an abstract of the dissertation of - Oregon State University
-
Upload
khangminh22 -
Category
Documents
-
view
0 -
download
0
Transcript of an abstract of the dissertation of - Oregon State University
AN ABSTRACT OF THE DISSERTATION OF Alexandra Buylova for the degree of Doctor of Philosophy in Public Policy presented on May 18, 2018. Risk Perceptions and Behavioral Intentions: Responses to the Threat of Cascadia Subduction Zone Earthquakes and Tsunamis. Abstract approved: _____________________________ Lori A. Cramer
This research examined public perceptions of risk, behavioral intentions in the event of the M9
Cascadia Subduction Zone (CSZ) earthquake and tsunami on the Oregon Coast, and factors that
may influence both attitudes and intentions. A household survey was conducted to understand
public opinion in Seaside, Oregon, which is located within the impact radius of the CSZ earthquakes
and tsunamis. Mediation analysis was applied to evaluate relationships between the three levels of
variables to examine ways in which human decision-making occurs under conditions of risk.
Research results found a positive association between risk perception and evacuation behavioral
intention, suggesting that higher risk perception motivates people to be proactive and adopt
recommended actions of immediate evacuation in a local tsunami emergency. In addition, relying on
mediation analysis, the study found that cognitive constructs of response efficacy and self-efficacy
played a mediating role between socio-environmental factors and behavioral intention, supporting
arguments advocated by theories of cognition that attitudes translate outside influences into
behavioral intentions. Older individuals were identified as vulnerable to a local tsunami risk due to
their lower levels of risk perception, self-efficacy, and physical mobility capacities.
Risk Perceptions and Behavioral Intentions: Responses to the Threat of Cascadia Subduction Zone Earthquakes and Tsunamis
by
Alexandra Buylova
A DISSERTATION
submitted to
Oregon State University
in partial fulfillment of the requirements for the
degree of
Doctor of Philosophy
Presented May 18, 2018 Commencement June 2018
Doctor of Philosophy dissertation of Alexandra Buylova presented on May 18, 2018
APPROVED: ______________________________________________________________________________ Major Professor, representing School of Public Policy ______________________________________________________________________________ Director of the School of Public Policy ______________________________________________________________________________ Dean of the Graduate School I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my thesis to any reader upon request. ______________________________________________________________________________
Alexandra Buylova, Author
ACKNOWLEDGEMENTS
I am grateful for continuous guidance and support of my major advisor Dr. Lori Cramer and my committee members: Dr. Brent Steel, Dr. Drew Gerkey, Dr. Denise Lach, and my Graduate Representative Dr. Marie Harvey. I am also thankful for financial and logistical support for this project and my education to the School of Public Policy, and Sea Grant and NSF grants of Dr. Lori Cramer and Dr. Haizhong Wang. And gods bless my friends!
TABLE OF CONTENTS Page
1 Introduction………………………………………………………………………………...…...1 2 Literature review……………………………………………………………………………...…9 2.1 Behavioral intentions……………………………………………………………...…...9 2.2 Cognitive and socio-environmental factors……………………………………….........13 2.2.1 Risk perception……………………………………………………………...15
2.2.2 Self-efficacy and response efficacy…………………………………………..19 2.2.3 Socio-environmental factors……………………………………………........21 2.2.4 Demographic variables……………………………………………………...28 3 Background and method………………………………………………………………………..32
3.1 Sample………………………………………………………………………….…..…32
3.2 Survey design…………………………………………………………………………35
3.3 Survey distribution………………………………………………………………....….37
3.4 Variables description and operationalization…………………………………………..39
3.4.1 Dependent variables………………………………………………………...39 3.4.2 Independent cognitive variables…………………………………………......42 3.4.3 Independent socio-environmental variables……………………………....…43 3.4.4 Demographic variables…………………………………………………...…50 3.5 Data cleaning……………………………………………………………………...….51 3.6 Analytical procedures……………………………………………………………...….53 4 Results………………………………………………………………………………………….54 4.1 Behavioral intentions descriptive statistics…………………………………………….54 4.2 OLS regression analysis…………………………………………………………....….56 4.3 Mediation analysis…………………………………………………………………….64 4.4 Analysis of variance……………………………………………………………....…...83
5 Discussion………………………………………………………………………………...……88
TABLE OF CONTENTS (Continued) Page
6 Limitations and suggestions for future research…………………………………………….....102
7 Key findings and policy recommendations………………………..…………………… ….…108 8 Conclusion……………………………………………………………………………….…...115 Bibliography……………………………………………………………………………....……117 Appendices………………………………………………………………………………....…..129 Appendix A Survey Cover Letter………………………………………………………130 Appendix B Survey Questionnaire………………………………………….…....……..131
LIST OF FIGURES Figure Page 1.1 Illustration of the Cascadia Subduction Zone extension from Northern California to
Canada.....................................................................................................................................................3 2.1 Research framework, indicating relationships between variables that are addressed in
research questions and hypotheses………………………………………………...………31 3.1 Map of Seaside and Tsunami Wave Arrival Time………………………………..……...…47 4.1 Descriptive statistics of behavioral intentions……………………………………………..55 4.2 Illustration of a total effect of X on Y…………………………………………………......65 4.3 Illustration of a simple diagram of mediation……………………………………...………66
LIST OF TABLES Table Page
2.1 Research questions and research hypotheses……………………………………...……….30 3.1 Sample bias, comparing U.S. Census statistics for Seaside with survey respondents….….....35 3.2 Descriptive statistics of behavioral intentions……………………………………………..40 3.3 Pattern matrix for behavioral intent………………………………………………....…….41 3.4 Factor matrix of perceived severity………………………………………………………..42 3.5 Summary of hazard-relevant knowledge survey responses………………………........……45 3.6 Factor matrix for knowledge confidence……………………………………....…………..45 3.7 Summary of dependent and independent variables……………………………………..…49 3.8 Summary of demographic variables……………………………………………………….51 4.1 OLS regression output for leadership dependent variable (coefficients are standardized).....57 4.2 Pearson’s correlations between independent variables……………………......……………60 4.3 Pearson’s correlations between demographic variables……………………………….........61 4.4 OLS regression output for delay dependent variable (coefficients are standardized)….........62 4.5 Mediation Analysis Results Leadership Dependent Variable……………………......…...…70 4.6 Mediation Analysis Results Delay Dependent Variable……………………………........….71 4.7 Summary of research hypotheses testing………………………………………….…….…82 4.8 Descriptive statistics and Shapiro-Wilk test of normality for dependent variables…..............84 4.9 Results of ANOVA and Kruskall-Wallis tests………………………….……......…....…..…86
1
1 Introduction
Potential for more intense natural hazards and disasters is increasing as Earth system
dynamics are changing rapidly and unexpectedly, creating more frequent and intense storms,
wildfires, droughts, and floods (Mileti 1999; EM-DAT(International Disaster Database)). In addition
to these slow-onset and high-frequency events, a substantial danger to human life is inherent in
rapid-onset low-frequency tsunamis, whose impacts on coastal areas have been significant in the past
century. In a recent large-scale event in Chile (1960) a tsunami killed over 1,000 people across Chile,
Japan, and Hawaii. In Alaska (1964) a tsunami killed 117 people; in Sumatra (2004) a tsunami
claimed more than 288,000 lives in a dozen countries; in the Samoa Islands (2009) a tsunami killed
192 people across the affected islands; and in Tohoku-oki (2011) a tsunami left 15,892 dead in Japan
(Apatu et al. 2016). Considering the devastating outcomes of these events, it is important to
understand how people prepare for and respond to tsunami threats to manage exposure to this
natural hazard on individual and organizational levels.
The importance of preparation for effective disaster response, including tsunamis, has been
voiced by the United Nations in the 2015-2030 Sendai Framework for Disaster Risk Reduction.
Reducing tsunami risks in the United States can be difficult due to the lack of information and
experiential knowledge of such events. In the U.S. the most recent tsunamis took place when a Chile
(1960) tsunami reached the Hawaiian coast killing 61 and injuring more than 100 people in the town
of Hilo (Lachman et al. 1961); an Alaska (1964) earthquake and tsunami killed 106 in Alaska, 5
people in Oregon, and 13 people in California; and a Samoa Islands (2009) earthquake and tsunami
claimed 34 lives in the American Samoa (Apatu et al. 2016). As a result, a majority of the available
data on tsunami preparedness and evacuation is based on studies of hurricanes, storms, floods, and
wildfires due to the frequency of such events in the U.S. (Lindell et al. 2011; Lindell and Prater 2007;
2
Toledo et al. 2018). However, these events are more predictable and require different preparedness
and response strategies in contrast to rapid-onset tsunamis.
Most often tsunamis are caused by earthquakes along the ocean seafloor. The biggest ones
occur at subduction zones like the Cascadia Subduction Zone (CSZ) along the west coast of the U.S.
and Canada, where an oceanic plate slides beneath a continental plate. Other causes of tsunamis
include underwater volcanic eruptions and landslides. Tsunamis that travel thousands of miles from
their source are called distant tsunamis. The more devastating events, however, are local or near-
field tsunamis. They can approach coastlines within minutes of an earthquake onset, leaving
minimum reaction time for populations in risk zones. By the time a tsunami wave hits the coastline,
it may have a speed of up to 30 mph and a height from 20 to 100 feet, depending on the
characteristics of the shoreline (water depth and shape of the shore) and the degree of fault
movement. Most local tsunamis are composed of a series of waves that can arrive within hours or
even days after the earthquake. The first wave may not be the most destructive (Oregon Department
of Land Conservation and Development 2016). An earthquake and a tsunami along the Cascadia
Subduction Zone would fall into the category of a local tsunami for the west coast of North
America. This research centers around understanding individual perceptions and behavioral
intentions under the risk of CSZ earthquakes and tsunamis on the Oregon coast.
The CSZ is a fault line that extends along the west coast of North America from the
Mendocino Ridge off the coast of northern California to northern Vancouver Island, British
Columbia. A magnitude 9.0+ earthquake (which is considered the worst-case CSZ scenario, or M9
scenario) along the CSZ is predicted to occur in the next 50 years, with different probabilities of the
strength, extent, and location of the earthquake epicenter (Goldfinger et al. 2012). Figure 1.1
illustrates the location of the CSZ relative to the west coast of North America and Oregon – the
3
primary place of interest in this research. The map indicates the edge of the subduction zone, where
the Juan de Fuca oceanic plate begins to slide underneath the North American continental plate,
creating a traction that can cause an earthquake.
Figure 1.1: Illustration of the Cascadia Subduction Zone extension from Northern California to Canada; Source: FEMA,
Secretary of State Audit report, January 2018.
A CSZ MW (moment magnitude) 9.0 earthquake is capable of generating a tsunami wave of 8
meters and higher that could inundate the coast in 15-30 minutes after the initial earthquake (Wood
et al. 2010). Earthquake magnitudes are measured on the seismic moment (MW) scale, which is one
of the most common metrics used by modern seismologists to describe the physical strength of an
earthquake. It is different from the Richter magnitude scale or local magnitude (ML) and surface-
4
wave magnitude (MS) estimates, which are becoming less popular. Some scientists argue that the
latter methodologies do not take into the account various characteristics and effects of the natural
event. For example, the 1964 Alaska earthquake using two different scales can be described as MS
8.3 and MW 9.2 event (Spence et al. 1989, USGS).
Several possible CSZ tsunami scenarios have been developed by the Oregon Department of
Geology (DOGAMI) for the Oregon coast using percentages of potential variability in the
subduction zone movement. The worst-case or M9 CSZ tsunami will happen in the case of a full
subduction zone slippage. M9 CSZ tsunami scenario was used to generate tsunami inundation maps
for local tsunami preparedness in Oregon coastal communities (Priest et al. 2016). There are several
methodologies used to estimate the probability of earthquakes along the CSZ, with varying degrees
of uncertainty. A probability of 7-11% of CSZ full rupture (M9 scenario) in the next 50 years was
estimated using both a Poisson and time-dependent calculations. According to failure-analysis, CSZ
full rupture will have exceeded approximately 25% of known recurrence intervals by 2060. For a
partial rupture along the southern segment of the CSZ, the probability of the event occurrence rises
to 18% (Poisson distribution) and 32–43% (time-dependent model), while the earthquake will have
exceeded approximately 85% of its known recurrence intervals by 2060 according to failure analysis
(Goldfinger et al. 2012). Full-length CSZ earthquake recurrence interval is estimated at 530 years,
with the last event taking place on January 26, 1700. These probabilities suggest a presence of real
risk for communities along the west coast of North America. According to Wood et al. (2015), the
CSZ tsunami hazard areas in Northern California, Oregon, and Washington contain 94,872
residents, 42,424 employees, 486 public venues, 440 dependent-care facilities, and 2,314 businesses
with a significant customer presence. In addition, continuous development of U.S. coastal regions is
exacerbating the vulnerability of the coast to tsunami impacts (Crossett 2005; Mileti 1999). Potential
5
damages to coastal communities in terms of mortality rates and impacts on infrastructure could be
substantial as a result of CSZ earthquakes and tsunamis.
It is important to remember that in the event of a local tsunami populations at risk are
generally subjected to two hazards: an earthquake and a tsunami. Individuals need to survive the
impacts of an earthquake before they attempt to evacuate outside of a tsunami hazard zone.
Earthquake preparedness and mitigation is not the focus of this research. In this study, earthquake is
regarded as the primary natural warning sign of an upcoming tsunami. However, it is important to
consider that people may be injured by an earthquake before they have a chance to escape from a
tsunami wave. Tsunami evacuation routes can also be severely damaged or completely impassable
due to earthquake impacts. When thinking about tsunami preparedness, one cannot exclude
earthquake impacts from consideration. However, because earthquakes and tsunamis are different
types of events that imply a different type of response, this research mainly focuses on the threat of
local tsunamis.
From an emergency management perspective, natural hazards are understood in terms of
mitigation, preparation, response, and recovery stages. Risk reduction activities that happen during
any of these stages contribute to the overall resilience of a system as they minimize risks to life and
impacts to physical, economic, and social structures. Investing in all stages of natural hazards
emergency management is important. However, the nature of a local tsunami is unique because of
the immediacy and the unpredictability of the event onset. In local tsunamis, many lives are lost in
the first hour of a disaster (Yun and Hamada 2012). Therefore, generating knowledge that can assist
emergency planners in minimizing human mortality during the response stage is essential and is the
primary focus of this study.
6
Contributions to risk reduction efforts for a successful tsunami response can include
developments in earthquake early warning systems (EWS), road networks and evacuation routes,
relocation of critical facilities outside of the tsunami inundation zone, construction of tsunami
evacuation towers, promotion of awareness educational programs, and others (Raskin and Wang
2016; Dunn et al. 2016; McCaughey et al. 2017). However, benefits of these activities cannot be fully
understood without measuring and forecasting individual and social behaviors. What actions people
adopt to protect themselves in reaction to and in the risk of natural hazards depends on a range of
cognitive, emotional, social, environmental, and institutional factors, as well as on built and natural
environments (Dunn et al. 2016). For example, building a tsunami evacuation tower may be the most
effective approach to ensure a successful evacuation during a tsunami event in certain locations. Yet,
people may be unaware of tsunami evacuation towers in their communities or they may not trust the
structures to utilize. Therefore, one may consider investing in a different risk reduction strategy or
try to change people’s attitudes toward built structures.
A successful protective response to tsunamis is impossible without active participation of
individuals, involved in making tough decisions. Effective planning for tsunamis requires advances
in engineering and hazard forecasting. However, hard protective structures do not guarantee full
safety (Mimura et al. 2011). In addition, they do not eliminate the importance of understanding how
people evaluate and interpret risks and what actions they take to protect themselves from hazard
impacts. Human behavior in general is difficult to predict and even more so in stressful and chaotic
events like disaster emergencies (Shapira et al. 2018). Individual response to risk involves cognitive
and emotional reactions that are influenced by social, institutional, and environmental contexts.
Because of a variety of possible behaviors that people can adopt in disaster situations, measuring and
7
predicting individual perceptions and behaviors is an integral part of preparedness and risk reduction
efforts (Dunn et al. 2016).
