also indicate a good fit. Figure 41 Hypothesized Hurricane Preparation Decision Model Risk Perception Hazard Awareness Motivational Relevance Perceived Control over Outcomes Self Efficacy Action Coping Individual Intentions Strong Network Weak Network Hurricane Preparation Stressors Coping Self Assessment Intention to Preparation Link Moderator Response Efficacy Personal Responsibility
86 Table 4 7. Results of path analysis in SEM of the hypothesized hurricane preparation decision model used to assess hurricane preparedness and factors infl uencing preparedness in 153 households in Escambia County, Florida, 2009 Model 2 df CFI TLI SRMR RMSEA Hypothesized Mod el 55.77* 16 .50 .2 2 .11 .13 (.09 .17) Model w/Vulnerability Model w/ Pers Res 12.96 8.62 7 7 .88 .96 .76 .93 .06 .05 .08 (.00 .14) .04 (.00 .11) p < .001 Note. All values are rounded to two decimal places. CFI = Comparative Fit Index; TLI = Tucker Lewis Index; SRMR = Standardized Root Mean Square Residual; RMSEA = Root Mean Square Error of Approximation. A 90% Confidence Interval is presented for the RMSEA (Kline, 2005). In order to test for moderation in structural equation modeling, samples must be split according to the hypothesized moderator. The model is then tested across both samples. In the present study, splitting the sample across any of the hypothesized moderators would result in a sample that is too small for this form of analysis. Therefore, moderation analyses were conducted for hypothesized pathways in a multiple regression framework (Baron & Kenny, 1986). Specifically, three regression analyses were conducted to examine the hyp othesized moderators (response efficacy, past experience, and perceived r esponsibility). Results of the hierarchical regression analyses revealed that hurricane p reparedness w p < .01), F (1, 151) = 9.74, p < 01), accounting for 6.1 percent of the variance of hazard preparedness. However, the addition of interaction terms with r es ponse e fficacy or perceived r esponsibility did not add significantly to the model. Therefore, there was no evidence that these variables moderate the relationship bet ween individual intentions and hurricane preparedness.
87 Vulnerability (0 = nonvulnerable; 1 = vulnerable) and personal responsibility were e xamined as predictors to the final model. In both cases, the model fit decreases and the pathways were not significant (vulnerability, p = 0 .35; personal responsibility, p = 0 .30) (Table 4 7) T he hypothesized hurricane preparation decision model di d not fit the data well (Table 47). Based on the correlation and structural equation model analyses, the data do not support the hypothesis that causal pathways found in Patons social -cognitive model for earthquake preparation decisions will be the same for the hurricane preparation decision process H2: Strong and weak social networks added as a link between intentions and hurricane preparations increase the explanatory power of the model. My model posits a strong and weak social network link between individual intentions and hurricane preparation (Fig 41). As noted previously, items under the strong and weak social network variables were recoded due to a large amount of missing data. The results reflected a significant correlation between weak social networks and individual intentions in the combined, nonvulnerable and vulnerable sample (Table 4 4, 4-5 and 46). Only the non vulnerable sample reflected a significant correlation between strong social network and individual intentions (corr. = .27, p < .01). Finally, there was no correlation in any of the samples between strong and weak social networks and hurricane preparation. The results show that strong and weak social networks do not add to the explanatory power of the model. This will be dis cussed further in Chapter 5. H3: Vulnerable and non -vulnerable populations differ in their hurricane preparation decision process.
88 T he correlation analysis supports a more complex model in the combined and nonvulnerable samples (Table 44 and 4-5) Ma ny of the linkages in the model are reflected in the correlations. Risk perception was significantly correlated with perceived control and individual intentions. Motivational relevance was also significantly correlated with perceived control. Although n ot reflected in the model, perceived control was directly correlated with individual intentions and, as reflected in the model, with action coping. Only t he nonvulnerable sample show ed a correlation between self efficacy and perceived control as modeled. Individual intentions were significantly correlated with hurricane preparations in the combined and nonvulnerable samples. Strong and weak social networks were not significantly correlated with individual intentions or disaster preparations in the combined and nonvulnerable samples. The moderator variable, response efficacy, was significantly correlated with the outcome variable, hurricane preparations in the combined and nonvulnerable sample. The results of the correlation analyses support a much le ss complex model for the vulnerable population. There were fewer significant correlations and some of the correlations were different than my model. For example, there was a correlation between motivation relevance and weak network ties, not perceived control. Similar to the model, there were correlations between perceived control, action coping, individual intentions, and weak ties. The only correlation linked to hurricane preparation was response efficacy. The results reflect that vulnerable and nonvulnerable populations differ in their hurricane preparation decision process.
89 Post hoc Modification to the Hypothesized Model Using structural equation modeling, t he hypothesized hurricane preparation decision model was modified using post hoc modification indices. Two sensible modifications were indicated. Specifically, it was sugge sted that a path be added from risk perception to individual intention and from perceived control to hurricane p reparation. These paths are consistent with the over all model in that the assumed directionality is consistent. The only difference is that some of the hypothesized mediational pathways are not supported fully. The revised model still did not adequately fit the data. While no additional modification indices were plausible, as further revisions were made by removing three variables that did not have significant relations with any other var iables in the previous models (self efficacy, hazard awareness, and motivational r elevance). With these variables removed, the model fit the data well, and 2 diff (12) = 66.61, p < .001. This revised model is presented in Figure 4 2. Standardized and unstandardized parameter estimates5 with standard errors are presented in Table 4-8 5 Standardized estimates are computed with standardized variables. Standardized variables are variables that have been transformed so that its mean is 0 and its standard deviation is 1.0. The most common method to standardize a variable is by converting t he raw score to a z score. Unstandardized estimates are derived with unstandardized variables, that is, variables in their original units (scale) rather than expressed as z scores (Kline, 2005)
90 Tab le 47. Results of path analysis in SEM of the hypothesized hurricane preparation decision model used to assess hurricane preparedness and factors influencing preparedness in 153 households in Escambia County, Florida, 2009 Model 2 df CFI TLI SRMR RMSEA Hypothesized Mod el 55.77* 16 .50 .2 2 .11 .13 (.09 .17) Revision 1 4 1.25* 14 .6 6 .3 7 .0 7 .11 (.07 .15 ) Revision 2 2.93 4 1.00 1.00 .03 .00 (.00 .11 ) p < .001 Note. All values are rounded to two decimal places. CFI = Comparative Fit Index; TLI = Tucker Lewis Index; SRMR = Standardized Root Mean Square Residual; RMSEA = Root Mean Square Error of Approximation. A 90% Confidence Interval is presented for the RMSEA (Kline, 2005). Figure 42 Revised Hurricane Preparation Decision Model .21** .17* .13 .31* .25* .22* Risk Perception Perceived Control over Outcomes Action Coping Individual Intentions Hurricane Preparation Stressors Coping Self Assessment Intention to Preparation Link
91 Table 4 8. Significance Levels and Unstandardized and Standardized Estimates and Correlations for the Exercise Model Parameter Estimate Unstandardized Standardized Correlations 6 Structural Model Indiv Intent Hurr. Prep .46 (.17)** .21 .25** Perceived Cont Hurr. Prep .26 (.16) .13 .17* Action Coping Indiv Intent .21 (.31 )* .17 .21** Risk Percept Indiv Intent .31 (.10)** .22 .25** Perceived Cont Action Coping .23 (.06 )** .31 .30* Risk Percept Perceived Cont .31 (.11 )** .22 .29** Note. 2 (4) = 2.93, p = ns CFI = 1.00, TLI = 1.00, SRMR = .03, RMSEA = .00 (CI90% = .00 .11 ). p < .05, ** p < .01 Binary Logistic Regression A binary logistic regression analysis was conducted to determine if the model could adequately predict membership in the vulnerability category. This served as an alternative way to compare the two populations given the limitation imposed by sample size Logistic binary regression carries fewer assumptions and multivariate normality or homogeneity of variance -covariance are not required. Additionally, logistic regression can incorporate a large number of variables, even when the sample is relatively small (Kinnear & Gray, 2010). The analysis showed a 2 Log likelihood of 129.04. This statistic is similar to the chi -square statistic in that a large value indicates that the regression model fits the d ata poorly. The omnibus test of model coefficients s hows the regression model improves significantly in predicting category membership as all the p values are 0 .001. The success rate of correct category assignments when the regression model has been applied to the data improved from 69% (Block 0absence of info about regression) to 6 The SEM unstandardized estimates reflect indirect effects. The correlation estimates reflect direct effects.