This study is a pre-impact assessment of individual perceptions and behavioral intentions of
the Oregon coastal population in the event of a local M9 CSZ earthquake and tsunami. Only
behavioral intentions could be empirically assessed. It is important to pay attention to the Oregon
coast in the context of this research for several reasons. The population of Oregon’s coastal counties
grew by 73.4% from 1960 to 2008 (Wilson and Fischetti 2010), with the population exposed to
tsunami hazard projected to increase by 3880 households and 6940 residents by 2061 in U.S. Pacific
Northwest (Sleeter et al. 2017). The increase in the number of people vulnerable to CSZ earthquakes
and tsunamis drives the importance of disaster preparedness to prevent injuries and reduce damages
to infrastructure along the coast.
Why do people have certain intentions and what factors may explain variation in their
behavioral intentions and preparation? Answers to these questions could assist in emergency
planning activities, from developing public education and awareness programs to improving
evacuation forecasting models. To address these questions, I offer a critical overview of natural
hazards scholarship and theory on the relationship between how people perceive and interpret risk
and the intentions they form as a consequence of such interpretations. The goal of my research is to
provide an additional insight into human decision-making process under conditions of natural
hazard risks.
The prominence of natural disaster research is underscored by a number of studies that have
examined public perceptions of risk, adoption of protective behaviors and behavioral intentions, and
factors that determine both perceptions and behavior in various emergency contexts including
technological, environmental, and man-made disasters. This research draws on the literature of
8
behavior in rapid-onset natural disasters such as earthquakes, tsunamis, and volcanic eruptions. Yet,
because of fewer studies on earthquakes and tsunamis, literature on slow onset natural disasters such
as hurricanes and floods is examined as well. I do not attempt to cover the literatures of risk
perceptions and behavior within contexts of technological and industrial hazards, war and terrorism,
disease outbreaks, and the like. There are continuities within literatures on natural hazards. People’s
behavior in certain non-natural hazard situations (e.g. terrorist attacks or nuclear plan meltdowns)
can resemble human reactions in rapid onset natural disasters (Riad et al. 1999). Yet, these
intersections mostly fall beyond the scope of this research.
The remainder of this manuscript is structured as follows. Chapter 2 provides a
comprehensive review of natural hazards literature about public perceptions and behaviors in
preparation for and during emergencies, focusing on rapid onset disasters such as earthquakes and
tsunamis. Findings from the literature are used as a benchmark for identifying variables of
importance in the analysis of behavioral intentions of individuals located in the CSZ risk area in
Seaside, Oregon. The data for the analysis was collected through a mail-based survey conducted in
the fall of 2017. Chapter 2 also proposes research hypotheses and research questions. Chapter 3
describes location of the study, sampling method, data collection procedures, variable
operationalization, and briefly reviews analytical approaches employed for data analysis. Chapter 4
explains in greater details analytical procedures employed in this research and presents results of data
analysis. Chapter 5 discusses the major research findings. Chapter 6 proposes practical applications
of the results for emergency preparedness. Chapter 7 concludes with research limitations and
suggestions for future work.
9
2 Literature Review
2.1 Behavioral intentions
Studies that examined behavior in natural disaster contexts often use terminology of
‘protective responses’ or ‘hazard adjustments’ to refer to people’s behavior in response to risk.
Hazard adjustments were conceptualized as a set of protective actions to reduce one’s vulnerability
to risk (Burton et al. 1993; Gregg et al. 2004). In different disaster settings ‘protective responses’ can
imply different types of behavior, e.g. retrofitting a house under a threat of an earthquake, storing
additional water and food at home in preparation for floods and storms, evacuating outside of
hazard area in the event of hurricanes and wildfires, and escaping to a high ground in the case of
tsunamis.
In terms of personal safety, evacuation from a hazard zone is considered the best protective
behavior in both slow- and rapid-onset natural disasters (Mas et al. 2012, 2013; Wei et al. 2017;
Charnkol and Tanaboriboon 2006; Takabatake et al. 2017; Wei and Lindell 2017). Research on
evacuation behaviors in slow-onset natural disasters (e.g. hurricanes, tornadoes, floods) is prevalent,
recognizing the frequent nature of such events (Dash and Gladwin 2007; Whitehead et al. 2000;
Kang et al. 2007; Baker 1991; Riad et al. 1999; Huang et al. 2016; Huang et al. 2012; Meyer et al. 2018;
Peacock 2003). Generally, during slow-onset disasters there is time for people to assess risks and
weight options of protective actions (e.g. evacuation vs. non-evacuation, reinforcing a home,
contacting/reuniting with family members, etc.). In slow-onset natural hazards, the ability to ensure
one’s physical safety is generally greater, because of the advanced warning that grants an advantage
of time for decision-making, compared to rapid-onset events such as earthquakes and tsunamis
(Dash and Gladwin 2007).
10
Immediate evacuation outside of hazard areas within the first minutes after an earthquake
shaking is considered the most effective personal protective response in local tsunamis (Mas et al.
2012, 2013; Wei et al. 2017; Charnkol and Tanaboriboon 2006; Takabatake et al. 2017; Wang et al.
2016). Ground shaking and informal channels of communication are considered the main signs of
an upcoming tsunami (Fraser et al. 2013). Bernard (2005) suggests that waiting for other natural
phenomena such as a rapid drawdown or a sudden rise of the ocean takes longer and requires
people to be in a visual proximity of a coastline, which would prevent people from evacuating in a
timely manner. In addition, official tsunami warning channels are likely to be damaged in an
earthquake (Fraser et al. 2013). The public is encouraged to rely on the natural warning sign of
earthquake shaking as a trigger for immediate evacuation.
However, post-impact tsunami analyses have shown that people tend to wait for tsunami
confirmation before engaging in immediate evacuation in local tsunami events. In the evaluation of
people’s behavior in the 2009 American Samoa tsunami, Lindell et al. (2015) reported that waiting
for tsunami confirmation from peers, authorities, and news media created delays in evacuation
among the local population. In a study of protective responses to a tsunami threat in 2011
earthquake in New Zealand and Japan, Wei et al. (2017) discovered that nearly half (47.7%) of
respondents said that they started to evacuate only when they saw a tsunami wave coming. In a
survey of Okitsu community following the 2011 Great East Japan earthquake and tsunami, Sun et al.
(2013) found that 10.9% of respondents would not evacuate without hearing a tsunami evacuation
order and 6.3% would wait to hear the evacuation order from community warning alarms system in
future tsunami events.
Other common behaviors that prevent people from evacuating outside of a hazard zone in a
timely manner include looking for family members, talking with neighbors, helping others, and
11
waiting for assistance from emergency personnel (Shibayama et al. 2013; Lindell and Perry 2012;
Murakami et al. 2012). While based on reports of survivors who may have conjectured the existence
of otherwise unknown behaviors, Yun and Hamada (2012) discovered that the most common
behaviors taken by people who died in the 2011 Great East Japan tsunami were helping others
(22.4%), doing rescue work (13.9%), and finding family or relatives (9.7%). There were reported
cases of parents travelling to collect their children from schools and individuals returning home to
look for elderly relatives, despite the imminent tsunami arrival. Similarly, Lindell et al. (2015)
reported that people delayed their evacuations in the 2009 American Samoa tsunami because they
were locating family members (36.8%), packing an emergency kit (26%), warning others (19.8%),
obtaining additional information from peers (24%), news media (15.6%), and authorities (11.5%),
protecting property (4.2%), and helping others (2.3%). To minimize such behaviors, following a
devastating 2011 Tohoku-oki earthquake and tsunami in Japan that left 5,892 people dead and 2,574
missing, a policy of immediate evacuation, known as tendenko, that discourages contacting family
members and waiting for tsunami confirmation, has been actively promoted in Japan (Mimura et al.
2011).
This study is set in a pre-impact context, therefore only assessment of evacuation behavioral
intentions was possible. This research inquired about people’s likelihood of adopting a range of
behaviors in the event of the M9 CSZ earthquake and tsunami (e.g. evacuating immediately following
ground shaking outside of tsunami inundation zone, contacting loved ones, waiting for tsunami
confirmation, collecting documents, etc.). Knowing people’s behavioral intentions is valuable for
several reasons. The information can be used to populate tsunami evacuation models with more
realistic data. Certainly, one has to be careful drawing conclusions from prediction analyses because
of possible inconsistencies between behavioral intentions and actual behavior. They can occur
12
because individuals report erroneous estimates about their intentions due to the lack of experience
and knowledge, variation of intentions over time, and unawareness of impeding or facilitating
factors to intended actions that may arise after intentions are first formed (Kang et al. 2007;
Grothmann and Reusswig 2006). In the evaluation of behavioral responses before and after
Hurricane Lili, Kang et al. (2007) found correlation between behavioral expectations and behavior.
Yet, empirical evaluations of association between intentions and behavior in studies of natural
disasters are rare. However, more importantly, information on behavioral intentions and how they
differ across different groups of individuals can assist in emergency preparedness by influencing
people’s attitudes, promoting more effective protective responses, and tailoring outreach efforts for
vulnerable populations to improve their level of preparedness. Therefore, I ask the following
research question.
Research question 1: What are the individual behavioral intentions within Oregon coastal
population in a response to M9 CSZ earthquake and tsunami scenario?
Certain populations may face greater risks and obstacles to disaster preparedness and
evacuation because of costs associated with those actions, physical abilities, access to information,
and capabilities to interpret and understand the information regarding preparedness guidelines and
evacuation options (Meyer et al. 2018). Identifying groups under risk of not adopting protective
behaviors or not intending to evacuate in the event of a local tsunami, and factors that influence
such decisions, could suggest channels for early intervention strategies that can, for example, create
access to resources necessary for successful preparation and protective actions adoption (Riad et al.
1999).
There is a range of factors that influence individual motivations and propensities to adopt
protective measures in a threat of disasters. Individual evacuation decision-making is a complex
13
phenomenon, influenced by cognitive and emotional (perceptions and attitudes) and external -
social, institutional, environmental, and demographic factors (e.g. previous experiences, knowledge,
location, gender, age, etc.) (Eiser et al. 2012). The following sections examine in detail cognitive,
socio-environmental, and demographic factors that influence people’s decision-making and explain
the variation in behavioral intent in emergency hazard situations. I rely on theories of cognition that
argue that human behavior is a product of processes of human thought and perceptions. Human
mental processes are arranged in a hierarchy of cognitions such as perceiving, remembering,
thinking, interpreting, which in turn form behavioral intentions and behavior. These mental
cognitive processes or perceptions, values, attitudes do not form on their own. They are influenced
by a number of factors – environmental influences, social norms and cultures, experiences, etc. The
following review identifies important factors that could explain people’s perceptions and behavioral
intentions.
2.2 Cognitive and socio-environmental factors
Theories of behavior (e.g. Theory of Reasoned Action, Theory of Planned Behavior,
Protection Motivation Theory, and Protective Action Decision Model) argue that psychological
processes such as perceptions and attitudes are the core components of cognition that translate
outside information into behavioral intent and ultimately behavior, serving as mediators between
observable factors and behavior, with Theory of Reasoned Action (Fishbein and Ajzen 1975) being
the first to synthesize the conceptual relationship between attitudes and behavior.
At the beginning of the 20th century, among social psychologists the notion that human
behavior is guided by attitudes was accepted as a given (Ajzen and Fishbein 2005). At that time, the
main approach to understanding the relationship between attitudes and behavior was based on the
notion of a logical consistency, i.e. a certain behavior was considered a logical outcome of a person’s
14
attitudes toward that behavior. In the 1930s, with the emergence of more empirical research, the
assumption that attitudes determine behavior started to weaken. The majority of studies that showed
support for the hypothesis had methodological flaws (e.g. LaPiere 1934, Corey 1937). Research
applied mostly arbitrary measures of attitudes and behavior, while attitude-behavior logical
connections were guided primarily by researcher intuition. By the 1960s and 1970s, many scholars
questioned the strength of the relationship and the utility of the attitude constructs (e.g. Blumer
1955; Campbell 1963; Deutscher 1966; Festinger 1964 in Ajzen and Fishbein 2005).
Fishbein and Ajzen (1975; 1977) revived the determinative power of attitudes on behavior
by improving attitude and behavior measurements. They contended that attitude and behavior
measurements must correspond in terms of action, time, target, and context to serve as a basis for
comparison. For example, an individual’s general attitude toward religion, may not be the most
reliable predictor of a specific behavior such as attending a church. Or, person’s voting preference
toward wolf reintroduction would be better determined by attitudes toward wolf reintroduction
rather than beliefs about or attitudes toward wolfs (Bright and Manfredo, 1996). In addition,
Fishbein and Ajzen (1977) proposed that behavior in question could be better explained by the
intention to perform an action rather than an attitude, creating a link in the cognitive process
between attitudes-behavioral intentions-behaviors. Their argument that behavioral intentions may
serve as a moderator between attitudes and behavior was a departure from the traditional view that
attitudes directly influence behavior. Yet, this is now a widely accepted principle and the foundation
of Theory of Reasoned Action (TRA), Theory of Planned Behavior (TPB) (Ajzen 1991), Protection
Motivation Theory (PMT) (Rogers 1983), and Protective Action Decision Model (PADM) (Lindell
and Perry 1992, 2003, 2012), the last two being commonly applied in natural hazards research.
15
TRA argues that behavioral intentions can be explained by two cognitive processes: (1) an
attitude toward a behavior (e.g. perception that immediate evacuation outside of a risk zone will
result in a positive outcome of being safe from injuries) and (2) a normative belief or a social
perception regarding the behavior (e.g. perception that significant others think one should or should
not evacuate during a tsunami) (Armitage and Christian 2003). TPB added an additional cognitive
construct to the behavioral model - perceived control over the performance of a behavior or
perception that one is capable of successfully executing the behavior (Montano and Kasprzyk 2015).
In addition, TRA and TPB advocated that personal and situational factors play a role in shaping
people’s beliefs and behaviors.
Protection Motivation Theory (PMT), originally proposed by Rogers (1983), expanded both
TRA and TPB by drawing attention to the influence of risk perceptions on behavioral intentions and
behavior. Originally developed to understand why people engage in unhealthy behaviors to provide
alternatives ways of changing those behaviors, it also found its application in studies of disaster
emergency because of high risk contexts. It offers a framework to examine why people engage or do
not engage in protective behaviors in risky situations. PMT proposed a two dimensional
measurement of risk perception: (1) perceived probability (e.g. perception of how likely a tsunami is
to occur) and (2) perceived severity (e.g. perception of how likely a tsunami is to cause personal and
property damages), in some contexts also referred to as a threat appraisal. In natural hazards
research risk perception plays a central role in explaining people’s decision-making processes in
emergency situations.
2.2.1 Risk perception
Risk is not an objective phenomenon that can be universally measured and ranked. It does
not exist independently of people’s minds and cultures. Risk arises out of an assessment of possible
16
impacts of a disaster and out of uncertainty surrounding those predictions (Sjoberg et al. 2004).
Assessment of risk, which occurs at institutional and individual levels, depends on many factors
including social and cultural beliefs, availability of resources to mitigate and control risks (e.g.
institutional, financial, and human), and physical and geographic facilitators or constraints. An
outcome of the assessment, which results in a certain level of risk perception, influences choices that
people make in regards to mitigation behaviors to address the threat of harm and uncertainty. Risk
perception is an important concept because it helps people to understand the dangers and
uncertainties of life and motivate certain behavioral choices to address those uncertainties (Krimsky
& Golding, 1992; Slovic, 1992; Wynne, 1992).
There are two main approaches that examine how risk perceptions are formed: the
psychometric paradigm and cultural theory. The psychometric paradigm, rooted in the discipline of
psychology, examines how individual cognitive characteristics, personality traits, needs, and
preferences influence formation of risk perceptions (Fischhoff et al. 1978; Goszczynska et al. 1991;
Rogers 1982; Jaeger et al.. 2013; Slovic et al. 1980; Slovic 1992; Slovic and Weber 2002; Lee and
Lemyre 2009; Gierlach et al. 2010). Cultural theory takes a broader view and focuses on how social
context and group norms and values influence what is perceived as dangerous. It argues that risk
perception is a product of culture, history; and ideology. Fear is not just an outcome of a rational
evaluation of existing risk, but a product of social learning (Douglas and Wildavsky 1982; Turner
1979; Tierney 1994). From cultural theory perspective, variation in a hazard severity or a type of
hazard may not necessarily change perceptions of risk. Rather, perceptions would be shaped by an
adherence to a certain worldview that is a product of interaction within a certain social order (Jones
and Faas 2017). Cultural theory is widely accepted within disciplines of anthropology and sociology.
These two approaches to understanding risk perception are not mutually exclusive as individual
17
cognitive processes of decision-making and cultural norms constantly interplay with one another, as
social and environmental factors interact with physiological and psychological makeups of
individuals, reinforcing or undermining different behaviors. This research does not adhere to a
specific paradigm of origins of risk perceptions. However, it examines factors that influence risk
perception that would align with both psychometric and cultural theory paradigms.
Risk perception (or threat appraisal) has been analyzed as a dependent, explanatory, and
mediating variable. It appears in studies of both behavior and behavioral intentions (Bourque et al.
2013). Following other studies that have examined the concept of risk perception in the emergency
context, this research measures risk perception as perceived probability (i.e. a judgement of
likelihood of an event occurrence) and perceived severity (i.e. a judgement of severity and likelihood
of suffering death or injury, property damage, severe disruptions to daily routines, and financial
losses) (Lindell and Hwang 2008; Huang et al. 2012, 2016; Lindell et al. 2016; Wei et al. 2017; Lindell
and Perry 2003; Peacock 2005).
Research on natural disasters often finds a positive correlation between risk perceptions and
adoption of protective responses in both natural and human-made disasters including earthquakes
and tsunamis (Fraser et al. 2016; Lindell and Whitney 2000; Lindell et al. 2009; Spittal et al. 2008;
Tekeli-Yesil et al. 2010; Wei et al. 2017; Apatu et al. 2013; Lindell and Perry 2000), hurricanes (Dash
and Gladwin 2007; Peacock 2003; Wu et al. 2015; Huang et al. 2012, 2016; Lindell and Hwang 2008;
Wachinger et al. 2013; Bird and Howes 2008), volcanic eruptions (Tobin et al. 2011; Perry and Lindell
2008; Johnston et al. 1999), floods and mudslides (Tobin et al. 2011; Grothmann & Reusswig, 2006;
Lin et al. 2008; Siegrist and Gutscher 2008; Lindell and Hwang 2008), and threats of terrorist attacks
(Bourque et al. 2013; Lee and Lemyre 2009). However, the direction and the degree of relationship
between risk perceptions, behavioral intentions, and behavior have also been reported to be
18
inconsistent across studies (e.g. Lindell and Prater 2000; Lindell and Whitney 2000; Paton et al. 2000;
Perry and Lindell 2008). Different research designs and measurements of risk perception contribute
to the variation in effects, suggesting that it is an important but not a sufficient factor that explains
individuals’ adoption of protective behaviors (Lindell and Whitney 2000). In the tsunami context,
Wei et al. (2017) and Lindell et al. (2016) found that higher levels of risk perception significantly
increased the likelihood of households attempting to evacuate in 2011 earthquakes in Christchurch,
New Zealand and Hitachi, Japan, while Lindell et al. (2015) reported that tsunami individuals’
evacuation decisions had a modest correlation with their risk perceptions in a post-impact evaluation
of the 2009 Samoa Islands earthquake and tsunami.
It is logical to assume that high levels of risk perception may lead to risk mitigating
behaviors, e.g. rapid evacuation in a case of an earthquake and tsunami. However, it is important to
be cognizant of the fact that sometimes people choose not to protect themselves even when
exhibiting high levels of perceived threat (Wachinger et al. 2013). Ruiter et al. (2001) argues that high
level of risk perception may cause denial behaviors as a protective mechanism. Consequently,
response to a threat may instead be manifested in a rejection of precautionary actions when
individuals believe that benefits of not performing a protective action overweight the negative
impacts, when they do not understand personal responsibility for taking actions, and when they do
not have resources to change a situation (Wachinger et al. 2013).
Based on the above review of the literature this study hypothesizes that risk perception
serves as a mediator that translates the influence of socio-environmental factors on the behavioral
intent. In addition, I hypothesize that risk perception will have a direct association with behavioral
intent. More specifically, the hypotheses are the following.
19
Hypothesis 1a: Perceived probability mediates the influence of socio-environmental factors on
behavioral intent.
Hypothesis 1b: Perceived probability is positively associated with the behavioral intent to evacuate
outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 1c: Perceived probability is negatively associated with the behavioral intent to engage in
delay and milling behaviors in the event of M9 CSZ tsunami.
Hypothesis 2a: Perceived severity mediates the influence of socio-environmental factors on
behavioral intent.
Hypothesis 2b: Perceived severity is positively associated with the behavioral intent to evacuate
outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 2c: Perceived severity is negatively associated with the behavioral intent to engage in
delay and milling behaviors in the event of M9 CSZ tsunami.
2.2.2 Self-efficacy and response efficacy
In addition to risk perception construct, TPB and PMT also argue that cognitive elements of
response efficacy and self-efficacy influence behavioral intentions. Response efficacy implies
attitudes toward effectiveness and value of a protective action. Self-efficacy is an evaluation of
personal abilities to perform a given action. It refers to respondents’ assessment of their knowledge,
skill, ability, energy, and financial resources in relation to demands of a protective action (Lindell and
Whitney 2000). Response efficacy and self-efficacy together are also referred to as a coping appraisal.
Generally, higher levels of self-efficacy and response efficacy are associated with higher
likelihood of adoption of protective actions. Johnston et al. (2005) showed a connection between
self-efficacy and decisions to prepare in the assessment of tsunami preparedness on the coast of
Washington. In the analysis of community preparedness in a threat of tsunamis in New Zealand,
20
Paton et al. (2008) found that individuals were more likely to prepare in the threat of a tsunami if
they thought that preparation was the response that would ensure their safety. On the other hand,
low self-efficacy (i.e. when the level of resources relative to the degree of threat is deemed
insufficient) can induce avoidance behaviors, even under high appraisal of threat (Abraham et al.,
1994; Rippetoe and Rogers, 1987; Van der Velde and Van der Pligt, 1991).
In this research self-efficacy is measured by assessing people’s confidence in their ability to
protect themselves to reduce chances of severe injuries in the event of the M9 CSZ earthquake and
tsunami. This study examines whether self-efficacy to perform protective actions has an influence on
behavioral intent and whether self-efficacy mediates effects of socio-environmental factors on the
dependent variable. Adhering to the common argument in the literature, I anticipate that individuals
with higher level of self-efficacy will display higher level of evacuation intent. The resulting
hypotheses are:
Hypothesis 3a: Self-efficacy mediates the influence of socio-environmental factors on behavioral
intent.
Hypothesis 3b: Self-efficacy is positively associated with the behavioral intent to evacuate outside
of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 3c: Self-efficacy is negatively associated with the behavioral intent to engage in delay
and milling behaviors in the event of M9 CSZ tsunami.
Response efficacy is measured by assessing people’s attitudes toward the value of emergency
plans and preparatory actions in increasing their chances of survival in the event of M9 CSZ
earthquake and tsunami. Similar to the assessment of self-efficacy, whether response efficacy has an
influence on behavioral intent and whether it mediates the influence of socio-environmental factors
on behavioral intent are examined. The specific hypotheses are:
21
Hypothesis 4a: Response efficacy mediates the influence of socio-environmental factors on
behavioral intent.
Hypothesis 4b: Response efficacy is positively associated with the behavioral intent to evacuate
outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 4c: Response efficacy is negatively associated with the behavioral intent to engage in
delay behaviors in the event of M9 CSZ tsunami.
2.2.3 Socio-environmental factors
Why do different individuals hold different perceptions and adopt different behavioral
intentions? What influences risk interpretation and what leads people to have intentions to adopt
either protective, risk-taking, or avoidance behaviors under conditions of risk? Lindell and Perry
(2012) argue that the process of adoption of protective actions starts with people’s reaction to and
interpretation of disaster warning signs. In a local tsunami, earthquake and informal communication
channels would be the only indicators of an approaching tsunami (Murakami et al. 2012; Fraser et
al. 2012). Knowledge about natural tsunami warning signs and the difference between warnings for
local and distance tsunamis is highly useful. The main natural sign for local tsunamis is ground
shaking. Others could include unusual wave formations and currents in ocean and estuaries that
expose portions of the ocean floor not normally visible even during low tides (Gregg et al. 2006). In
the event of a distant tsunami, people are encouraged to wait for official government issued
information about status of a tsunami and a possible mandatory evacuation, before evacuating to a
high ground.
In addition to tsunami warning signs, knowledge of tsunami evacuation routes, tsunami safe
places, recommendations for tsunami preparedness and response, and knowledge of extent of
tsunami inundation zones are important factors that influence how people react when a disaster
22
occurs. Studies of earthquakes and tsunamis that report failures by victims to evacuate to tsunami
safe zones, attribute such behaviors to a lack of prior knowledge about tsunamis, natural tsunami
warning signs, tsunami evacuation routes, and tsunami safe places in respective communities (Kurita
et al. 2007; Said et al. 2011; Iemura et al. 2006; Gregg et al. 2006; Okumura et al. 2011). Gregg et al.
(2007) reported that majority of survey participants in a post-tsunami impact assessment in Hilo,
Hawaii expected a tsunami alert via official sources. On the other hand, Fraser et al. (2016) reported
high level of awareness among respondents regarding their location relative to a tsunami hazard. The
evidence suggests that knowledge of tsunami warning messages, tsunami evacuation routes, tsunami
safe places, and other tsunami relevant knowledge can have an influence on behavior in tsunami
emergencies.
Because of the pre-impact nature of this study, it is not possible to assess people’s
interpretation of tsunami warning signs, however, it is possible to inquire about their knowledge of
tsunami warning signs and other tsunami-relevant information. Hazard knowledge has been shown
to be effective in promoting preparation and evacuation response in disasters (Walshe and Nunn
2012; Dudley et al. 2011; McAdoo, 2006; Yogaswara and Yulianto 2008; Said et al. 2011). In addition
non-conclusive evidence exists from a qualitative study of 2009 tsunami in American Samoa that
knowledge of natural tsunami warning signs prompted individuals to evacuate in a timely manner
during the event (Dudley et al. 2011).
Due to resource constraints and the nature of the data collection instrument, instead of
testing people’s actual knowledge, this study examines individual confidence in knowing tsunami
relevant information. More specifically it inquires whether people know the difference between local
and distant tsunamis, whether they understand natural warning signs of tsunamis, whether they
know where to get the information about preparation for tsunamis and about tsunami evacuation
23
routes, and other tsunami-relevant information. This study assumes that tsunami-relevant knowledge
confidence will have a positive influence of behavioral intent. Specific hypotheses are:
Hypothesis 5a: Tsunami-relevant knowledge confidence is positively associated with the behavioral
intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 5b: Tsunami-relevant knowledge confidence is negatively associated with the
behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami.
It is also assumed that tsunami-relevant knowledge confidence will be positively associated
with self-efficacy, response efficacy, perceived probability, and perceived severity. Relationship with
perceived severity is assumed to be negative. Higher confidence in knowing tsunami-relevant
information may reduce perceptions of danger. Related hypotheses are:
Hypothesis 5c: Tsunami-relevant knowledge confidence is positively associated with self-efficacy.
Hypothesis 5d: Tsunami-relevant knowledge confidence is positively associated with response
efficacy.
Hypothesis 5e: Tsunami-relevant knowledge confidence is positively associated with perceived
probability.
Hypothesis 5f: Tsunami-relevant knowledge confidence is negatively associated with perceived
severity.
Efficient and appropriate response to threats involves an ability to discriminate between
different situations, some more dangerous than others. Prior direct experience with hazards can
increase people’s knowledge and ability to recognize threats. This ability depends on how people
interpret previous experience and what information and skills they learn from it, depending on the
recency, frequency, and the degree of damage and casualties experienced personally and by relatives
and friends (Lindell and Perry 2012). Not all experiences ensure a more efficient response in the
24
future. There are examples of individuals with prior hazard experience without major damages,
develop a false sense of security, lower risk perception, and a belief that future events will unlikely
affect them (Becker et al. 2017; Shapira et al. 2018). Similarly, experience of evacuation in a tsunami
warning that was not followed by a tsunami can instill a sense of false security or an intention to wait
for confirmation of tsunamis in future emergencies regardless of their type. The situation is known
as Normalization Bias (Mileti and Fitzpatrick, 1992; Russell et al. 1995; Johnston et al. 1999). In
addition, individuals who previously experienced natural disasters, may develop a sense of security
and confidence in their skills to survive future events. For example, Yun and Hamada (2012) in the
study of Tohoku-Oki earthquake and tsunami showed that the respondents’ confidence in their
ability to ward off the impact of tsunamis was correlated with previous experience with earthquakes
without consequent tsunamis.
Assessment of experience is important because decisions made from experience are
different from choices made from description (Eiser et al. 2012). Previous studies confirm the
importance of hazard specific (and recent) experience in influencing adoption of protective actions.
Charnkol and Tanaboriboon (2006) in a study of tsunami evacuation behavioral intent in Thailand,
showed that respondents with prior experience (e.g. 2004 Indian Ocean tsunami) were three times
more likely to have an intent to evacuate faster and earlier than those without such experiences.
After a community lives through a disaster, the likelihood of adopting emergency preparedness
activities increases (Lindell et al. 2016). In the review of case studies of floods, droughts, earthquakes,
volcanic eruptions, wildfires, and landslides in Europe, Wachinger et al. (2013) summarized that in
majority of studies direct experience with disasters reinforced precautionary behaviors in the future.
In addition, experience has been shown to have both direct and indirect (via risk perception) effects
on hazard adjustment adoption (Lindell and Prater 2000; Lindell and Hwang 2008).
25
Because of the rarity of local tsunamis on the Oregon coast, with the last CSZ tsunami
occurring in 1700, this study examines individual experience with any natural hazard, assuming that
those with greater experience may be more aware of risks and damages that natural hazards can
bring to personal health and property. This study evaluates how influence of previous hazard
experience on behavioral intent is mediated by risk perception, self-efficacy, and response efficacy.
The corresponding hypotheses are:
Hypothesis 6a: Hazard experience is positively associated with the behavioral intent to evacuate
outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 6b: Hazard experience is negatively associated with the behavioral intent to engage in
delay behaviors in the event of M9 CSZ tsunami.
Hypothesis 6c: Hazard experience is positively associated with self-efficacy.
Hypothesis 6d: Hazard experience is positively associated with response efficacy.
Hypothesis 6e: Hazard experience is positively associated with perceived probability.
Hypothesis 6f: Hazard experience is positively associated with perceived severity.
Even though the M9 CSZ earthquake and tsunami is expected to make a large geographic
impact, it is still logical to assume that people’s location relative to the source of risk can make a
difference in their motivation to react to the threat of tsunami. Fraser et al. (2016) reported that
reasons for not evacuating during the local-source earthquake in Wellington, New Zealand included
living at an elevation high enough to be impacted by tsunamis. Responses to their survey revealed
that 95% of survey respondents made a correct assessment regarding their location relative to the
tsunami inundation zone. Other studies of behavior in tsunamis have examined correlation between
coastal distance, risk perception, and evacuation. Looking at the relationship between physical
measures of risk and cognitive measures of risk perception is especially interesting. It allows to
26
determine to what extent personal risk assessment is connected with evaluation of physical reality,
and how well people understand and comprehend their location relative to risk (Lindell and Perry
2012).