92 77.7%. All variables were used in the regression model. Using the enter method, it was found that only self efficacy (.04), action coping (.03) and strong social networks (0 .00) were significant at p < 0.05. At p < 0.10, perceived control (.07) and weak social network ( 0 .09) are significant. Additional Analyses of Full Sample T o incorporate some cases with missing data, I conducted some analyses with the complete sample of 245 cases. I focused my analysis on the variables used in the revised model. Starting with the complete sample of 245 cases, I explored the data using the following variables: risk perception, perceived control, action coping, individual intent and hurricane preparation. The result was elimin ation of 18 cases that were found to be outliers for one or more of the variables. A t test for independent samples was conducted to evaluate differences between the vulnerable and nonvulnerable samples with regard to these variables. The only variable s that dif fered significantly were risk perception and individual intentions (p < 0.01). Means scores for the other variables were not significant. Table 4 9. Groups statistics and independent samples test for correlation coefficients between 5 variables used to assess hurricane preparedness and factors influencing preparedness in 220 households in Escambia County, Florida, 2009 Variables N Mean Std Dev t Test for Equality of Means NV = Nonvulnerable; V = Vulnerable t Sig. (2 tailed) Outcome NV V 118 106 2.89 3.04 .89 1.02 1.15 .25 Risk Per NV V 110 93 3.76 4. 03 .67 .4 6 3.43 .00 Perc Con NV V 107 88 3.8 5 3.9 6 .52 60 1.52 .13 Action Cope NV V 114 100 3.61 3.5 7 .51 54 62 .53
93 Ind Int NV V 115 101 1.72 1.9 8 .54 .5 6 3.52 .00 Summary This chapter presented the analysis and results of the study. First, reliability tests were conducted post hoc on all instruments. Indicators were eliminated from risk perception, motivational relevance and perceived control. Additionally, the moderator v ariable of personal responsibility was deleted from the model because the reliability scores were too low. Initially, it was found that the primary statistical analysis would be limited due to size of the useable sample (n = 154). Based on the independent samples t -test conducted on all variables, equal variances were assumed as significance levels were close to or greater than 0 .05. Therefore, the sample could be treated as one population. The research for this study is based on comparing a vulnerable population to a non vulnerable population. Comparing two groups would be difficult as the sample sizes were even smaller (nonvulnerable = 53 and vulnerable = 100). However, in an effort to delineate any differences, if possible, a descriptive and correlation analysis was conducted to analyze the 11 different variables for each population. Structural equation model analyses were also conducted to test hypotheses one and two. T he correlation and structural equation model results reflected a similarity between the hypothesized model and Patons social -cognitive model. However, the results only reflected similarities. Overall, the results do not support the hypothesis that causa l pathways found in Patons social -cognitive model for earthquake preparation
94 decisions will be the same for the hurricane preparation decision process. Additionally, the results did not support the hypothesis that strong and social networks add explanatory power to the model. Finally, t he correlation analysis showed little difference between the combined sample of 153 and the nonvulnerable sample of 100. The correlation analysis reflected a more complex model in the combined and non vulnerable samples. Many of the linkages in the model are reflected in the correlations. The results of the correlation analyses support a much less complex model for the vulnerable population. There were fewer significant correlations and some of the correlations were different than my model. Therefore, the hypothesis that v ulnerable and non vulnerable populations differ in their hurricane preparation decision process is supported by the data. The binary logistic regression analysis indicated that the model did not fit the data well. This was also the case with the structural equation model. On the other hand, the binary logistic regression model indicated that by examining the variables of self -efficacy, action coping, and strong and weak social networks, one could s uccessfully categorize individuals (vulnerable vs nonvulnerable) at a 77% rate. Finally, a few additional analyses were conducted with the complete sample. Recognizing that one cannot draw too many conclusions from the t -test and correlation analysis, there are still similarities between the full sample (n=245) and the smaller sample (n = 153). Discussion of the results of the different analyses and their policy implications will be presented in the next chapter.