Tsunami research that has examined the influence of risk proximity on perceptions and
behavior reported evidence that people living in households closer to coastlines (as a measure of risk
proximity) were more likely to make a decision to evacuate, evacuate earlier and faster, and had
higher level of risk perception (Apatu et al. 2013, 2016; Lindell et al. 2015; Charnkol and
Tanaboriboon, 2006; Wei et al. 2017). This study assessed risk proximity via predicted tsunami wave
arrival time. The assumption is that greater tsunami wave arrival time means a further location from
the coastline, therefore lower risk of being hit by a tsunami wave. Similarly to previous socio-
environmental factors, this study examines the association between risk proximity and behavioral
intent, and whether influence of risk proximity on behavioral intent is mediated by cognitive factors
of risk perception, self-efficacy, and response efficacy. Based on previous studies, this research
assumes that risk proximity will be positively associated with risk perceptions. While there is a lack
of research on the connection between proximity to hazard and self-efficacy and response efficacy, it
is logical to assume that risk proximity could be negatively associated with self-efficacy (being closer
to risk may imply lower confidence in one’s ability to protect themselves) and positively associated
with response efficacy (being closer to risk may imply greater confidence and trust in protective
actions). The associated hypotheses are:
Hypothesis 7a: Risk proximity is positively associated with the behavioral intent to evacuate
outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 7b: Risk proximity is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
27
Hypothesis 7c: Risk proximity is negatively associated with self-efficacy.
Hypothesis 7d: Risk proximity is positively associated with response efficacy.
Hypothesis 7e: Risk proximity is positively associated with perceived probability.
Hypothesis 7f: Risk proximity is positively associated with perceived severity.
Engaging in tsunami preparatory measures in anticipation of the Cascadia earthquakes and
tsunamis can have an influence on people’s confidence in their ability to respond successfully to
tsunamis, confidence in the utility of protective actions and preparation, risk perceptions, and the
intent to evacuate in the case of emergency. Preparation can take different forms: from education to
acquiring skills of an emergency responder. The process of preparation suggests that people who
engage in such behaviors acquire new information and skills, possibly leading to changes in
perceptions and behavioral intent. In the evaluation of responses to the 2009 Samoa tsunami,
Lindell et al. (2015) found that preparation in a form of earthquake hazard awareness meetings
correlated with higher levels of risk perception, yet it did not show correlations with the likelihood
to evacuate to higher ground. This research assumes that greater level of preparedness is associated
with higher level of risk perceptions, self-efficacy, and response efficacy. Similarly to other socio-
environmental factors, this study hypothesizes that preparation will be associated with the behavioral
intent directly and will be mediated by risk perception, self-efficacy, and response efficacy. Specific
hypotheses are:
Hypothesis 8a: Preparation is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Hypothesis 8b: Preparation is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
Hypothesis 8c: Preparation is positively associated with self-efficacy.
28
Hypothesis 8d: Preparation is positively associated with response efficacy.
Hypothesis 8e: Preparation is positively associated with perceived probability.
Hypothesis 8f: Preparation is positively associated with perceived severity.
2.2.4 Demographic variables
Finally, examining demographic characteristics of the population at risk is an essential part of
any study of behavior in natural hazard. Knowing groups under higher risk for not adopting
protective behaviors or not intending to evacuate in the event of a tsunami, could suggest channels
of reaching out to those individuals to encourage them to evacuate and provide options for early
intervention strategies that can, for example, create access to the resources necessary for successful
evacuation and preparation (Riad et al. 1999). This research examines variation in behavioral intent
among demographic characteristics of individuals and households: gender, age, education, income,
community tenure, homeownership, household size, having difficulty walking, presence of children
and presence of elderly in a household. Depending on a hazard context, influence of demographic
characteristics on perceptions and behavioral intentions can vary, but there are certain common
trends in the association between demographics and behavior in emergency situations.
There is limited and uncertain evidence that females tend to exhibit higher levels of risk
perception in disaster situations (Lindell et al. 2015, Terpstra and Lindell 2012). Also, in post-tsunami
impact studies, distribution of human loss is often skewed towards older populations, with fatality
rates growing with an increase in age. Older populations tend to have restrictions in the ability to
evacuate (e.g. difficulty walking) and tend to stay behind to protect property (Murakami et al. 2012,
Rofi et al. 2006; Sun et al. 2013; Lindell et al. 2015). Yun and Hamada (2015) reported that in the
2011 Tohoku-Oki earthquake and tsunami 63% of the survivors were younger than 39 years of age,
while only 3% were older than 60 years of age. Sun et al. (2013) reported that 64.3% of victims in the
29
Great East Japan Earthquake were over 60 and 45.5% were over 70. Charnkol and Tanaboriboon
(2006) found that respondents between 20 and 40 years of age were more likely to be the first and
the fastest to respond to a disaster and evacuate compare to older persons.
Household size and the presence of elderly in a household may affect household evacuation
decisions as well (Lindell et al. 2015). In disaster research, older populations are typically defined as
65 years and older. Elderly persons may require physical assistance in the evacuation process due to
mobility issues, a reluctance to evacuate, or lack of other necessary resources for successful
evacuation (Wood et al. 2007; Sun et al. 2017; Yun and Hamada 2012, 2015). The presence of
children can also influence the evacuation process as they may also have mobility issues and require
special assistance (Charnkol and Tanaboriboon 2006). Dash and Gladwin (2007) and Apatu et al.
(2016) define the category of children for evacuation purposes as 10 years of age and younger. Wei et
al. (2017) showed a correlation between tsunami evacuation and community tenure, measured in a
number of years lived in a community, where community tenure was expected to influence
knowledge and awareness of the tsunami hazard. They also found that homeowners were more
likely to exhibit higher levels of risk perception compared to non-homeowners. Having education
could provide people with necessary resources to prepare themselves for emergency situations. And
finally, Apatu et al. (2016) in the study of Samoa islands tsunami showed that households reporting
higher incomes had an increased probability of evacuation.
Research question 2: What is the variation in behavioral intent among the following
demographic characteristics: gender, age, formal education, income, community tenure,
homeownership, household size, having difficulty walking, presence of children, and presence of
elderly in a household?
Table 2.1 displays research questions and hypotheses addressed in this study.
30
Table 2.1: Research questions and research hypotheses
Research question 1
What are the variations in individual behavioral intentions within Oregon coastal population in a response to M9 CSZ earthquake and tsunami scenario?
Research question 2
What is the variation in behavioral intent within the following demographic characteristics: sex, age, formal education, income, community tenure, homeownership, household size, having difficulty walking, presence of children and presence of elderly in a household?
Hypothesis 1
a) Perceived probability mediates the influence of socio-environmental factors on behavioral intent. b) Perceived probability is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. c) Perceived probability is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami.
Hypothesis 2
a) Perceived severity mediates the influence of socio-environmental factors on behavioral intent. b) Perceived severity is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. c) Perceived severity is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami.
Hypothesis 3
a) Self-efficacy mediates the influence of socio-environmental factors on behavioral intent. b) Self-efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. c) Self-efficacy is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami.
Hypothesis 4
a) Response efficacy mediates the influence of socio-environmental factors on behavioral intent. b) Response efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. c) Response efficacy is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami.
Hypothesis 5
a) Tsunami-relevant knowledge efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. b) Tsunami-relevant knowledge efficacy is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami. c) Tsunami-relevant knowledge confidence is positively associated with self-efficacy, response efficacy, and perceived probability. d) Tsunami-relevant knowledge confidence is positively associated with response efficacy. e) Tsunami-relevant knowledge confidence is positively associated with perceived probability. f) Tsunami-relevant knowledge confidence is negatively associated with perceived severity.
Hypothesis 6
a) Hazard experience is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. b) Hazard experience is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami. c) Hazard experience is positively associated with self-efficacy. d) Hazard experience is positively associated with response efficacy. e) Hazard experience is positively associated with perceived probability. f) Hazard experience is positively associated with perceived severity.
31
Hypothesis 7
a) Risk proximity is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. b) Risk proximity is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami. c) Risk proximity is negatively associated with self-efficacy. d) Risk proximity is positively associated with response efficacy. e) Risk proximity is positively associated with perceived probability. f) Risk proximity is positively associated with perceived severity.
Hypothesis 8
a) Preparation is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami. b) Preparation is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami. c) Preparation is positively associated with self-efficacy. d) Preparation is positively associated with response efficacy. e) Preparation is positively associated with perceived probability. f) Preparation is positively associated with perceived severity.
Diagram of the research framework, indicating three levels of variables, demographic
characteristics, hypotheses, and research questions is displayed below in Figure 2.1
Figure 2.1: Research framework, indicating relationships between variables that are addressed in research questions and
hypotheses.
32
3 Background and Method
A cross-sectional research design was implemented to gauge understanding of public
behavioral intentions and factors associated with behavioral intentions and to address the proposed
research questions and hypotheses. To do so, a structured, self-administered, anonymous household
survey instrument was applied to a randomly selected sample of households in a purposively
targeted community of Seaside, Oregon.
3.1 Sample
Survey questionnaire is a popular method of data collection used to assess public knowledge
and perceptions in studies of natural hazards (Bird 2009; Wei et al. 2017; Lindell et al. 2016; Lindell et
al. 2015; Jon et al. 2016; Fraser et al. 2016; Apatu et al. 2013; Sun et al. 2013; Rofi et al. 2006). It allows
for the possibility to reach large number of people with diverse perceptions and behaviors, which is
useful when trying to understand behaviors of the general public. A large sample also yields an
opportunity to generalize results to a broader population and to use these results in general
evacuation modeling. Survey instruments also allow for comparison of the results to other research
that used similar methodologies. Targeting households is a common method in a survey
methodology because it allows researchers to reach a diverse audience by covering a wide geographic
space within a community, making the sample representative of the vulnerable population. In the
context of natural hazards, in particular, examining behavioral intentions at a household or a family
unit level is central to understanding evacuation decision-making, because families tend to reunite,
search, and account for family members before evacuating together (Fraser et al. 2013).
Seaside, Oregon was selected as a representative example of a coastal community in Oregon
with high exposure to CSZ earthquake and tsunami risks. Seaside has a large number of people who
live and work within a local tsunami inundation zone, while its flat geography increases the time
33
required for evacuation outside of the tsunami zone. Seaside is a small town on the central coast of
Clatsop County, Oregon (2010 population: 6,457, U.S. Census Bureau), located in a close proximity
to the Cascadia Subduction Zone. It is situated in an estuary at the confluence of two small rivers
that run parallel to the coast and is surrounded by hills to the south and east. Numerous bridges that
connect the community to tsunami safe zones are critical elements in the effective response to
tsunamis. A number of bridges have been seismically upgraded, but the city expects seven (out of
10) bridges to fail during a CSZ event (Raskin and Wang 2017). Location on the seismic Pacific
basin, fairly flat topography, and the location of the tsunami shelter areas at more than 1.5 km from
the shoreline makes Seaside susceptible and vulnerable to both distant and local tsunamis (Mostafizi
et al. 2017). Currently Seaside evacuation plans assume horizontal evacuation on foot as the official
tsunami evacuation policy (Mostafizi et al. 2017).
In the past, distant tsunamis with origins in Alaska (1952), Chile (1960), and Alaska (1964)
affected Seaside. The Cascadia Subduction Zone earthquake of 1700 was the latest local tsunami
event on the Oregon coast (Priest et al. 2016). In 1964, the tsunami caused 1 fatality in Seaside
(attributed to a heart attack), damaged two bridges and created an estimated damage to the city in
the amount of $41,000 and to private entities in the amount of $235,000 in 1964 dollars. In 2011 as a
result of an earthquake off the coast of Japan an official evacuation warning was issued in Seaside,
triggering mandatory city evacuation. No tsunami arrived in Seaside in 2011, but the Oregon Coast
experienced damages in the amount of $6.7 million (Natural Hazards Mitigation Plan, State of
Oregon, 2015). Today, CSZ earthquakes and tsunamis pose a great risk to the community of Seaside
and its residents (Natural Hazards Mitigation Plan, State of Oregon, 2015).
Considering the characteristics of Seaside, drawing a sample from its population can yield
interesting and useful results for understanding people’s perceptions and behavioral intentions in the
34
face of tsunami risks in other Oregon coastal communities, and perhaps coastal communities of
neighboring states of Washington and California. Communities of Aberdeen, Eureka, Crescent City,
Hoquiam, Ocean Shores, Port Townsend, and the unincorporated portions of Grays Harbor and
Pacific Counties are similar to Seaside in terms of the number of residents, employees, public
venues, dependent-care facilities, and community businesses located in tsunami hazard zones and
therefore exposed to tsunami risks (Wood et al. 2015). Moreover, because physical exposure to
tsunami risk is only one factor that is assumed to influence people’s perceptions and behavioral
intentions, in addition to non-location dependent factors of experience, knowledge, preparation, and
demographic characteristics, results of this study could be applied to other Oregon coastal
communities with more confidence.
To examine the representativeness of the survey sample of the population of Seaside, Table
3.1 compares demographic characteristics of the sample and the population of Seaside. This
information is useful for purposes of being able to apply the results of the study to the population of
Seaside. Within the sample, 45.7% are persons 65 years and older, 63.8% of respondents are home
owners, 56.7% are female, the average household is comprised of two individuals, 43.4% of the
sample have a Bachelor’s degree and higher, and the median income ranges from $40,000-$49,000.
Compared to the overall population of the study area, the survey sample included more persons 65
years and older and individuals with Bachelor’s degree or higher. Persons 65 years and older
represent almost half of the sample population, while in Seaside older persons comprise about 22%
of the city population. Individuals with the Bachelor’s degree of higher also represent almost half the
sample, while in Seaside persons with the Bachelor’s degree of higher comprise about 21% of the
population. The sample is representative of gender, household size and income, but also contains
slightly more respondents who are homeowners.
35
Table 3.1: Sample bias, comparing U.S. Census statistics for Seaside with survey respondents.
Category Census data Seaside Total Sample
Total population (18 and over) 5,163 211
Persons 65 years and older (percent of total population over 18 years) 21.8% 45.7%
Owner occupied housing units (percent of total occupied units) 44.4% 63.8%
Female (18 and over) 52% 56.7%
Average household size 2.21 2.0
Bachelor's degree or higher (population over 25, 2012-2016) 20.9% 43.4%
Median household income (in 2016 dollars), 2012-2016 $36,373 $40,000-49,000
The survey sample over-represents older populations. Therefore during the analysis the
sample is weighted by the age category, using 65 years old as a cut-off point to make the sample
more representative of the population. Weighting cannot eliminate every source of nonresponse
bias. However, random sampling combined with weighting techniques create better conditions for
unbiased results (Meyer et al. 2018).
3.2 Survey design
This research primarily follows Dillman’s et al. (2014) “Internet, phone, mail, and mixed-
mode surveys: the tailored design method” for the design and distribution of the questionnaire.
Designing a questionnaire to produce reliable and valid results is a difficult task. The order and
format of questions, the sequence of the wording and length of the questionnaire, color choices,
layout all factor in participants’ perceptions and therefore their answers. Initial recruiting letters can
also influence decisions to take a survey. Groves (2004) elaborates on written questionnaire
instrument characteristics that can give rise to errors in reporting by respondents. These errors can
36
include: failure to correctly interpret the questions, failure to recall information from the past
required to answer the questions, flawed judgments and estimates, difficulty in formatting answers
due to open ended questions, deliberate misreporting and failure to follow the instructions. Many of
these errors exist because of human cognitive limitations that arise due to various factors, including
internal and external circumstances. For instance, depending on a person’s daily mood survey
responses can vary. One cannot fully control for the above limitations. However, there are
techniques that can reduce shortcomings in survey interpretation, for instance survey visual design
and order and format of questions. They have been applied in this research questionnaire design.
The following paragraph describes the principles that help with a question’s reliability or consistency
(the probability of obtaining the same result if a question is duplicated) and a question’s validity (i.e.,
a question measures what it was intended to measure).
One set of principles focuses on the visual aspect of the survey design. Bird (2009)
recommends sequencing questions in a logical order, allowing for smooth transitions from one idea
to the next, grouping questions by theme and providing a short description of each section. Also,
sufficient space is left between questions. Research shows that spreading survey out on more pages
does not reduce the response rate. However, it helps respondents to organize information on the
page and group related information (Dillman et al. 2014). Cognitive process known as ‘encoding’,
forming memories and attitudes toward the subject of the survey, can be influenced by a decision to
insert certain questions upfront to trigger vivid memories and knowledge to help participants to
answer questions in the remainder of the survey (Groves 2004). This technique was employed in this
study by placing a short description of the M9 Cascadia earthquake and tsunami scenario at the
beginning of the survey. In addition, questions about natural disaster experiences were placed at the
beginning of the survey to trigger respondent memories. Evoking past experiences at the beginning
37
of the questionnaire can help create context, which makes it easier to answer the subsequent
questions.