95 CHAPTER 5 DISCUSSION AND CONCL USION The purpose of this research was to explore hurricane preparation decisions by vulnerable and non vulnerable populations. This last chapter conta ins five sections. The first section discusses conclusions in regards to the s pecific hypotheses as stipulated in Chapter 2. The second section reviews the theoretical perspectives and how the research supports or corroborates the theories used in the research. The third section discusses the overarching conclusion of the research and policy implications The fourth section briefly discusses limitations of the study. Finally, the last section suggests future work and areas of focus for research. Discussion of Hypotheses H1: The causal pathways found in Patons social -cognitive m odel for earthquake preparation decisions will be the same for the hurricane preparation decision process. The hypothesized model indicates that individuals decide to prepare based on a sequence of social cognitive activities (see Figure 4-1). Paton (2003) argues that motivators or precursors lead to intention formation that eventually results in preparation actions. The hypothesis was not supported by the data. The same causal pathways in Pato ns (2003) model were not evident in my hypothesized model. However, further analysis of the model reflected similarities in variables and directionality. Numerous variables were eliminated during the analysis, but the foundational concept that social cognitive factors lead to individual intentions that result in hurricane preparations was proven to be correct. The low number of cases analyzed for the complex hypothesized model may have been a factor in eliminating many variables. The revised model shows that two stressor variables, hazard awareness and
96 mot ivational relevance, were eliminated These are two variables that Patons (2005) earthquake research indicated are important. Similar to my revised model, Patons (2005) model also shows that control and coping are factors that lead to individual intentions. Finally, intention to prepare shows a link to hurricane preparations. Analysis failed to confirm moderat ing roles for response efficacy and personal responsibility. Figure 51. Revised Model. Findings from this study indicate that there are three directional paths to disaster preparation (Fig. 5 1) In the first path individuals go from risk perception directly to individual inte ntions with the result being hurricane preparations. A second path indicates risk perception leads to perceived control resulting in hurricane preparations. In t he final path individuals go from risk perception to perceived control to individual intenti ons through action coping. This model is much less complex than my hypothesized model. The simpler model excludes a number of cognitive variables normally found in decision making .21** .17* .13 .31* .25* .22* Risk Perception Perceived Control over Outcomes Action Coping Individual Intentions Hurricane Preparation Stressors Coping Self Assessment Intention to Preparation Link
97 theories, such as the theory of planned behavior or the theory of reasoned action. Many de cisionmaking theories posit that individuals weigh their knowledge, abilities and resources against the strength of the event (Basolo et al., 2009; Lindell & Whitney, 2000). As Basolo et al. state individuals perform a cognitive calcu lus assessing qualities about themselves and attributes of the hazard adjustments or actions to be taken. Attitudes about the actions to take are more influential than the perception of the hazard (Basolo et al., 2009). Although this cognitive calculus is not evident in the revised model, it is evident in the correlation analyses. H2: Strong and weak social networks added as a link between intentions and hurricane preparations increase the explanatory power of the model. It was my assumption that vulnerable populations would utilize social networks (strong or weak) in hurricane preparation actions. Unfortunately, due to the amount of missing data, I was unable to analyze these variables as fully as I would have liked. They were not used in the s tructural equation model analysis for this reason. However, I made an attempt to glean some information from the data I had in regards to social networks. As noted in Chapter 4 I recoded missing data on surveys where some but not all items had been chec ked in the weak and strong social network questions. If some had been checked and others were blank, I recoded the blanks as never. I assumed that many respondents chose to select only items that applied to them. If all items were left blank, the ques tionnaire was not used in the analysis. None of the samples showed a correlation between strong and weak social networks and disaster preparation. However, there were strong correlations between
98 weak social networks and individual inten tions in the combi ned, non vulnerable and vulnerable sample. Strong social networks were classified as family, friends, neighbors and coworkers. The nonvulnerable group showed a significantly higher mean score in the use of strong social netwo rks for hurricane preparati on. T he vulnerable population showed mean scores of 3.1 for family and 2.9 for friends and the non vulnerable had mean scores of 3.8 for family and 3.3 for friends. Weak social networks were classified as churches, local law enforcement, local emergency offices, social clubs, TV, radio and county extension. A review of the responses indicates that between 70 and 80 percent of the vulnerable and nonvulnerable populations use TV or radio almost always or always to obtain hurricane information. The mean s core (scale of 1 -5) for the vulnerable population for use of TV was 4.58 and 4.35 for radio. The non vulnerable population had scores of 4.32 for TV and 4.06 for radio. The local emergency management office had scores of 3.54 for the vulnerable populati on and 3.41 for the non vulnerable population. The hypothesis that adding strong and weak social networks as a link between intentions and hurricane preparations would increase the explanatory power of the model was not supported by the data. The overwhel ming majority (93%) of the respondents in the useable sample believe they are personally responsible for preparing for a hurricane. They also agree that the government is not responsible for making sure they are prepared. A factor that may have influence d this result in Escambia County is the increased amount of information that the county may be disseminating as part of their emergency planning. A cursory review of various
99 websites for Escambia County indicates that the county is taking aggressive action to inform residents about hurricane preparations. For example, the creation of BRACE was a county decision. The BRACEs statement of purpose is primarily to engage the community in disaster preparedness, response and recovery (BRACE, 2009). In fact the Countys emergency management plan states, BRACE will coordinate social and support services utilizing local, donated and purchased resources from its membership and the community abroad to meet individual unmet needs and manage and maintain social service casework where appropriate (BRACE, 2009). The results o f the correlation analysis may indicate that the strong and weak social networks should fall under coping self assessment and not intention to prepare. Eisenman, et al.s (2007) research on disaster planning and risk communication in the wake of Hurricane Katrina showed that social networks (family, friends and neighbors) influenced decision making in regards to evacuation. The correlations between social networks and the coping self asses sment mediator variables in my model would support the idea that social networks play a more important role in deciding to prepare rather than actual preparation actions. The vulnerable sample only showed correlations with the weak social network variabl e. As stated previously, weak social networks included indicators such as media outlets. Media outlets (TV and radio) are generally used extensively in the dissemination of hurricane preparation information. Therefore, the weak social network appears to have played a predominant role for the vulnerable sample. Eisenman et al. (2007) found that individuals tend to integrate media messages with information from family and friends. I conducted a correlation analysis combining
100 the indicators from both my strong and weak social networks as that would combine family and friends with media to determine if social networks would be correlated with hurricane preparations. Once again, there was no correlation between social network and hurricane preparation. Th e same correlations appeared when using separate variables (stong and weak) and in the analysis combining the two. This would support the conclusion that strong and weak social networks are more useful in the coping self assessment phase. H3: Vulnerab le and non -vulnerable populations differ in their hurricane preparation decision process. Hurricane preparedness was determined by calculating a score based on responses to a yes/no hurricane checklist from the Ready.gov website. Based on a total possib le score of 31 for the outcome variable, the mean scores were 22.31 for the nonvulnerable sample and 21.0 for the vulnerable sample. After normalizing responses by using the squar e root adjustment, there is no significance between the vulnerable and non vulnerable populations in regards to hurricane preparations. Several factors support this assumption. First, 96 percent of all respondents have experienced at least one hurricane. In fact, 80 percent have experienced more than one hurricane. Addition ally, 74 percent of respondents experienced loss or damage from a hurricane. Lindell and Whitney (2000) found in their study about seismic hazard adjustment that increasing peoples knowledge about a hazard will positively affect their adjustment to the hazard. The findings in my research support this idea. If one has experienced a hurricane, one knows what to expect and thus might be encouraged to seek information about how to prepare. Respondents in my study were asked questions regarding the
101 definit ion of hurricane watch versus warning and how hurricanes were classified. Over 94 percent knew that hurricanes were classified based on wind speed and 67 percent knew that a hurricane watch meant that a hurricane could hit within 36 hours. Interestingly, only 38 percent knew that a hurricane warning meant a hurricane could hit within 24 hours. Another supporting factor that may have influenced preparation levels is that Escambia County has an aggressive hurricane preparation program. Since Hurricane Ivan struck Escambia County in 2004, county leadership has stressed the importance of identifying vulnerable populations so appropriate hurricane preparations and response can be planned. As a result, they have created BRACE. This is an umbrella organization that has 280 partners that work together in disaster preparation planning. T hey have a database of over 57,000 individual vulnerable households/members that they can track in times of emergencies. Another factor that may have influenced this particular outcome is the timing that the questionnaire was administered. The questionnaire was administered in September and October. This was mid-hurricane season. Therefore, it could be assumed that a large amount of preparation information was most likely being publicized. Additionally, shortly before this questionnaire was mailed, Tropical Storm Claudette struck Escambia County and the surrounding areas. Thus, any preparation actions taken for Tropical Storm Claudette may have still been in place when the questionnaire was mailed. The literature on disaster preparation and vulnerable communities generally indicates that vulnerable populations are less inclined to prepare or have the resources to prepare for hurricanes. My research shows this is not the case. My model indicates
102 that response efficacy will serve as a moderator variable between individual intentions and hurricane preparation. All samples (combined, nonvulnerable and vulnerable) showed a significant correlation between response efficacy and hur ricane preparation. However, there was no correlation between individual intention and response efficacy. An interesting observation is that the combined and vulnerable sample indicated a significant correlation between self efficacy and response efficac y. This would be consistent with the personrelative -to event (PrE) theory that postulates that self efficacy refers to self assessments of ones knowledge, skill, ability, energy, and financial resources in relation to a hazard event (Lindell and Whitn ey, 2000, p. 14). Indicators for self -efficacy were based on generalized notions of how one might deal with problems faced in everyday life. Response efficacy indicators queried respondents in regards to specific hurricane preparations. Respondents were asked to indicate to what degree did cost, skill or knowledge required, time, other things to think about, and need for cooperation, prevent them from preparing for a hurricane. These indicators are the same indicators used in the PrE theory. Bandura (19 89) argues that self efficacy, a construct of social cognitive theory, is not a global personality trait but is specific to a given behavior (Wenzel et al., 2002). Based on the correlations reflected in the combined and vulnerable sample, one might conc lude that the results reflect a high self efficacy for hurricane preparation not necessarily self -efficacy for other behaviors. The high self efficacy could be related to past experience. Although the variable of past experience was eliminated from analy sis due to poor reliability concerns, researchers posit that individuals who have experienced
103 a natural disaster are more likely to prepare, even more so if it was a negative experience (Basolo, et al., 2009; Mileti, 1999). The correlation analyses for the combined and nonvulnerable sample support the decision making theories in regards to the number of cognitive variables used in the process. However, the vulnerable sample reflected something very different. The correlati on analysis for the vulnerable sample fell more in line with the revised model in that there were, by far, the fewest amount of correlations between the variables. It appears the vulnerable population is not inclined to utilize all the cognitive processes in deciding to prepare or not prepare. Based on the correlation analysis, the vulnerable sample appears to react by preparing rather than contemplate what would happen if they did not prepare. I tentatively conclude that the vulnerable population was responding to media messages (weak soc ial network) disseminated by the local government (emergency planners). Theoretical Retrospective Social Cognitive Theory An analysis of the revised model indicates that the theories used in the hypothesized model formulation were appropriate. Social cognitive theory addresses the psychosocial constructs that explain why people make certain choices. Individuals decide to pr epare or not prepare for hurricanes. In discussing social cognitive theory, Bandura (2001) refers to core features of human agency as intentionality, forethought, self -reactiveness and self -reflectiveness. The hypothesized and revised models show a consi stent directionality that supports human agency features of intentionality and forethought. Intentionality refers to human actions that may involve certain inducements. As the model indicates, the inducements of risk perception, hazard
104 awareness and moti vational relevance (referred to as stressors) would lead an individual to intentions. Analysis of the model indicates that the stressor of risk perception is, in fact, an inducement to intention to act. Bandura (2001) argues that intentions also center on plans of action. The construct of action coping as shown in the model supports Banduras argument in that action coping is linked to individual intentions. Action coping is also referred to as problem -solving coping. Respondents were asked questions such as to the extent to which they make plans of action, develop strategies, decide what steps to take and decide how best to handle a problem in dealing with everyday life problems. These questions get at the heart of intentionality as described by Band ura (2001). Forethought provides direction, coherence and meaning (Bandura, 2001). Individuals anticipate consequences of future events and decide on courses of action accordingly. Bandura (2001) argues that future events cannot be a motivator because t hey do not actually exist However, if represented cognitively in the present, future events can motivate and direct behavior This would be the case with hurricanes. Hurricanes are presented in the present in many ways. For example, hurricane season runs from June through November. People know that th is is the most likely time for a hurricane to occur. Emergency planners publicize hurricane preparation information in the spring to encourage households to start preparation action. In fact, hurricane projections issued in De cember for the following year are an attempt to cognitively represent the threat of a hurricane. T his cogni tive representation would be an impetus for individual disaster preparation behavior.
105 Stress and Coping The t ransacti onal model of stress and c oping identifies cognitive appraisal and coping as factors that influence future outcomes for individuals (Wenzel et al, 2002; Folkman et al, 1986). The theory links coping process es with stressful events. The t ransactional model of s tress and coping starts with a medi ating process that identifies stressors. These stressors are viewed through primary appraisal and secondary appraisal. The primary appraisal is the initial evaluation of the event. In my research, the event would be the hurricane. As stated previously, these future occurring events are represented cognitively through various forms of media and information dissemination. My hypothesized model used variables such as risk perception, hazard awareness and motivational relevance as stressors. These are factors found in the primary appraisal of the stress and coping model. Secondary appraisal determines what actions can be taken to improve the outcome of the event or encounter (Folkman et al, 1986). I incorporated s eco ndary appraisal factors like perceived control over outcomes and self efficacy in my m odel. My revised model shows that primary and secondary ap praisal factors (stress ors) are linked to intentions and consistent with the theory of stress and coping. However, self -efficacy was not a significant factor in my revised model. Perceived control over outcomes was linked directly with action coping and hurricane preparations. My research attempted to compare a vulnerable population to a non vulnerable popula tion with regard to the hypothesized model. Stress literature indicates that vulnerable populations may be more susceptible or reactive to stressors (Thoits, 1995). However, ot her research indicates that stress ors differ and that the particular event