Another set of principles of sound questionnaires focuses on the design of questions for
soliciting the most accurate and thorough responses. Dillman et al. (2014) recommends to avoid
vague quantifiers (e.g. “often”, “always”), pay attention to scale questions construction (e.g. inclusion
or exclusion of a “zero” point, providing “I don’t know” option, etc.), be consistent with wording
throughout the questionnaire, avoid double-barreled and long questions.
This study developed an 8-page survey questionnaire containing three sections. Section 1
asked participants about their experience with natural disasters, knowledge of tsunamis, tsunami
warning signs, emergency guidelines in the event of a tsunami, and preparation for CSZ tsunamis.
Section 2 solicited opinions about behavioral intentions and factors that may influence behavioral
intentions. Section 3 asked demographic questions. The full questionnaire, including the cover letter,
can be found in the Appendix.
3.3 Survey distribution
The survey data were collected from October-December 2017. The survey was mailed to a
random sample of 951 households in Seaside, Oregon, generated by the commercial company
Marketing Systems Group (MSG). A systematic sampling approach was applied within each
household by asking those with the latest birthday and over 18 years old to take the survey (Fraser et
al. 2016). Out of 951 valid addresses, 211 questionnaires were completed (with 14 questionnaires
completed online), resulting in a 22.2% response rate. The response rate was calculated by
comparing total number of completed questionnaires to the number of valid solicited questionnaires
(Lance and Vandenberg 2009).
38
Before being distributed to a random sample of residents in Seaside, the study questionnaire
was pre-tested on a small group of targeted individuals using paper- and online-versions of the
questionnaire. Participants consisted of graduate and undergraduate students at Oregon State
University and professionals involved in emergency preparedness on the Oregon coast. Five
responses were received from the online version of the questionnaire, allowing to test not only the
content of the survey but also the technical aspects of the online instrument. Five responses were
received from the paper-based version of the questionnaire. Pilot participants were asked to pay
attention to the soundness and clarity of the questions, ease of the visual layout (both paper and
online version), and the content of the survey (applicability and relevance of the questions).
Three waves of the survey using first-class mailings were distributed, starting with a pre-
survey postcard that advertised the project (Wolters et al. 2017). Each two consequent mailings
contained a cover letter explaining the project and participants’ rights, a copy of a questionnaire and
a business postage pre-paid return envelope. The second wave of questionnaires was sent to 813
qualifying addresses, after completed and non-deliverable addresses were removed from the
database. An additional wave with modified addresses was mailed to 167 households (they were
excluded from the second wave). Addresses were modified replacing recipients’ first and last names
to a ‘current resident’ in an attempt to reach more people and solicit more responses. Phone call
reminders were conducted using 286 phone numbers available in the database, during which the
participants were encouraged to respond to the survey. Voicemail was left in a case of non-response.
Out of 286 households that were targeted with the phone calls, 17 households completed the
questionnaire. After all waves of mailing and follow-up phone calls, 672 addresses that did not
return the questionnaire and were not contacted by phone, comprised a sample for in-person visits.
From 672 addresses 11% (98 addresses) were drawn using spatial random sampling technique via
39
ArcGIS software. These 98 addresses were visited in person over a span of two days by the author
of this study. Residents were encouraged to take the survey. A new copy of the survey was provided
if respondents indicated that they did not have one. Out of 98 households that were visited in
person, 8 returned the completed questionnaire.
3.4 Variables description and operationalization
3.4.1 Dependent variables
The behavioral intent dependent variable was measured by asking respondents to indicate
the likelihood (Very unlikely = 1 to Very likely = 5) of engaging in a set of behaviors if they were at
home in the event of a M9 Cascadia earthquake and tsunami (Table 3.2). Behavioral intent items are
the following (for each item N=206): continuing normal routine behavior (M=1.68, s.d. =1.127),
following earthquake behavior guidelines (M=3.93, s.d. =1.147), evacuating to higher ground
immediately after ground shaking (M=4.03, s.d. =1.253), waiting for tsunami confirmation for an
authority/sirens (M=2.88, s.d. =1.418), checking social media, radio, TV for additional
information(M=3.68, s.d. =1.316), contacting loved ones (M=4.1, s.d. =1.195), collecting documents
and other items (M=3.47, s.d. =1.389), travelling to gather family or friends (M=2.66, s.d. =1.365),
talking with neighbors (M=3.24, s.d. =1.365), taking the lead to evacuate and encouraging others
(M=3.57, s.d. =1.250), waiting for family at home (M=2.08, s.d. =1.194), following what friends or
neighbors are doing (M=2.03, s.d. =1.147), trying to help and rescue people (M=3.58, s.d. =1.194),
panicking (M=2.33, s.d. =1.275).
40
Table 3.2: Descriptive statistics of behavioral intentions (numbers are percentages of responses on each behavioral item)
Behavior Very
unlikely Unlikely Neither Likely
Very likely
Mean (s.d.)
Continue normal behavior 63.1 21.4 4.4 6.3 4.9 1.68 (1.127)
n=206
Follow earthquake behavior guidelines
7.3 6.8 5.3 47.1 33.5 3.93 (1.147)
n=206
Evacuate to higher ground immediately
6.8 8.3 11.2 22.8 51.0 4.03 (1.253)
n=206
Wait for tsunami confirmation from authority, sirens
22.8 24.3 9.2 29.6 14.1 2.88 (1.418)
n=206
Check social media, TV 10.2 9.2 5.3 35.0 40.3 3.86 (1.316)
n=206
Contact loved ones 6.8 6.8 5.3 31.6 49.5 4.1 (1.195)
n=206
Collect documents, other items 10.1 24.6 6.8 27.1 31.4 3.47 (1.389)
n=206
Travel to gather family, friends 23.8 30.6 15.0 17.0 13.6 2.66 (1.365)
n=206
Talk with neighbors 18.9 10.7 13.6 40.8 16.0 3.24 (1.365)
n=206
Take the lead to evacuate and encourage others
9.7 12.6 12.6 41.3 23.8 3.57 (1.250)
n=206
Wait for family at home 47.1 18.9 16.5 13.6 3.9 2.08 (1.194)
n=206
Follow friends or neighbors 43.7 27.2 14.6 11.7 2.9 2.03 (1.147)
n=206
Try to help and rescue 8.7 9.7 18.9 39.8 22.8 3.58 (1.194)
n=206
Panic 34.5 28.6 12.1 19.4 5.3 2.33 (1.275)
n=206
Exploratory factor analysis was used to form behavioral intent scales. Factor analysis is often
used to represent latent concepts and for dimension reduction in human dimensions studies (Vaske
2008). In this study maximum likelihood extraction with a promax rotation was used because it is
more rigorous. Items whose communalities were less than 0.3, and items that cross-loaded on
several factors and loaded below 0.5 were removed from the factor analysis. The analysis yielded two
41
factors. The KMO (Kaiser-Meyer-Olkin) measure of sampling adequacy was 0.612 and significant.
Cumulative variance explained was 47.8%. This is an acceptable outcome for small number item
scales.
Two components with eigenvalues ≥1 (1.966 and 1.429) were extracted as a result of factor
analysis as displayed in Table 3.3. Factor loadings greater than 0.5 were chosen as a minimum cut off
(Field 2009). The eigenvalue is a measure of the amount of variance within the observed variable
that a factor explains. Eigenvalue is equivalent to the number of variables that a factor represents.
Therefore, any factor with an eigenvalue ≥1 explains more variance than a single observed variable.
Cronbach’s alphas, the measure of scale reliability, were 0.638 and 0.700, respectively. Cronbach’s
alpha at 0.6 is acceptable for scales with few items.
Table 3.3: Pattern matrix for behavioral intent
Behavioral intent Leadership Delay
Evacuate to high ground immediately after an earthquake 0.600 0.002
Take the lead to evacuate and encourage others 0.777 0.031
Check social media, TV -0.165 0.793
Contact loved ones 0.127 0.633
Collect documents and other items 0.088 0.583
Cronbach’s alpha 0.638 0.700
Two behavioral intent indices were formed from the arithmetic mean of items in each factor
weighted by factor scores (DeVellis 2016): leadership behavioral intent index (M= 2.59, s.d. = 0.74,
N = 206) and delay behavioral intent index (M= 2.56, s.d. = 0.69, N= 206). Leadership behavioral
index represents behavioral intentions of immediate evacuation outside of the tsunami zone
following an earthquake, while delay behavioral index represents behavioral intentions that involve
42
contacting others, checking other sources of information for tsunami confirmation and collecting
household items, i.e. delaying the start of evacuation.
3.4.2 Independent cognitive variables
Perceived severity was measured by asking respondents about the likelihood (Very unlikely =
1 to Very likely = 5) that CSZ earthquakes and tsunamis will: lead to injuries to you or your family
(M=3.85, s.d. =1.041, N =206); create a life-threatening situation for you or your family (M=4.10,
s.d. =0.950, N =206); create a life-threatening situation for people in your town; severely damage or
destroy your home (M=4.26, s.d. =1.020, N =206); destroy or severely damage roads, homes, etc. in
your town; and create a severe financial burden for you or your family (M=4.12, s.d. =1.017, N
=206) (Wei et al. 2017; Lindell et al. 2016; Lindell and Hwang 2008).
Due to kurtosis and skewness, “damage to town” and “damage to people in town” items
were deleted, because they showed high kurtosis of 16 and 11.6, respectively. “Damage to people in
town” item showed skewness level of -3.45. The two items with high kurtosis can be removed
without undermining reliability of the scale due to the other four items still remaining.
Exploratory factor analysis was used to form a perceived severity index. One component
with an eigenvalue ≥1 (2.678) was extracted. Factor loadings greater than 0.5 were chosen as a cut
off. Scale reliability Cronbach’s alpha is 0.830, as displayed in Table 3.4. The KMO measure of
sampling adequacy is 0.753 and significant. Cumulative variance explained is 57.7%, which is
acceptable for a four-item scale.
Table 3.4: Factor matrix of perceived severity
Component matrix Perceived severity
Lead to injuries 0.855
Create life-threatening situations 0.925
Damage or destroy your home 0.684
Create financial burden 0.504
Cronbach’s alpha 0.830
43
The perception of risk index is formed from the arithmetic mean of items weighted by factor
scores (M=3.02, s.d. = 0.62, N =206).
Perceived likelihood was measured by asking respondents about the likelihood (Very unlikely
= 1 to Very likely = 5) that CSZ earthquakes and tsunamis will happen in the next 50 years (M= 3.76,
s.d. = 1.085, M = 206).
Self-efficacy was measured by asking respondents about their confidence (Not at all
confident=1 to Very confident=5), given their personal skills and resources, to effectively protect
themselves and reduce the chance of severe injuries in the event of a M9 Cascadia earthquake and
tsunami (M= 2.83, s.d. = 1.114, M = 206).
Response efficacy was measured by asking respondents about their agreement (Strongly
disagree = 1 to strongly agree = 5) with whether people can increase their chances of surviving a M9
Cascadia earthquake and tsunami if they adopt emergency plans and protective actions (M= 4.14,
s.d. = 0.995, M = 206).
3.4.3 Independent socio-environmental variables
Experience is typically measured by asking respondents to indicate the degree of experience
with disasters under investigation. It is also often measured as records of casualties and damages
experienced by respondents and their family members in similar situations in the past (Jon et al.
2016; Lindell and Hwang 2008; Lindell and Prater 2000). For example, Jon et al. (2016), in a study of
responses to earthquakes in New Zealand and Japan, asked respondents to report their experience
with earthquakes prior to the one being studied by indicating the degree of damage and danger
previous earthquakes posed to their community, personal property, and personal and family health.
Seaside has not experienced severe earthquakes and a tsunamis in the past. Therefore, limiting
measurement of experience to only these events will not create enough variation in the results.
44
Instead, the questionnaire asked respondents to indicate whether they have ever experienced any
type of natural disaster, for instance floods, earthquakes and wildfires (No=1, Not sure=2, Yes=3).
The result was coded as a count variable, the number of disasters a person has experienced in their
lifetime. The range of the number of disasters experienced by respondents in the survey was from 0
to 7 (M= 1.34, s.d. = 1.467, N=206, Range = 0-7).
To assess confidence in tsunami-related knowledge respondents were asked about their
confidence (Not at all confident=1 to Very confident=5) in their knowledge in the following items: the
difference between local and distant tsunami events (M= 3.82, s.d. = 1.065, M = 206); natural
warning signs of tsunamis (M= 3.63, s.d. = 1.139, M = 206); emergency warning messages of
tsunamis (M= 3.99, s.d. = 1.106, M = 206); where to get the information about preparation for
tsunamis (M= 3.79, s.d. = 1.249, M = 206); what to do when there is a warning about a tsunami
(M= 3.98, s.d. = 1.079, M = 206); tsunami evacuation routes from their homes (M= 4.20, s.d. =
1.070, M = 206); tsunami safe places close to their homes (M= 3.92, s.d. = 1.354, M = 206). The
summary of responses to each survey knowledge item are displayed in Table 3.5.
45
Table 3.5: Summary of hazard-relevant knowledge survey responses
Percentage
Items Not at all
confident
A little
confident
Somewhat
confident Confident
Very
confident Mean (s.d.)
Local vs. distant
tsunami 3.9 6.8 23.3 35.4 30.6 3.82 (1.065)
Natural warning
signs 6.3 9.2 24.3 35.4 24.8 3.63 (1.139)
Emergency warning
messages 4.4 7.8 12.1 36.4 39.3 3.99 (1.106)
Sources of
information for
preparation
8.3 9.2 13.1 34.0 35.4 3.79 (1.249)
Government
recommendations 13.1 13.1 21.4 27.7 24.8 3.38 (1.337)
What to do when
there is a warning 3.9 5.8 18.4 32.5 39.3 3.98 (1.079)
Evacuation routes 3.4 5.8 11.2 26.7 52.9 4.20 (1.070)
Tsunami safe places 11.2 5.8 11.2 23.8 48.1 3.92 (1.354)
Exploratory factor analysis examined dimensions of the tsunami-relevant knowledge
confidence scale. One component with an eigenvalue ≥1 (5.427) was extracted as a result. Factor
loadings greater than 0.5 were chosen as a cut off. Cronbach’s alpha is 0.929, as shown in Table 3.6.
KMO measure of sampling adequacy is 0.919 and significant. Cumulative variance explained is
63.4%.
Table 3.6: Factor matrix for knowledge confidence
Component matrix Knowledge perception
Local vs. distant tsunami 0.693
Natural warning signs 0.802
Emergency warning messages 0.826
Sources of information for preparation 0.854
Government recommendations 0.783
What to do when there is a warning 0.906
Evacuation routes 0.778
Tsunami safe places 0.707
Cronbach’s alpha 0.929
46
A knowledge confidence index was formed from the arithmetic mean of items weighted by
factor scores (M=3.05, s.d. = 0.76, N=206).
Proximity or exposure to risk was measured in conjunction with the tsunami wave arrival
time findings by Priest et al. (2016). ArcGIS 10.5 was used to geocode household data from the
survey and then merge the spatial data (i.e., tsunami inundation zone, roads, city boundaries and
tsunami wave arrival time) into GIS layers. Twenty-one respondents out of 211 live outside of M9
CSZ tsunami inundation zone. West coast ocean data portal provided geospatial data of Oregon
tsunami inundation zones for XXL1 scenario (9.0. Magnitude earthquake) developed by Priest et al.
(2013). City of Seaside provided the geospatial data for Seaside city roads and city boundaries.