106 det ermines whether a vulnerable population reacts differently (Thoits, 1995 and 1987). My research does not support the proposition that vulnerable populations respond differently to stressors as they relate to hurri canes. T he stressor of a future hurricane event differs greatly from the types of events in the stress resear ch. Much of the research about stress deals with health and well -being issues. The refore, it is difficult to generalize that vulnerable populations respond differently without further re search about different types of stressors. Social Network Theory Social interactions have always been at the heart of sociological inquiry and social networks provide the mechanism for social interactions (Pescosolido, 2006 and 1992). Pescosolido (1992) argues that interaction is a primary element in the decisionmaking process. Granovetter (1985, p. 486) states that the approach to social relations has the paradoxical effect of preserving atomized decision making even when decisions are seen to involve more than one individual. The exploration of social interactions can be traced to Georg Simmels (1955) work Conflict and the Web of Group Affiliations. Simmel (1955, p. 163) starts by stating, Society arises from the individual and the indivi dual arises out of association. The strong and weak networks that Granovetter (1982) refers to can be traced to Simmels ideas of organic and rational motivations for joining groups (Allen, 2007). Organic motivations are based on natural connections s uch as family. Rational motivations are based on choice. Allen (2007) uses the terms of primary and secondary group and Granovetter (1982) uses the term of strong and weak networks. My research examined the variables of strong and weak networks and how they impact hurricane preparations. My research showed minimal to no correlation between
107 soci al networks (weak or strong) and hurricane preparation. There are factors that have produc ed this result. First, the number of useable surveys may not have been large enough to present a statistically strong analysis. Second, my choices of what constituted a strong or weak network may have been flawed. While family and friends can easily be considered a strong network, it is not as clear how TV and radio shoul d be classified. Should TV and radio be considered part of an individuals network or are they simply resources used for information dissemination? I believe further research in this regard is warranted. Finally, my hypothesized model had social networ ks situated as a possible link between individual intentions and hurricane preparations. Research in stress and coping literature discusses social support instead of social networks. Thoits (1986) argues that coping and social support have similar functi ons they are both directed at changing or managing stressful situations. Reconceptualizing social support as coping assistance would warrant relocating the social network or social support construct in the model to precede individual intentions. Again, I believe future research in this area would be beneficial to understanding how vulnerable communities use social support systems in disaster preparation. Policy Implications This study has shown that vulnerable communities choose to prepare for disaster a t the same level as nonvulnerable communities. The only difference between the two communities is that nonvulnerable communities use a more complex cognitive analysis before deciding to prepare. The vulnerable community, on the other hand, does not employ a complex cognitive analysis and appears to react based on hazard awareness. The extent to which vulnerable communities r eceive information in about natural
108 hazards or hurricanes influences their decisions to prepare. Therefore, a program that includ es extensive information dissemination to all residents would be useful to ensure vulnerable and non vulnerable populations prepare. The State of Florida Comprehensive Emergency Management Plan 2004 (FL CEMP 2004) directs counties to implement a broad based public awareness, education and information program designed to reach all citizens of the county Hayek (1945) discusses how certain individuals in society may have an advantage over others as they may have more information that would be beneficial in making decisions. This theory is applicable to the study of vulnerable communities and hurricane preparations. Factors that define a vulnerable community such as age, income, and disability are often the same factors that limit the amount of informat ion provided or, if provided, understood by the vulnerable populations. Therefore, the goal for any hurricane preparedness policy should be that every citizen of the state of Florida has equal access to information that could improve their preparation act ions for hazard events. The current statutes and plans make no distinction for vulnerable populations. Current policies only address special needs populations such as disabled, visual or hearing impaired (FL CEMP 2004). The lack of addressing other vuln erable populations in emergency management plans may result in a large group of citizens not being prepared for hazard events. Yet, my study and other studies have shown that vulnerable populations will take preparations prior to the onset of a disaster event if the population is aware of what needs to be done (Tobin et al., 2006). Escambia County has a proactive emergency preparation program and has identified over 50,000 local residents that are classified as vulnerable. Personal and
109 contact information for each of these residents is maintained in a database. Working through service and support agencies and organizations, the county is able to ensure residents receive the information and support they need to prepare for hurricanes. I believe the ef forts the county has taken to ensure increased preparedness has increased the level of awareness by its vulnerable populations with the result of a higher preparedness level by its vulnerable residents. This study shows that vulnerable populations are abl e, willing and do prepare for hurricanes if they are aware of the hazard. The State of Florida should consider any model or program that improves dissemination of hurricane preparation information to vulnerable populations. Escambia County may serve as a model as they have found a way to identify vulnerable residents and are using supporting service agencies and organizations to channel the information to the residents they serve. Limitations of the Study The primary limitation for this study was the number of useable cases for analysis. Although 245 questionnaires were returned from a mail out of 1500, only 153 cases were useable due to missing data. Funding constraints limited the mail out to 1500 residents. For the complex model I was using and t he fact that I was attempting to obtain a large enough sample from the vulnerable population, I believe it would have taken another 3000 questionnaires in order to obtain a statistically powerful sample. Another possibility, had funding been available, would have been to conduct the questionnaire via phone or in person to ensure all data was obtained. Future Research I believe that my research and other research have shown there is a directional link between risk perception and individual intentions resu lting in disaster preparations.
110 Unfortunately, my research provided little evidence as to the impact of social networks on hurricane preparation decisions. This is an area that I believe should be examined further, especially in regards to vulnerable pop ulations. There has been an abundant amount of research conducted examining how social support impacts coping with stressful situations. However, most of this research has been done in the health behavior areas. It would be beneficial and constructive t o examine social support in regards to coping with disaster. Some questions for consideration in future research: 1. What does vulnerability really mean? 2. What degree does vulnerability and social isolation correlate? 3. Self -helpwould it work with vulnerable populations? 4. How does one create conditions for empowering vulnerable populations to take disaster preparation actions?
111 APPENDIX A LIST OF VARIABLES AN D SCALAR RESPONSE RANGE Variables Items Scalar Response Range Risk Perception A hurricane could pose a threat to your personal safety A hurricane could pose a threat to your daily life A hurricane could pose a threat to your property The most likely time within which a damaging hurricane could affect me is: this year; next year; within 3 years; within 5 years 1 5 Strongly disagree to strongly agree Hazard Awareness Please describe how much you: think about hurricanes talk about hurricanes 1 5 Not at all to a great deal Motivational Relevance I get nervous when there is discussion about approaching hurricanes When hurricane ads come on TV, I change the channel or dont pay attention I avoid things that remind me of hurricanes If I believe a hurricane is approaching, I make sure I know the evacuation route If I believe a hurricane is approaching, I make sure I know where my shelter is I avoid thinking about hurri canes 1 5 Not at all to a great deal Perceived Control Hurricanes are too destructive to bother preparing for A serious hurricane is unlikely to occur during your lifetime Preparing for hurricanes will reduce damage to my home should a hurricane occur Preparing for a hurricane will improve my everyday living conditions Preparing for a hurricane will improve the values of my house/property Preparing for hurricanes will reduce the disruption to family/community life following a hurricane Preparing for hurricanes is a hassle for me 1 5 Strongly disagree to strongly agree Self Efficacy I have considerable control over what happens in my life I can solve most of the problems I have by myself What happens to me in the future mostly depends on me I can do a lot to change many of the important things in my life I can do just about anything if I really set my mind on it I rarely feel helpless in dealing with the problems of my life 1 5 Strongly disagree to strongly agree Action Coping I try to come up with a strategy about what do do I make a plan of action I think hard about what steps to take I think about how I might best handle the problem 1 4 I usually dont do this at all to I usually do this a lot Individual Intention Respond to the following with regards to hurricane preparation actions: --Check your level of preparedness for hurricanes --Increase your level of 1 3 No, possibly, definitely
112 preparedness for hurricanes --Become involved with a local group to discuss how to reduce hurricane damage --Seek information on hurricane risk --Seek information on things to do to prepare Strong Social Network Where do you look to get information about how to prepare for hurricanes? --Family --Friends --Neighbors --Co workers Who do you turn to when you need to prepare for a hurricane? --Family --Friends --Neighbors --Co workers 1 5 Never to always Weak Social Network Where do you look to get information about how to prepare for hurricanes? --Churches --Local emergency management ofc --Local law enforcement --Social clubs --County Cooperative Extension --TV --Radio Who do you turn to when you need to prepare for a hurricane? --Churches --Local emergency management ofc --Local law enforcement --Social clubs --County Cooperative Extension --TV --Radio 1 5 Never to always Response Efficacy To what extent might each of the following prevent you preparing for hurricanes? --The cost --The skill or knowledge required --Time to do them --Other things to think about --Need for cooperation with others 1 5 Not at all to a great deal Personal Responsibility I feel responsible for preparing for a hurricane The local government is responsible for making sure that I am prepared for a hurricane 1 5 Strongly disagree to strongly agree Past Experience Have you been in a hurricane in the last 10 years If yes, what year(s) did this occur? If yes, did you experience damage or loss (i.e., requiring repairs/insurance claims)? -Nominal responses Hurricane Prep A 31 item checklist obtained from ready.gov website 0,1 No or Yes
113 APPENDIX B PERCENT POVERTY BY F IRE DISTRICT ESCAMBIA COUNTY
114 APPENDIX C 65+ POPULATIONS ESCAMBIA COUNTY
115 APPENDIX D SPECIAL NEEDS REGIST RANTS ESCAMBIA COUNTY
116 APPENDIX E ZIP CODE COMPARISON
117 APPENDIX F Escambia County Zip Codes
118 APPENDIX G HURRICANE PREPARATIO N QUESTIONNAIRE
124 LIST OF REFERENCES Adcock, R. and Collier, D (2001). Measurement validity: a shared standard for qualitative and quantitative research. American Political Science Review 93(3), 529-546. Allan, K. (2007). The Social Lens: A n Invitation to Social and Socio logical Theory. Thousand Oaks, CA: SAGE Publications. Agresti, A. and Finlay, B. (1997). Statistical Methods for the Social Sciences, 3rd Ed. New Jersey: Prentice Hall. American Red Cross Report (2005). Th e Aftershock of Katrina and Rita: Public Not Moved to Prepare. Prepared by Peter D. Hart Research and Public Opinion Strategies for the Council for Excellence in Government and the American Red Cross. Aspinwall, L.G. (2003). Proactive coping, well -being, and health. In The International Encyclopedia of the Social and Behavioral Sciences, ed. NJ Smelser, PB Baltes. Oxford: Elsevier. Aspinwall, L.G. and Taylor, S.E. (1997). A stitch in time: self -regulation and proactive coping. Psychological Bulletin 121, 417436. Bagozzi, R.P. and Dabholar P.A. (2000). Discursive psychology: an alternative conceptural f oundation to means end chain theory. Psychology and Marketing 17(7), 535586. Ballantyne, M., Paton, D., Johnston, D., Kozuch, M. and Daly, M. (2000). Information on Volcanic and Earthquake Hazards: The impact on awareness and preparation. I nstitute of Geological and Nuclear Sciences Limited Science Report No. 2000/2, Wellington. Bandura, A. (1989). Human Agency in Social Cognitive Theory. American Psychologist 44(9), 11751184. Bandura, A. (2001). Social Cognitive Theory: An Agentic Perspective. Annual Review of Psychology 52, 1 -26. Baranowski, T., Perry C. L., and Parcel G. S. (2002). How individuals, environments, and heal th behavior interact: social cognitive theory. In K. Glanz, B. K. Rimer and F. M. Lewis (Eds.), Health Behavior and Health Education Theory, Research and Practice, 3rd Edition (pp. 165-184). San Francisco, CA: Jossey -Bass.
125 Basolo, V., Steinberg, L.J., Burby, R.J., Levine, J, Cruz, A.M. and Huang, C. (2009). The effects o f confidence in government and information on perceived and actual preparedness f or disasters. Environment and Behavior, 41, 338-364. Blaikie, P., Cannon T., Davis I., and Wisner B (1994). At risk: natural hazards, peoples vulnerability, and disasters. New York: Routledge. Blake, E.S., Jarrell, J.D., and Rappaport E.N. (2006). The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2005 (and Other Frequentl y Requested Hurricane Facts). NOAA Technical Memorandum NWS TPC -4. National Hurricane Center, Miami, FL. Blake, E.S., Rappaport, E.N., and Landsea, C.W. (2007). The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2005 (and Other Frequently Requested Hurricane Facts). NOAA Technical Memorandum NWS TPC -5. National Hurricane Center, Miami, FL. Blau, Peter (1974). Parameters of social structure. Amer ican Sociological Review 39(5), 615-635. Bolin, Bob (2006). Race, class, ethnicity, and disaster vulnerability (pp. 113-129) In Handbook of Disaster Research. (Eds: Havidan Rodriguez, Enrico L. Quarantelli, and Russell R. Dynes), New York: Springer. BRACE (2009). What is BRACE? Our Mission, Purpose and Vision. Retrieved from http://www.bereadyalliance.org/what_is/ Bradshaw, Sarah (2004). Socio economic impacts of natural disasters: a gender analysis. Santiago, Chile: United Nations Publications. Bennett, P. and Murphy S. (1997). Psychology and Health Promotion. Buckingham: Open University Press. Bryman, Alan (2004). Social Research Methods, 2nd Ed. Oxford: University Press. Burger, J.M. and Palmer M. I. (1992). Changes in and generalizatio n of unrealistic optimism following experiences with stressful events: reactions to the 1989 California earthquake. Journal of Personality and Social Psychology, 18. 39 -43. Cannon, W.B. (1932). The Wisdom of the Body New York: Norton. Cannon,T. (1994) Vulnerability analysis and the explanation of natural disasters (pp. 1330) In Disasters, Development and Environment. (Ed: Ann Varley), New York: John Wiley & Sons.
126 Carr, L. (1932). Disaster and the sequence-pattern concept of social change. Americ an Journal of Sociology 38, 207-218. Carver, C.S., Scheier M.F., and Weintraub J.K. (1989). Assessing coping strategies: A theoretically based approach. Journal of Personality and Social Psychology, 56, 267-283. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. Comfort, L., Wisner B., Cutter S., Pulwarty R., Hewitt, K., Oliver Smith, A., Wiener, J., Fordham, M., Peacock W. and Krimgold F. (1999). Reframing disaster policy: the global evolution of vulnerable communities. Environmental Hazards 1, 39-44. Cutter, S. L. (2005). Are we asking the right question? (pp. 39-48) In What is a disaster? New answers to old questions (Eds: Ronald W. Perry and E. L. Quarantelli), US: International Research Committee on Disasters. Cutter, S. L., Brouff B.J., and Shirley W.L. (2003). Social vulnerability to environmental hazards, Social Science Quarterly 84(2), 242261. Cutter, S. L., Emrich, C. T., Mitchell J. T., Boruff, B. J., Gall, M., Schmidtlein, M.C., Burton, C. G., and Melton, G. (2006). The long road home: race, class, and recovery from Hurricane Katrina, Environment 48(2), 220. Cutter, S.L., Johnson, L.A., Finch, C., and Berry M. (2007). The US Hurricane Coasts: Increasingly Vulnerable? Environment 49(7), 8 -20. Dalton, J.H., Elias, M.J. and Wandersman, A. (2001) Community Psychology. Belmont, CA: Wadsworth. Darlington, R. ( 2006) Factor Analysis. [cited 3 May, 2007]. Available from internet site: http://comp9.psych.cornell.edu/darlington/factor.htm DeVaus, David (2001). Research Design in Social Research. London: SAGE Publications Ltd Dillman, Don A. (2007). Mail and Internet Surveys: The Tailored Design Method. Hoboken, New Jersey: John Wiley & Sons, Inc. Dombrowsky, Wolf R. (1998). Again an again: Is a disaster what we call a disaster? (pp. 19-30) In What is a disaster? Per spectives on the Question (Ed: E. L. Quarantelli), London: Routledge. Downing, Steven M. (2004). Reliability: on the reproducibility of assessment data. Medical Education, 38, 1006 -1012.