Department of Geological and Mineral Services (DOGAMI) provided the geospatial data
for CSZ tsunami advance time in the community of Seaside. Geospatial data was set in GCS North
American 1983 HARN geographic coordinate system and NAD 1983 HARN Lambert Conformal
Conic projected coordinate system. Using the Spatial Analyst tool “Extract values to points” in
ArcGIS, each household was assigned a wave arrival time (in minutes) using spatial location of each
household as a reference point (M=25.4, s.d. = 4.11, N = 206, Range = 19.3 to 32.9 minutes). The
resulted map, displayed as Figure 3.1, shows the community of Seaside: city boundaries and streets,
tsunami inundation zone for M9 CSZ scenario, tsunami wave arrival time in different shades of blue,
and households displayed in orange. Black and green circles represent bridges, where black indicated
bridges are expected to fail in the CSZ earthquake, while green indicated retrofitted bridges may
survive in the CSZ earthquake. Wave arrival serves as a proxy for assessment of actual risk. Only
households that are located within the tsunami inundation zone were assigned measure of tsunami
wave arrival time (a characteristic of the wave arrival time data file). Therefore, observations outside
of the inundation zone (N=21) originally had missing data for the wave arrival time. To avoid
47
missing data, these observations were given a value of 34 minutes, because the highest value of the
existing wave arrival data was 33 minutes. Additional values do not negatively impact the analysis.
On the other hand, they support the conceptual idea of risk proximity.
Figure 3.1: Map of Seaside and Tsunami Wave Arrival Time
48
Preparedness was measured by asking respondents to mark all the measures that they have
undertaken in the past 2 years: attended a meeting or received written information on how to better
prepare for Cascadia earthquakes and tsunamis (52.2% - Yes, 47.8% - No); prepared an ‘emergency
supply kit’ (stored extra food, water, batteries, etc.) (47.4% - Yes, 52.6% - No); discussed the topic
of preparedness for Cascadia earthquakes and tsunamis with others in the community (47.4% - Yes,
52.6% - No); developed an ‘emergency plan’ with family or friends in order to decide what everyone
would do in the event of Cascadia earthquakes and tsunamis (44.5% - Yes, 55.5% - No); developed a
communication plan (e.g. a plan that establishes how to get in touch with each other if meeting at
home is not possible and phones do not work) (21.1% - Yes, 78.9% - No); attended First Aid or
Cardio-Pulmonary Resuscitation (CPR) training (32.5% - Yes, 67.5% - No); and identified an
emergency contact person outside of the Northwest (27.8% - Yes, 72.2% - No). The preparation
variable was coded as a count variable, the number of measures that a respondent has undertaken
(M = 2.83, s.d. = 2.014, N = 206, Range 0-8). Study followed Dunn et al. (2016) who evaluated
connection between preparedness and perceptions of risks in a survey in Washington, Oregon, and
California regarding people’s perceptions of earthquake hazards and earthquake early warning
systems.
Table 3.7 displays the summary statistics for two dependent variables and both cognitive
constructs and socio-environmental factors. Based on the results, on average respondents are as
likely to have intentions to engage in the leadership behavior (M=2.59) as in the delay behavior
(M=2.56). They also show above average perception of danger (M=3.02) and perceptions of the
likelihood of CSZ tsunami occurrence in the next 50 years (M=3.76). Respondents’ confidence in
being able to protect themselves in the event of the disaster centers on the medium of the scale
(M=2.83), while their belief in the efficacy of preparation is high (M=4.14). In terms of experience,
49
results show that respondents have minimal prior experience with natural disasters, with 1.34
average number of natural disasters experienced in the past. Respondents’ confidence in tsunami-
relevant knowledge is just above the mean (M=3.05). Respondents’ households that are within local
tsunami inundation zone are located within 19.3 to 32.9 minutes of tsunami wave arrival time after
the earthquake. Respondents’ average preparedness is M= 2.83 measures undertaken in the past 2
years.
Table 3.7: Summary of dependent and independent variables
Variable Description Mean (s.d.)
Leadership
behavior
Leadership behavior index based on weighted average
1=very unlikely to 5=very likely
2.59 (0.74)
n=206
Delay behavior Delay behavior index based on weighted average
1=very unlikely to 5=very likely
2.56 (0.69)
n=206
Perceived severity Index based on assessment of 4 risk statements
1=very unlikely to 5=very likely
3.02 (0.62)
N=206
Perceived
likelihood
Perceived likelihood of disaster occurrence in the next 50 years
1=very unlikely to 5=very likely
3.76 (1.085)
N=206
Self-efficacy Confidence in being able to protect yourself in the disaster
1=not at all confident to 5=very confident
2.83 (1.114)
N=206
Response efficacy Preparation can increase chances of surviving the disaster
1=strongly disagree to 5=strongly agree
4.14 (.995)
N=206
Experience Number of natural disasters experienced
[Range: 0-7]
1.34 (1.467)
N=206
Knowledge
confidence
Index based on assessment of 8 knowledge statements
1=not at all confident to 5=very confident
3.05 (0.76)
N=206
Wave arrival M9 Cascadia tsunami wave arrival time
[Range: 19.3 to 32.9 minutes]
25.4 (4.11)
N=206
Preparation Number of preparatory disaster measures already taken
[Range: 0-8]
2.83 (2.014)
N=206
50
3.4.4 Demographic variables
Gender was measured by asking respondents to indicate which sex they identify with
(Female=1, Male=0). Age was measured as a continuous variable by asking respondent to fill in their
age. Household size was measured by asking respondents to indicate the number of people and their
ages who are currently living in their household. Based on the answers, two dummy variables were
constructed, whether there were children of 10 years and younger (Yes=1, No=0) and elderly of 65
years and older (Yes=1, No=0) present in a household. Homeownership was measured by asking
respondents whether they are primary owners=1 of the residence where they received the survey,
whether they are renters or occupiers=2, or other=3. Community tenure was measured as a continuous
variable by asking respondents to indicate how many years they have lived in the community of
Seaside (0 if less than a year). The level of formal education was measured by asking respondents to
indicate the highest year of formal education completed (Less than high school diploma=1 to Graduate
degree or other professional degree=7). Household income was measured by asking respondents to indicate
which category best described their household income before taxes in 2016 (Less than $10,000=1 to
$75,000 or higher=9). Table 3.8 displays summary statistics of demographic variables.
The sample is almost evenly split between female and male respondents. The average age of
respondents is M=60.81. Average household size is M=2.01. There are almost no respondents with
children of 10 years old and younger present in a household. There are more households with the
presence of elderly. The majority of the respondents were homeowners. The average community
tenure of respondents is M=20.79, while the mean income of respondents equals to M=$40,000-
$49,000.
51
Table 3.8: Summary of demographic variables
Variable Description Mean (s.d.)
Gender Female=1, Male=0 0.59 (0.494)
N=206
Age Continuous variable 60.81 (15.18)
N=206
Household size Total number of people living in a household 2.01 (1.055)
N=206
Presence of
children
Presence of children 10 years and younger
Yes=1, No=0
0.09 (0.283)
N=206
Presence of
elderly
Presence of elderly 65 years and older
Yes=1, No=0
0.23 (0.424)
N=206
Homeownership Primary owner=1, Renter=2, Other=3 1.35 (0.537)
N=206
Community
tenure Continuous variable
20.79 (18.388)
N=206
Education Less than high school diploma=1 to Graduate degree or other
professional degree=7
4.87 (1.648)
N=206
Income Less than $10,000=1 to $75,000 or higher=9 5.99 (2.547)
N=206
Difficulty walking Yes=1, No=0 0.27 (0.446)
N=206
3.5 Data cleaning
Several of the sample observations had more than 20% of data missing. Five observations
with missing data were deleted. Two of the five were missing 21 values, two were missing 17 values,
and one was missing 12 values. These responses were deemed significantly incomplete, since the
missing data included important independent and dependent variables such as behavioral intentions,
risk and knowledge perception questions. The deletion led to a working sample of 206 observations.
The sample also had 35 variables with missing values. However, all variables had less than 5 percent
of data missing. Therefore, the missing data was imputed using the median value for ordinal scales
52
and the mean for continuous scales. One of the main reasons for imputing missing values is to be
able to do a mediation analysis explained later in this paper, which calls for no missing values.
Moreover, preserving observations in a small sample gives the results more power. Income has the
most number of missing values—15, representing 7.3% of the sample. If the proportion was more
than 10%, then imputation could cause problems, because it would dilute data, bringing it closer to
the mean.
For variables that are measured on scales from 1-5 using several items within one question,
there is a possibility that respondents are not engaged and answer all questions with the same
answer. I checked answers for knowledge and perceived severity measures by examining standard
deviations of each scale. There were 7 observations with a standard deviation of 0, which means
respondents provided exactly the same answer for all questions. There were also 17 observations in
which standard deviation ranges from 0.25 to 0.4. However, behavioral scale for these observations
displays variation, indicating that these respondents were not necessarily unengaged. There is a
possibility that some scales were not constructed in a way to allow room for variation in answers.
These observations were not deleted, with the sample size preserved at 206 observations.
Outliers – unusual observations – that can cause the most problem on continuous variables
that are characterized by a wide range of values that can qualify as a valid answer. In contrast, scale
questions that are limited to 1-5 range are unlikely to have outliers. In this study, there is one
continuous variable where outliers were possible—community tenure. Outlier values can be defined
as any value beyond 3 standard deviations of the mean (Vaske 2008). Community tenure did not
show a presence of outliers. Therefore, no replacement or deletion was performed.
Examination of skewness and kurtosis allows to check if observations have a normal distribution.
Values over 3 and below -3 were chosen as a cut off. Any value above the absolute value of 3 is
53
considered problematic. Behavioral item, “go towards the ocean,” had a kurtosis of 8.8 and
skewedness just barely over 3. It is acceptable to remove this item from the behavioral intentions
scale because there are many items in that scale to preserve variation in the data. Household size had
kurtosis of 3.47 and presence of children had a kurtosis of 6.7. If they show unusual behavior later,
it could be related to high kurtosis.
The sample was weighted to account for overrepresentation of adults over 65 years old. The
weight of 0.47 was assigned to respondents 65 years and older and a weight of 1.44 was assigned to
respondents of 64 years and younger. During the analysis both weighted and unweighted samples
were compared.
3.6 Analytical procedures
To answer the first research question of the distribution of behavioral intentions in the event
of a local tsunami in the population under the threat of M9 CSZ tsunami, the results section
provides a descriptive statistics table which displays percentage variation among behavioral items.
To evaluate hypotheses 1-8 (i.e. to understand relationships between socio-environmental,
demographic, cognitive, and behavioral intentions variables), Ordinary Least Squares (OLS) linear
regression and mediation analyses were performed. OLS regression was used primarily for an initial
overview of the data. OLS regression that includes all independent variables in one model does not
allow to fully assess the mediating role of the cognitive variables. Therefore, mediation analysis using
PROCESS package in SPSS (Hayes 2012) was utilized to examine the effect of socio-environmental
variables on behavioral intent directly and via cognitive constructs of perceptions. To answer the
third research question of the variation in behavioral intent within the following demographic
characteristics, analysis of variance (ANOVA) was applied. The next chapter 4 talks about analyses
and results.
54
4 Results
4.1 Behavioral intentions descriptive statistics
To answer research question #1, figure 4.1 displays the distribution of behavioral intentions
in the event of a local tsunami in the population under the threat of M9 CSZ tsunami. In the survey
respondents were asked how likely they were to adopt each of the following behaviors if they were
at your home in the event of a M9 Cascadia earthquake and tsunami? (1=very unlikely to 5=very
likely). The graph below combined “very unlikely” and “unlikely” responses into “unlikely” and
“very likely and “likely” into “likely”. According to the graph, the top three behaviors that the
respondents are unlikely to engage in include: continuing normal routine behavior (84.5%),
following friends and neighbors (70.9%), and waiting for family at home (66%).The top three
behaviors that respondents indicate they are likely to engage in include: contacting loved ones
(81.1%), following earthquake behavioral guidelines (80.6%), and checking social media, TV (75.3
%). Relevance and explanation of these results are discussed in detail in chapter 5.
55
Figure 4.1 Descriptive statistics of behavioral intentions
0 10 20 30 40 50 60 70 80 90
Panic
Continue normal behavior
Follow earthquake behavior guidelines
Try to help and rescue
Wait for family at home
Follow friends or neighbors
Wait for tsunami confirmation
Talk with neighbors
Travel to gather family, friends
Check social media, TV
Contact loved ones
Collect documents, other items
Take the lead to evacuate and encourage others
Evacuate to higher ground immediately
Likely Neither Unlikely
56
4.2 OLS regression analysis
To evaluate hypotheses 1-8, OLS regression analysis was performed first, using two
dependent variables: leadership behavior and delay behavior. Linear regression analysis, testing
model specifications that include all independent variables simultaneously, was applied for data
exploratory purposes. Yet, this is not the main technique for hypotheses testing in this research. The
goal of this study is to understand relationships between different levels of variables (i.e. socio-
environmental vs. cognitive vs. behavioral intentions), which are difficult to assess with a simple
linear regression. However, OLS regression results can provide a general understanding of the
association between behavioral intentions and explanatory variables. Regression output for
leadership dependent variable is displayed in Table 4.1. Two sets of regressions, using weighted and
unweighted samples, were conducted. Two model specifications, with and without demographic
variables, were examined. Standardized beta coefficients were reported to compare the effects
among variables.
57
Table 4.1: OLS regression output for leadership dependent variable (coefficients are standardized)
Variable Unweighted 1 Unweighted 2 Weighted 1 Weighted 2
Perceived severity 0.101 0.110 0.137 0.167*
Perceived probability 0.012 0.002 -0.001 -0.015
Self-efficacy 0.064 0.092 0.062 0.082
Response efficacy 0.142 0.147* 0.153* 0.172*
Wave arrival -0.151* -0.157* -0.123 -0.115
Experience -0.041 -0.049 -0.041 -0.061
Preparation 0.178* 0.192* 0.156* 0.182*
Knowledge confidence 0.133 0.084 0.136 0.076
Age 0.040 0.017
Difficulty walking -0.100 -0.062
Household size 0.023 -0.040
Children present 0.096 0.135
Elderly present -0.010 0.005
Ownership 0.060 0.044
Community tenure -0.078 -0.066
Education -0.090 -0.104
Income 0.016 0.080
Gender 0.152* 0.186**
N 206 206 206 206
F-statistic 2.94*** 4.48*** 2.99*** 4.28***
R-squared 0.220 0.154 0.223 0.148
Adjusted R-squared 0.145 0.119 0.148 0.113
* p<.05; ** p<.01; *** p<.001
There is only a slight difference in the outcomes between weighted and unweighted samples.
In the unweighted sample models, wave arrival time, preparation, and gender show a significant
effect on the outcome. In the weighted sample, the effect of perceived severity becomes significant,
while wave arrival time loses its significance. The effect of perceived severity in the weighted sample
is positive and significant, meaning that higher level of perceived risk is associated with higher
likelihood of having an intention to adopt the leadership behavior. This outcome is consistent with
previous empirical findings that shows the importance of risk perception in influencing behavior.
58
The effect of response efficacy in both weighted and unweighted samples is positive and significant,
meaning that higher level of response efficacy is associated with higher likelihood of having an
intention to engage in the leadership behavior. Significance of response efficacy and non-significance
of self-efficacy, shows that the perception of the effectiveness of protective actions is more
influential in leading to immediate evacuation decisions than the perception of one’s ability to do so.
In the unweighted sample the effect of wave arrival time (or risk proximity) is negative and
significant, meaning that increase in wave arrival time (a household is located further away from the
coastline) is associated with lower likelihood of having an intention to engage in the leadership
behavior. This outcome makes logical sense. It could be a reflection of the fact that people living
further away from the coastline have more time to escape the tsunami wave and more time to
engage in other types of behavior before evacuating. Effect of preparation, in both weighted and
unweighted samples, is positive and significant, suggesting that greater level of preparedness is
associated with higher likelihood of having an intention to engage in the leadership behavior. This
outcome shows that those who have done certain preparations (e.g. made a first-aid kit, established a
communication plan with friends and family) are ready to evacuate immediately. Gender is positive
and significant in both weighted and unweighted samples, meaning that being female is associated
with a higher likelihood of having an intention to adopt leadership behavior.
Beta coefficients in the regression output in Table 4.1 may not display their true value
and/or significance because of the possibility of an omitted variable bias or multicollinearity
problems, caused by correlation between variables. Pearson’s correlations between independent
variables are reported in Table 4.2 and between demographic variables are reported in Table 4.3.