127 Duval, T. S. and Mulilis J.P., (1999). A person-relativeto event (PrE) approach to negative threat appeals and earthquake preparedness: a field study. Journal of Applied Social Psychology, 29, 495-516. Dynes, R. (1998). Coming to terms with community disasters (pp. 109-126) In What is a disaster? Perspectives on the Question (Ed: E. L. Quarantelli), London: Routledge. Ericksen, E, and Yancey W. (1977). The locus of strong ties. Unpublished manuscript, Department of Sociology, Temple University. Eisenman, D.P., Cordasco, K.M., Asch, S ., Golden, J.F., and Gilk, D. (2007). Disaster planning and risk communication with vulnerable communities: lessons from Hurricane Katrina. American Journal of Public Health 91(S1), 109115. Federal Emergency Management Agency (FEMA). (2003). The Robert T. Stafford Disaster Relief and Emergency Assistance Act, (U.S.C. 5121, et seq.) Washington, DC: Federal Emergency Management Agency. Federal Emergency Management Agency (FEMA) Emergency Preparedness Demonstration Project (2009). http://srdc.msstate.edu/fema/ Folkman, S. and Lazarus, R.S. (1980). An analysis of coping in a middleaged community sample. Journal of Health and Social Behavior 21, 219-239. Folkman, S., R.S. Lazarus, C. Dun kel -Schetter, A. DeLongis and R.J. Gruen (1986). Dynamics of a stressful encounter: cognitive appraisal, coping and encounter outcomes. Journal of Personality and Social Psychology 50(5), 9921003. Folkman, S. and Moskowitz, J. T. (2004). Coping: pitfal ls and promise. Annual Review of Psychology 55, 745774. Fothergill, A., Maestas, E.G.M., and Darlington J.D. (1999). Race, ethnicity and disasters in the United States: a review of the literature. Disasters 23(2), 156173. Gilbert, Claude (1998). Studying disaster: Changes in the main conceptual tools (pp. 1118), in What is a disaster? Perspectives on the Question (Ed: E. L. Quarantelli), London: Routledge. Granovetter, Mark S. (1973). The strength of weak ties. American Journal of Sociology 78(6), 13601380. Granovetter, Mark S. (1983). The strength of weak ties: a network theory revisited. Sociological Theory 1, 201-233.
128 Hair, J.F., Anderson, R.E., Tatham R.L. and Black W.C. (1995). Multivariate data analysis (4th Ed). Upper Saddle River, NJ: Prentice Hall. Hammersley, Martyn (1987). Some notes on the terms validity and reliability. British Educational Research Journal 13(1), 7381. Haque, E. E. and Etkin, D. (2007). People and community as constituent parts of hazards: the significance of societal dimensions in hazards analysis. Natural Hazards 41, 271 -282. Hayek, F. A. (1945). The use of knowledge in society. The American Economic Review 35(4), 519530. Heaney, C.A. and Israel ,B.A., (2002). Social networks and soc ial support. In K. Glanz, B. K. Rimer and F. M. Lewis (Eds.), Health Behavior and Health Education Theory, Research and Practice, 3rd Edition (pp. 185209). San Francisco, CA: Jossey -Bass. Hewitt, Kenneth (1998). Excluded perspectives in the social co nstruction of disaster (pp. 7591) in What is a disaster? Perspectives on the question (Ed: E. L. Quarantelli). New York: Routledge. Homans, George (1950). The Human Group. New York: Harcurt, Brace and World. House, J.S. (1981). Work Stress and Social Support. Reading, MA: Addison-Wesley. Johnst on, D. M., Bebbington, M.S., Lai, C.D., Houghton, B.F., and Paton D. (1999). Volcanic hazard perceptions: comparative shifts in knowledge and risk. Disaster Prevention and Management. 8, 118126. Johnston, M. (2000). Anxiety beliefs in the lower North island, New Zealand. Unpublished report, School of Psychology, Massey University, Palmerston North. Kaplan, Abraham (1964). The Conduct of Inquiry. San Francisco: Chandler. Kapucu, Naim (2008). Collaborat ive emergency management: better community organizing, better public peparedness and response. Disaster 32(2), 239-262, \ _________(2008). Culture of preparedness: household disaster preparedness. Disaster Prevention and Management 17, 526 -535. Katast rophenVorschrift (1988), Bonn: Deutsches Rotesse Dreuz (in German).
129 Kline, Rex B. (2005). Principles and Practice of Structural Equation Modeling (2ed.), New York: The Guilford Press. Kreps, G. (1985). Disaster and the social order. Sociological Theory 3, 4965. _________(1998). Disaster: systemic event and social catalyst (pp. 31-55) In What is a disaster? Perspectives on the question (Ed: E. L. Quarantelli), London: Routledge Lamontaigne, M. and La Rochelle S. (2000). Earth sci entists can help people who fear earthquakes. Seismological Research Letters, 70, 14. Lazarus, R. S. (1966). Psychological stress and the coping process. New York: McGraw -Hill. Lazarus, R.S. and Folkman, S. (1984). Stress, appraisal and coping. New Yo rk: Springer -Verlag. Lechliter, G.J. and Willis, F.N. (1996). Living with earthquake beliefs and information. The Psychological Record 46, 391-396. Lindell, M. and Perry R.W. (2000). Household adjustment to earthquake hazard: a review of the research. Environment and Behavior 32, 461-501. Lindell, M. K. and Whitney D.J. (2000). Correlates of household seismic hazard adjustment adoption. Risk Analysis 20(1), 13 -25. Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M., and Combs B. (1978). Judged frequency of Lethal Events. Journal of Experimental Psychology: Human Learning and Memory, 4, 551 578. Lomnitz, Larissa (1977). Networks and marginality. New York: Academic Press. McIvor, David and Douglas Paton (2007). Prepar ing for natural hazards: normative and attitudinal influences. Disaster Prevention and Management 16(1), 7988. Lopes, R. (2000). Community disaster education. Proceedings of the Planning for an Earthquake in New Zealand Conference, IRL Conference C enter, Gracefield, 28 29 February. Lower Hutt: Institute of Geological and Nuclear Sciences. Marsh, C. (1982). The Survey Method: the Contribution of Surveys to Sociological Explanation. London: George Allen and Unwin. McIvor, D. and Paton, D. (2007). Preparing for natural hazasrds: normative and attitudinal influences. Disaster Prevention and Management 16(1), 79 -88.
130 Mileti, D. L. (1999). Disasters by designa reassessment of natural hazards in the United States. Wash DC: Joseph Henry Press. Mulilis, J. and Duval T.S. (1995). Negative Threat Appeals and Earthquake Preparedness: A Person -Relative-to -Event (PrE) Model of Coping with Threat. Journal of Applied Social Psychology 25, 1319 -1339. Murphy, B. L. (2007). Locating social capital in resilient community level emergency management. Natural Hazards, 41, 297 -315. National Weather Service (nd). (Landfalling United States major hurricanes. Retrieved 11 December 2009 from NWS http://www.nhc.noaa.gov/pasta11.shtml#tracks_us Nunnaly, J. (1978). Psychometric theory. New York: McGraw -Hill. Oliver -Smith, Anthony (1996). Anthropological research on hazards and disasters. Annual Review of Anthropology 25, 303 -328. ________(1998). Global changes and the definition of disaster (pp. 177 -194) In What is a disaster? Perspectives on the question (Ed: E. L. Quarantelli), London: Routledge. _________(1999). What is a disaster? (pp. 1834), in The Angry Earth: Disaster in anthropological perspective (Eds: Anthony Oliver -Smith and Susanna M. Hoffman). Paton, D. (2003). Disaster preparedness: a social -cognitive perspective. Disaster Prevention and Management 12(3), 210216. Paton, D., Smith, L.M. and Johnston, D. (2000). Volcanic hazards: risk perception and preparedness. New Zealand Journal of Psychology, 29, 8488. Paton, D., Smith, L.M., Johnston, D., Johnston M., and Ronan, K. (2003). Developing a Model of Intentions to Predict Natura l Hazard Preparation. EQC Research Report No. 01 479, New Zealand Earthquake Commission, Wellington. Perry, Ronald W. (2006). What is disaster? (pp 1-12) In Handbook of Disaster Research (Eds: Havidan Rodriguez, Enrico L. Quarantelli and Russell R. Dynes), New York: Springer. Pescosolido, Bernice A. (2006). The sociology of social networks. 21ST Century Sociology. Sage Publications. Retrieved from http://www.sage ereference.com/sociology/Article_n20.html __________(1 992). Beyond rational choice: the social dynamics of how people seek help. American Journal of Sociology 97: 1096 -1138.
131 Porfiriev, Boris N. (1998) Issues in the definition and delineation of disasters and disaster areas (pp. 5672) In What is a disaster? Perspectives on the question (Ed: E. L. Quarantelli), London: Routledge Quarantelli, E. L. (1985). What is a disaster? The need for clarification in definition and conceptualization in research (pp. 41-37), in Disasters and Mental Health: Selected C ontemporary Perspectives (Ed: S. Solomon), Washington DC: US Government Printing Office _________(1987). What should we study? International Journal of Mass Emergencies and Disasters 5, 7-32. _________(1995). What is a disaster? (Editors Introduction). International Journal of Mass Emergencies and Disasters 13(3), 221-230. _________(2005). A social science research agenda for the disasters of the 21st century: theoretical, methodological and empirical issues and their professional implementation (pp. 325 -396) in What is a disaster? New answers to old questions (Eds: Ronald W. Perry and E. L. Quarantelli) US: International Research Committee on Disasters. Rapoport, A. and Horvath, W. (1961). A study of a large sociogram. Behavioral Science 6, 279-291. Rosenthal, U. (1998). Future disasters, future definitions. In E.L. Quarantelli (Ed.) What is a disaster? Perspectives on the Question, (pp. 146-159). London: Routledge. Rustemli, A. and Karanci, A.N. (1999). Correlates of earthquake cognitions and preparedness behavior in a victimized population. The Journal of Social Psychology 139, 91 101. Sheskin, David J. (2007). Handbook of Parametric and Nonparametric Statistical Procedures. Boca Raton, FL: Taylor & Francis Group, LLC. Sjoberg, Lennart (2000). Factors in risk perception. Risk Analysis, 20(1), 1-11. Smith, K. (1993). Environmental Hazards: Assessing risk and Reducing Disaster. London: Routledge. Smith, S.K. an d McCarty C. (2009). Fleeing the Storm(s): An Examination of Evacuation Behavior During Floridas 2004 Hurricane Season. Demography 46(1), 127145. Stack, Carol (1974). All our kin. New York: Harper & Row.
132 Thoits, Peggy A. (1986). Social support as coping assistance. Journal of Consulting and Clinical Psychology 54, 416 -423. __________ (1995). Stress, coping, and social support processes: Where are we? What next? Journal of Health and Social Behavior 35, 53 -79. Thompson, B. (2000). Ten commandments of structural equation modeling. In Grimm, L.G. and Yarnold, P.R. (Eds.) (p. 261-283) Reading and understanding more multivariate statistics. Washington, D.C.: American Psychological Association. Thompson, C.P. and Mingay D. (1991). Estimating the frequency of everyday events. Applied Cognitive Psy chology 5, 497510. Tierney, K.J., Lindell M.K. and Perry R.W. (2001). Facing the Unexpected: Disaster Preparedness and Response in the United States. Washington, D.C.: Joseph Henry Press. Tobin, G.A., Bell, H.M., Whiteford, L.M., and Montz, B.E. (2 006). Vulnerability of displaced persons: relocation of park residents in the wake of Hurricane Charley. International Journal of Mass Emergencies and Disasters 24(1), 77109. U.S. Census Bureau (n.d.) Current Population Survey (CPS) Definitions an d Explanations. Retrieved January 3, 2010 from http://www.census.gov/population/www/cps/cpsdef.html Von Winterfeldt, D., John, R.S., and Borcherding K. (1981). Cognitive components of risk ratings. Risk Analysis, 1, 277287. Wenzel, L., Glanz K., and Lerman C. (2002). Stress, coping and health behavior. In K. Glanz, B. K. Rimer and F. M. Lewis (Eds.), Health Behavior and Health Education Theory, Research and Practice, 3rd Edition (pp. 210 -239). San Francisco, CA: Jossey -Bass. Wikman, Anders (2006). Reliability, validity and true values in surveys. Social Indicators Research 78, 85 -110. Wisner, B., Blaikie, P., Cannon, T., and Davis I. (2004). At risk: natural hazards, peoples vulnerability and disasters (2nd Ed). New York: Routledge
133 BIOGRAPHICAL SKETCH Molly Moon enlisted in the United States Air Force directly out of high school in 1975. After earning her Bachelor of Science degree in business m anagement while on active duty from Troy State University in Montgomery, Alabama, Molly was commissioned an officer in 1984. During her 29year Air Force career, she earned a Masters of Business Administration degree from Barry University. She also compl eted Squadron Officer School, Air Command and Staff College and Air War College. Upon retiring from the Air Force in 2004, Molly moved to Gainesville, Florida where she volunteered her time to serve as the Project Coordinator for a Habitat for Humanity Wom an Build house. Upon completion of the Woman Build project, Molly returned to school at the University of Florida to pursue her doctorate degree. Molly is currently a full -time lecturer in the Department of Family, Youth and Community Sciences where she teaches Principles of Family, Youth and Community Science and Introduction to Social and Economic Aspects of Community.