Correlations between variables do not invalidate results. In fact, the research hypotheses assume a
presence of association between them, especially between socio-environmental and cognitive
59
variables. For instance, Table 4.2 shows that self-efficacy and knowledge are significantly and
strongly correlated (r = 0.5337, p < 0.001), and so are self-efficacy and preparation (r = 0.4523, p <
0.001). However, presence of correlation makes it more difficult to evaluate variables’ true influence
on the dependent variable when all of them are present in the model. That is why using mediation
analysis that relies on a series of hierarchical regressions could yield more clear results. In addition,
Table 4.3 displays Pearson’s correlations between the demographic variables. The output shows, for
instance, that difficulty walking is associated with age (r = 0.3764, p < 0.05), meaning that older
respondents are more likely to have walking difficulties than younger respondents. Presence of
children in a household and a household size are positively correlated as well (r = 0.6344, p < 0.05).
However, it is important to remember that both variables display high kurtosis and could be
showing unreliable behavior.
60
Table 4.2: Pearson’s correlations between independent variables
Variable Self-efficacy Response
efficacy
Perceived
severity
Perceived
probability Knowledge Experience Wave arrival Preparation
Self-efficacy - - - - - - - -
Response efficacy 0.404*** - - - - - - -
Perceived severity -0.2676*** 0.0091 - - - - - -
Perceived probability -0.0286 0.099 0.2104** - - - - -
Knowledge 0.5337*** 0.2282*** -0.0332 0.0175 - - - -
Experience 0.2247** 0.0483 0.0185 -0.0338 0.2331*** - - -
Wave arrival 0.2725*** 0.0789 -0.3448*** 0.1084 0.1557* 0.0147 - -
Preparation 0.4523*** 0.2217*** 0.0017 0.0769 0.5405*** 0.2583*** 0.1941** -
* p<.05; ** p<.01; *** p<.001
61
Table 4.3: Pearson’s correlations between demographic variables
Variable Age Difficulty
walking
Household
size
Children
present
Elderly
present Ownership
Community
tenure Education Income
Difficulty walking 0.3764**
Household size -0.4096** -0.1509**
Children present -0.4264** -0.1504** 0.6344**
Elderly present 0.2867** 0.0504 0.1259* -0.1705**
Ownership -0.0793 0.1864** -0.1267* 0.0520 -0.2574**
Community tenure 0.2853** 0.0594 0.0328 -0.1529** 0.1905** -0.2383**
Education 0.0200 -0.1057 -0.0947 -0.0913 0.0563 -0.1476** -0.1810**
Income -0.0669 -0.3154** 0.2343** 0.0485 0.1738** -0.3755** -0.0060 0.3077**
Gender -0.0735 0.0245 0.0359 0.0498 -0.0512 0.1311* -0.1081 -0.0284 -0.0692
** p<.05 or better
62
Table 4.4 shows the output of OLS regression analysis for unweighted and weighted models
respectively for delay dependent variable. Compared to leadership behavior regression output, there
are fewer variables that show significant influence on the dependent variable.
Table 4.4: OLS regression output for delay dependent variable (coefficients are standardized)
Variable Unweighted 1 Unweighted 2 Weighted 3 Weighted 4
Perceived severity -0.105 -0.087 -0.120 -0.078
Perceived probability -0.068 -0.055 -0.104 -0.097
Self-efficacy 0.134 0.145 0.060 0.091
Response efficacy 0.066 0.032 0.053 0.012
Wave arrival -0.046 -0.033 -0.007 0.005
Experience -0.086 -0.113 -0.109 -0.150*
Preparation 0.005 0.035 -0.049 -0.007
Knowledge
perception
-0.040 -0.055 0.008 -0.025
Age 0.701 0.009
Difficulty walking -0.067 -0.087
Household size 0.285** 0.286**
Children present -0.097 -0.126
Elderly present -0.075 -0.024
Ownership -0.004 0.009
Community tenure -0.067 -0.074
Education -0.095 -0.081
Income -0.049 -0.052
Gender 0.174* 0.199**
N 206 206 206 206
F-statistic 1.66* 1.17 1.93* 1.24
R-squared 0.138 0.045 0.156 0.048
Adjusted R-squared 0.055 0.007 0.075 0.009
* p<.05; ** p<.01; *** p<.001
In Table 4.4 in the unweighted model household size and gender show significant positive
association with delay intent dependent variable. Interestingly, for both leadership and delay
dependent variables gender shows significant positive influence on the outcome, suggesting that
females are more likely than males to have both leadership and delay intent. The reason that gender
63
variable shows such an outcome, could be reflective of the fact that females are more aware of risks
and the need to act proactively, but they also feel the importance of protecting their families (Shapira
et al. 2018). In this study female gender is also associated with higher level of perceived severity
(r=0.171, p < 0.05) and lower level of self-efficacy (r = -0.142, p < 0.05). Being female is the only
demographic characteristic that is consistently shown to be associated with higher levels of risk
perception and the adoption of protective actions in the natural hazard literature (Lindell et al. 2015;
Terpstra and Lindell 2012; Barberi et al. 2008; Fothergill 1996; Griffin et al. 1999; Lindell and Hwang
2008; Peacock et al. 2005). In the weighted model, household size and gender variables perform
similarly to the unweighted model.
In addition, experience shows significant negative association with the delay intent
dependent variable, meaning that higher number of natural disasters experienced in the past is
associated with lower likelihood of delay intent. This outcome suggest that perhaps experiential
knowledge plays a role in influencing people’s intentions to avoid delay behaviors. However,
experience variable did not show the opposite influence on the leadership intent, suggesting that
respondents who experienced natural disasters in the past are less likely to have a delay intention,
but are not necessarily more likely to have a leadership intention. Relationships between variables are
going to be analyzed in more details in the following section relaying on the mediation analysis
technique. OLS regression analyses show significant influence of some variables on behavioral
intentions, more specifically response efficacy, preparation, wave arrival time, gender, and household
size, supporting findings of previous research. Therefore, this research framework has value and is
going to be analyzed in more details in the following section.
64
4.3 Mediation analysis
To evaluate research hypotheses 1-8 regarding the relationships between socio-
environmental and cognitive perceptions variables and behavioral intent (leadership behavior and
delay behavior), a mediation analysis approach suggested by Preacher and Hayes (2004) was used. A
simple mediation model, the most commonly employed type of a mediation model, demonstrates a
sequence in which independent variable X can affect dependent variable Y indirectly through
mediator variable M and directly without the influence of M. Baron and Kenny (1986) define
mediator as a mechanism through which central independent variable influences the dependent
variable of interest, partially or fully accounting for the effect of a predictor on an outcome. The
impetus for using mediation analysis in this study is that it has the potential for improving
understanding of the mechanisms through which behavioral intent is formed, evaluating whether
cognitive variables partially or fully account for the influence of socio-environmental variables on
the outcome and through which socio-environmental variables have the most influence on the
outcome (Zanocco et al. 2017).
It is important to note that using mediation analysis does not improve the process of
assessing a causal effect. Mediation, as a process, is a causal phenomenon, but a statistical model
cannot prove causality. Especially in simple mediation models it is difficult to avoid key effects from
being confounded by omitted variables (Haynes and Preacher 2014). Yet, mediation statistical
analysis can help illuminate certain alternative explanations of phenomena. It has been chosen for
this study because perceptions and attitudes are hypothesized to be core components of cognition
that translate outside information into behavioral intent, serving as mediators between observable
factors and behavioral intent (Ajzen and Fishbein 2005).
65
OLS regression analyses conducted earlier shows the influence of socio-environmental,
demographic, and cognitive variables on behavioral intent. Yet, a model that contains all variables
does not evaluate the full complexity of the relationship between independent variables and between
independent and dependent variables that exists as suggested by theories of behavior. In addition,
the OLS regression independent variables may not be displaying their true effects and significance
because of possible multicollinearity. Mediation analysis also relies on the OLS estimator to
understand effects of independent variables on the outcome.
Using diagrammatic conventions common to mediation analysis (Preacher and Hayes 2004),
a generalized model of the relationship between socio-environmental factors, cognitive factors, and
behavioral intent is presented in Figure 4.2. and 4.3. The simple relationship between X and Y is
known as the total effect (path c) (see Figure 4.2).
Figure 4.2. Illustration of a total effect of X on Y
Behavioral
intentions
(Y)
c
Socio-
environmental
factors
(X)
66
Figure 4.3: Illustration of a simple diagram of mediation
In Figure 4.2 path c is called the direct effect of X on Y. The direct effect is estimated
through equation (1), where c’1 coefficient is the direct effect of X on Y when controlled for the
mediator. It quantifies how much variations in the levels of X account for variations in Y,
independent of the effect of M on Y, i.e. it represents the effect of X on Y that is unique to X.
Y=iY+c1’ X+b1M+eY (1)
M=iM+a1X+eM (2)
The indirect effect of X on Y is represented as the two paths linking X to Y through M
(paths a and b in Figure 4.3). The indirect effect of X on Y through M is quantified as the product of
a and b paths (a*b). It is estimated as a1b1, meaning the product of the effect of X on M equation (2)
Perceptions
(M)
Behavioral
intentions
(Y)
c’
Socio-
environmental
factors
(X)
Model effects Direct effect of X on Y, while controlling for M -- (c’) Direct effect of M on Y, while controlling for X -- (b)
Indirect effect of X on Y -- (ab)
Model variables Primary independent variable (X) Mediator variable (M) Dependent variable (Y)
67
and the direct effect of M on Y while controlling for X equation (1) (Baron and Kenny 1986, Vaske
2008).
A widely accepted approach developed by Baron and Kenny (1986) argues that for a
mediation to exist three criteria must be met: (1) X significantly predicts Y (i.e. total effect is
significant); (2) X significantly predicts M (path a); and (3) M significantly predicts Y controlling for
X (path b) (Preacher and Hayes 2004). Mediation occurs when the effect of X on Y decreases (effect
c’) with the inclusion of M, either to zero (full mediation) or by a nontrivial amount (partial
mediation). Baron and Kenny (1986) also argue that to claim mediation there should be no
measurement error in the mediator and Y should not cause M, the former being a difficult criterion
to meet considering the latent nature of cognitive variables.
Preacher and Hayes (2004) have argued for the necessity to perform a formal and direct
significance test of the indirect effect (ab) rather than following a series of separate significance tests
that do not involve effect ab, as described by Baron and Kenny (1986). One such direct tests has
been developed by Sobel (1982). The Sobel test provides a direct test of the indirect effects (ab) and
has been argued to be a more statistically rigorous test of the presence of indirect effects. It relies on
unstandardized regression coefficients of paths a and b and the standard errors associated with those
coefficients. However, few researchers have relied on the Sobel test compared to following the
criteria for mediation established by Baron and Kenny (1986), where significance of the indirect
effect is not a required condition for mediation. Preacher and Hayes (2004) argue that another
reason for the unpopularity of the Sobel test lies in the fact that common programs for regression
analysis such as SPSS and SAS would provide all the information needed to conduct the Sobel test
manually, but researchers would not bother to make additional computational efforts.
68
Hayes (2013) developed an alternative to the Sobel test, examining the significance of the
indirect effects using bootstrapped confidence intervals that can be executed in the PROCESS
package in SPSS, making it easier for researchers to implement. Bootstrapping is a nonparametric
approach to effect-size estimation and hypothesis testing that makes no assumption about the shape
of the distribution of the indirect effect (Bollen and Stine, 1990; MacKinnon and Dwyer, 1993;
Stone and Sobel, 1990; Preacher and Hayes 2004). Using Sobel significance test which assumes
normal distribution when testing for mediation is not appropriate and results in underpowered tests
of mediation (MacKinnon et al. 1995). Bootstrapping has been applied to analyses with issues of
asymmetries and nonnormality in the distribution of the data, which is the case with several variables
in the sample of this research. It can also be applied to smaller samples with more confidence
because it produces a test that does not rely on large-sample theory. The indirect effect of X on Y is
considered statistically different from zero when the bootstrap confidence interval does not span
zero (Hayes 2013).
It is important to keep in mind the distinction between terms: mediated effects and indirect
effects, which are sometimes used interchangeably. A claim that mediation exists implies the initial
presence of a total effect between X and Y to be mediated. No such assumption is necessary in the
assessment on the indirect effects. Variable X can influence Y through M, even if there is no direct
relationship between X and Y (Preacher and Hayes 2004). Therefore, following Preacher and Hayes
(2004) a more power strategy to test for mediation, compared to Baron and Kenny’s (1986) steps,
would require satisfying two criteria: (1) there is an effect to be mediated (effect c is different from
zero) and (2) the indirect effect is statistically significant. Yet, even if a mediation effect cannot be
established, it is still important to evaluate the presence of indirect effects, because it allows for a
better understanding of the complexity of the relationship between variables.
69
Tables 4.5 and 4.6 show the output of a mediation analysis following Hayes (2013) method
of bootstrapping for direct testing of the indirect effects (ab) using the PROCESS package in SPSS.
Table 4.5 describes results for the leadership behavior dependent variable (column Y), while Table
4.6 describes the output for the delay behavior dependent variable (column Y). Weighted sample
with robust standard errors was used in the analysis. All the effects displayed in the table are
unstandardized except the indirect (ab) output, which is standardized to allow for comparison of
effects across several independent variables and mediators. PROCESS produces standardized
coefficients only for the indirect effects, therefore other coefficients are recorded using their true
effect.
In Tables 4.5 and 4.6, effect c’ (X>>Y) represents the direct effect of socio-environmental
variables (column X) on the outcome while controlling for mediator variables (column M). Tables
also display the relationship between outside variables and mediators (effect a, X>>M) and between
mediators and outcome variables (effect b, M>>Y). In addition, ab is the indirect effect of the
outside controls on the behavioral intent via the mediator. Effect c is the total effect between X and
Y. To claim mediation, total initial effect must be significant.
Results for demographic variables are not displayed in the tables. However, demographic
characteristics considered in this study were included as controls in the mediation procedure in the
PROCESS tool. Reasons they are not displayed in the output table is because most of them did not
show any significant effect on the outcome and because they primarily serve as control variables and
not the main explanatory variables. This study takes an in-depth look at the variation of perceptions
and behavior among different demographic groups using the analysis of variance (ANOVA).
70
Table 4.5: Mediation Analysis Results Leadership Dependent Variable; SE are HC0, weighted sample
X M Y c’
(X>>Y) SE (c’) a (X>>M) SE (a)
b
(M>>Y) SE (b)
ab
standardized SE (ab)
(c)
Total
Effect
SE (c)
Wave arrival
Self-
efficacy
Leadership
behavior
.0305** .0136 .0656*** .0152 .1501*** .0509 .0547** 0.0044 -.0207 .0128
Knowledge .2085*** .0776 .6248*** .0845 .0489 .0554 .0317 .0380 .2390*** .0683
Preparation .0817*** .0254 .2082*** .0349 .0538 .0523 .0305 .0322 .0929*** .0228
Experience .0139 .0347 .1660*** .0410 .1115** .0500 .0369** .0195 .0325 .0335
Wave arrival
Response
efficacy
-.0235* .0130 .0167 .0171 .1679*** .0550 .0155 .0192 -.0207 .0128
Knowledge .2016*** .0711 .2859*** .0923 .1308** .0534 .0388** .0206 .2390*** .0683
Preparation .0788*** .0235 .1095*** .0341 .128 ** .0520 .0384** .0203 .0929*** .0228
Experience .0273 .0335 .0326 .0420 .1593*** 0.0524 .0103 .0152 .0325 .0335
Wave arrival
Perceived
severity
-.0139 .0134 -.0539*** .0102 .1267 .0832 -.0379 .0271 -.0207 .0128
Knowledge .2417*** .0683 -.0164 .0533 .1646** .0772 -.0028 .0108 .2390*** .0683
Preparation .0944*** .0220 -.0087 .0237 .1672** .0749 -.0040 .0125 .0929*** .0228
Experience .0301 .0330 .0150 .0332 .1566** .0797 .0047 .0127 .0325 .0335
Wave arrival
Perceived
probability
-.0225 * .0131 .0360** .0173 .0496 .0496 .0099 .0127 -.0207 .0128
Knowledge .2367*** .0685 .0767 .1008 .0303 .0486 .0024 .0072 .2390*** .0683
Preparation .0917*** .0227 .0556 .0442 .0224 .0474 .0034 .0098 .0929*** .0228
Experience .0328 .0335 -.0087 .0616 .0388 .0490 -.0007 .0083 .0325 .0335
71
Table 4.6: Mediation Analysis Results Delay Dependent Variable; SE are HC0, weighted sample
X M Y c’
(X>>Y) SE (c) a (X>>M) SE (a) b (M>>Y) SE (b)
ab
standardized SE (ab)
(c)
Total
effect
SE (c)
Wave arrival
Self-
efficacy
Delay
behavior
-.0035 .0127 .0656 *** .0152 .0860* .0491 .0335 .0229 .0021 .0122
Knowledge -.0678 .0754 .6248 *** .0845 .1042 ** .0526 .0723 .0405 -.0026 .0683
Preparation -.0202 .0225 .2082 *** .0349 .0978 ** .0488 .0592** .0345 .0001 .0221
Experience -.0489 * .0283 .1660 *** .0410 .1004 ** .0475 .0355** .0208 -.0322 .0286
Wave arrival
Response
efficacy
.0012 .0122 .0167 .0171 .0584 .0486 .0058 .0094 .0021 .0122
Knowledge -.0203 .0669 .2859*** .0923 .0618 .0484 .0196 .0191 -.0026 .0683
Preparation -.0066 0.0223 .1095*** .0341 .0614 .0500 .0196 .0189 .0001 .0221
Experience -.0342 .0283 .0326 .0420 .0613 .0479 .0043 .0076 -.0322 .0286
Wave arrival
Perceived
severity
-.0070 .0129 -.0539*** .0102 -.1691 * .0929 .0542 .0305 .0021 .0122
Knowledge -.0051 .0670 -.0164 .0533 -.1528* .0867 .0028 .0113 -.0026 .0683
Preparation -.0012 .0219 -.0087 .0237 -.1528* .0863 .0039 .0131 .0001 .0221
Experience -.0300 .0298 .0150 .0332 -.1501 .0861 -.0048 .0140 -.0322 .0286
Wave arrival
Perceived
probability
.0040 .0123 .0360** .0173 -.0524 .0429 -.0112 .0124 .0021 .0122
Knowledge .0012 .0674 .0767 .1008 -.0504 .0419 -.0043 .0083 -.0026 .0683
Preparation .0030 .0221 .0556 .0442 -.0508 .0423 -.0082 .0111 .0001 .0221
Experience -.0327 .0282 -.0087 .0616 -.0509 .0424 .0009 .0091 -.0322 .0286
72
The following discussion examines outcomes of the mediation analysis, addressing
previously proposed hypotheses.
Hypothesis 1a: Perceived probability mediates the influence of socio-environmental factors on the
behavioral intent.
For two independent variables (knowledge confidence and preparation) there is an observed
reduction in the association with the leadership behavioral intent when perceived probability is
added to the model, however, there is no significant influence of the independent variables on the
mediator nor mediator on the leadership intent. Following Baron and Kenny (1986) criteria to
establish mediation is not met. In Hayes (2013) direct test of indirect effect, perceived probability
does not show a significant influence of the leadership intent either. Perceived probability does not
mediate the influence of socio-environmental variables on the delay behavioral intent.
Hypothesis 1b: Perceived probability is positively associated with the behavioral intent to evacuate outside
of the tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output does not show a significant association between perceived
probability and leadership behavioral intent when controlling for four independent variables (path
b), even though the association is positive. Pearson’s correlation between perceived probability and
leadership behavior is also positive, but non-significant (r=0.018, p=0.793). This result is similar to
the OLS regression output, where there were found no association between leadership behavioral
intent and perceived probability. In the interpretation of this and the following hypotheses, where
there is no analysis of a mediation effect, output from the mediation model is prioritized over OLS
regression output. In the OLS regression analysis, the presence of all variables in one regression
model, many of which are highly correlated with each other, is dampening the true effect of
constructs. In the mediation model, even though there is a possibility for the omitted variable bias,
73
focus only on one variables prevents multicollinearity and type II error, retaining false null
hypothesis.
Hypothesis 1c: Perceived probability is negatively associated with the behavioral intent to engage in delay
behaviors in the event of a M9 CSZ tsunami.
Mediation analysis output does not show a significant association between perceived
probability and delay behavioral intent when controlling for four independent variables (path b),
even though the association is negative. Pearson’s correlation between perceived probability and
delay behavior is also negative, but non-significant (r=-0.099, p=0.160).Hypothesis 2a: Perceived
severity mediates the influence of socio-environmental factors on the behavioral intent.
For two independent variables (knowledge confidence and preparation) there is an observed
reduction in the association with the leadership behavioral intent when perceived severity is added to
the model. There is also a significant association between M and Y. However, path a is not
significant. Therefore, following the Baron and Kenny (1986) criteria to establish mediation is not
met. In Hayes (2013) direct test of indirect effect, perceived severity does not show a significant
influence of the leadership intent either. Perceived severity does not mediate the influence of socio-
environmental variables on the delay behavioral intent as there is no established significant
association between X and Y (c’ and c) and X and M (a).
Hypothesis 2b: Perceived severity is positively associated with the behavioral intent to evacuate outside of
the tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows a positive significant association between perceived severity
and leadership behavioral intent. Pearson’s correlation between perceived severity and leadership
behavior is also positive and significant (r=0.145, p<0.05).
74
Hypothesis 2c: Perceived severity is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
Mediation analysis output shows a significant negative association between perceived
severity and delay behavioral intent when controlling for four wave arrival, knowledge confidence
and preparation (path b). Pearson’s correlation between perceived severity and delay behavior is also
negative and significant on the 90% confidence level (r=-0.133, p<0.1).
Hypothesis 3a: Self-efficacy mediates the influence of socio-environmental factors on the behavioral intent.
Not all conditions to satisfy criteria for mediation according to Baron and Kenny’s (1986)
approach exist for the leadership dependent variable. For knowledge confidence and preparation
variables, there is an observed reduction in the association with the leadership behavioral intent
when self-efficacy is controlled for in the model. There is also a significant association between X
and M (path a) However, path b is not significant. For the wave arrival variable, path c, a, and b are
significant, but there is no established initial effect of X on Y. However, following Hayes (2013)
direct test of indirect effect, wave arrival and experience show positive significant influence on
leadership behavior through self-efficacy, wave arrival showing greater influence compared to
experience.
Similarly, not all conditions to satisfy criteria for mediation according to Baron and Kenny’s
(1986) approach exist for the delay dependent variable. None of the independent variables shows a
significant total association with the delay dependent variable. However, following Hayes (2013)
direct test of indirect effect, preparation and experience show positive significant influence on delay
behavior through self-efficacy, preparation showing greater influence compared to experience.
Hypothesis 3b: Self-efficacy is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
75
Mediation analysis output shows positive significant association between self-efficacy and
leadership behavioral intent when controlling for wave arrival and experience. Pearson’s correlation
between self-efficacy and leadership behavior is also positive and significant (r=0.182, p<0.01).
Hypothesis 3c: Self-efficacy is negatively associated with the behavioral intent to engage in delay behaviors
in the event of M9 CSZ tsunami.
Mediation analysis output shows positive significant association between self-efficacy and
delay behavioral intent when controlling for all four independent variables. This result contradicts
the hypothesis. Pearson’s correlation between self-efficacy and leadership behavior is also positive
but non-significant (r=0.072, p=0.310).
Hypothesis 4a: Response efficacy mediates the influence of socio-environmental factors on the behavioral
intent.
Criteria for mediation according to Baron and Kenny’s (1986) approach exist for leadership
dependent variable when looking at the relationship between knowledge confidence and the
outcome, and preparation and the outcome. There is a positive significant total influence of X on Y.
There is an established reduction in the association between X and Y when M is controlled for in
the model. Paths a and b are significant. Response efficacy mediates the influence of knowledge
confidence and preparation on leadership behavioral intent. In addition, following Hayes (2013)
direct test of indirect effect, shows a significant indirect influence of knowledge confidence and
preparation on the outcome via response efficacy. The indirect effect of the two variables is similar
to each other. For delay behavioral intent, not all conditions exist to satisfy criteria for mediation
according to Baron and Kenny’s (1986). None of the independent variables shows a significant total
association with the delay dependent variable. In addition, following Hayes (2013) direct test of
indirect effects, there is no established significant indirect effects.
76
Hypothesis 4b: Response efficacy is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows positive significant association between response efficacy
and leadership behavioral intent when controlling for all independent variables. Pearson’s correlation
between response efficacy and leadership behavior is also positive and significant (r=0.231, p<0.01).
Hypothesis 4c: Response efficacy is negatively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output does not show a negative significant association between response
efficacy and delay behavioral intent. Pearson’s correlation between response efficacy and delay
behavior is positive and non-significant (r=0.037, p=0.599), showing no support for the hypothesis.
Hypothesis 5a: Tsunami-relevant knowledge confidence is positively associated with the behavioral intent
to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows that knowledge confidence is positively and significantly
associated with leadership behavioral intent with and without control of mediator variables.
Pearson’s correlation between knowledge confidence and leadership behavior is also positive and
significant (r=0.209, p<0.01), showing support for the hypothesis.
Hypothesis 5b: Tsunami-relevant knowledge confidence is negatively associated with the behavioral intent
to engage in delay behaviors in the event of M9 CSZ tsunami.
Mediation analysis output shows no support for this hypothesis. Pearson’s correlation
between knowledge confidence and delay behavior is also positive and non-significant (r=0.003,
p=0.971), showing no support for the hypothesis.
77
Hypothesis 5c: Tsunami-relevant knowledge confidence is positively associated with self-efficacy.
Unstandardized coefficient of association between knowledge confidence and self-efficacy
from the mediation analysis output is positive and significant (b=0.6248, p<0.01). Pearson’s
correlation (or standardize coefficient) is also positive and significant (r=0.560, p<0.01), showing
support for the hypothesis.
Hypothesis 5d: Tsunami-relevant knowledge confidence is positively associated with response efficacy.
Unstandardized coefficient of association between knowledge confidence and response
efficacy from the mediation analysis output is positive and significant (b=0.2859, p<0.01). Pearson’s
correlation (or standardize coefficient) is also positive and significant (r=0.223, p<0.01), showing
support for the hypothesis.
Hypothesis 5e: Tsunami-relevant knowledge confidence is positively associated with perceived probability.
There is no support for this hypothesis in the mediation analysis output. Pearson’s
correlation (or standardize coefficient) is also non-significant (r=0.043, p=0.538), showing no
support for the hypothesis.
Hypothesis 5f: Tsunami-relevant knowledge confidence is negatively associated with perceived severity.
There is no support for this hypothesis in the mediation analysis output. Pearson’s
correlation (or standardize coefficient) is also non-significant (r=-0.107, p=0.129), showing no
support for the hypothesis.
Hypothesis 6a: Hazard experience is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows that hazard experience is not significantly associated with
leadership behavioral intent, non-significance holds in models with and without control of mediator
78
variables. Pearson’s correlation between hazard experience and leadership behavior is also non-
significant (r=0.052, p=0.458), showing no support for the hypothesis.
Hypothesis 6b: Hazard experience is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
Mediation analysis output shows support for the hypothesis only in the model with the
presence self-efficacy. In that situation, hazard experience shows negative significant association
with delay behavioral intent (b=-0.0489, p<0.1). Pearson’s correlation between hazard experience
and delay behavior is also negative and significant (r=-0.138, p<0.05), showing support for the
hypothesis.
Hypothesis 6c: Hazard experience is positively associated with self-efficacy.
Unstandardized coefficient of association between hazard experience and self-efficacy from
the mediation analysis output is positive and significant (b=0.1660, p<0.01). Pearson’s correlation
(or standardize coefficient) is also positive and significant (r=0.217, p<0.01), showing support for
the hypothesis.
Hypothesis 6d: Hazard experience is positively associated with response efficacy. There is no support for
this hypothesis in the mediation analysis.
Pearson’s correlation (or standardize coefficient) is also non-significant (r=0.087, p=0.218),
showing no support for the hypothesis.
Hypothesis 6e: Hazard experience is positively associated with perceived probability. There is no support
for this hypothesis in the mediation analysis.
Pearson’s correlation (or standardize coefficient) is also non-significant (r=-0.044, p=0.531),
showing support for the hypothesis.
79
Hypothesis 6f: Hazard experience is positively associated with perceived severity.
There is no support for this hypothesis in the mediation analysis. Pearson’s correlation (or
standardize coefficient) is also non-significant (r=0.079, p=0.260), showing support for the
hypothesis.
Hypothesis 7a: Risk proximity is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows no total effect of wave arrival time on the leadership intent.
Wave arrival time has a positive significant association with leadership intent when controlled for
self-efficacy, which does not support the hypothesis because that means that greater tsunami arrival
time is associated with a higher likelihood of having an intention to engage in the leadership
behavior. Wave arrival time is negatively significantly associated with leadership intent when
controlled for response efficacy and perceived probability, which supports the hypothesis. Pearson’s
correlation between wave arrival time and leadership behavior is negative, but non-significant (r=-
0.095, p=0.177), showing no support for the hypothesis.
Hypothesis 7b: Risk proximity is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
Mediation analysis output shows no support for this hypothesis. Pearson’s correlation
between wave arrival time and leadership behavior is negative, but non-significant (r=0.046,
p=0.511), showing no support for the hypothesis.
Hypothesis 7c: Risk proximity is negatively associated with self-efficacy.
Unstandardized coefficient of association between wave arrival time and self-efficacy from
the mediation analysis output is positive and significant (b=0.0656, p<0.01). Pearson’s correlation
(or standardize coefficient) is also positive and significant (r=0.251, p<0.01), showing support for
80
the hypothesis. This result suggests that higher wave arrival time (i.e. lower risk proximity) is
associated with higher level of self-efficacy.
Hypothesis 7d: Risk proximity is positively associated with response efficacy.
There is no support for the association between wave arrival time and response efficacy in
the mediation analysis. Pearson’s correlation (or standardize coefficient) is also non-significant
(r=0.076, p=0.292), showing no support for the hypothesis.
Hypothesis 7e: Risk proximity is positively associated with perceived probability.
Unstandardized coefficient of association between wave arrival time and perceived
probability from the mediation analysis output is positive and significant (b=0.0360, p<0.05).
Pearson’s correlation (or standardize coefficient) is also positive and significant (r=0.165, p<0.01),
showing the opposite direction of the relationship from the one assumed in the hypothesis.
Hypothesis 7f: Risk proximity is positively associated with perceived severity.
Unstandardized coefficient of association between wave arrival time and perceived severity
from the mediation analysis output is negative and significant (b=-0.0539, p<0.01). Pearson’s
correlation (or standardize coefficient) is also negative and significant (r=-0.374, p<0.01), showing
support for the hypothesis, that further location of a household from the coastline is associated with
lower level of perceived severity (or closer risk proximity is associated with higher level of perceived
severity).
Hypothesis 8a: Preparation is positively associated with the behavioral intent to evacuate outside of
tsunami inundation zone in the event of M9 CSZ tsunami.
Mediation analysis output shows presence of total significant effect of preparation on the
leadership intent. Preparation is also positively significantly associated with leadership intent when
controlled for all cognitive variables, strongly supporting the hypothesis. Pearson’s correlation
81
between preparation and leadership behavior is positive and significant (r=0.251, p<0.01), showing
support for the hypothesis.
Hypothesis 8b: Preparation is negatively associated with the behavioral intent to engage in delay
behaviors in the event of M9 CSZ tsunami.
Mediation analysis output shows no support for this hypothesis. Pearson’s correlation
between wave arrival time and leadership behavior is negative, but non-significant (r=-0.020,
p=0.773), showing no support for the hypothesis.
Hypothesis 8c: Preparation is positively associated with self-efficacy.
Unstandardized coefficient of association between preparation and self-efficacy from the
mediation analysis output is positive and significant (b=0.2082, p<0.01). Pearson’s correlation (or
standardized coefficient) is also positive and significant (r=0.446, p<0.01), showing support for the
hypothesis.
Hypothesis 8d: Preparation is positively associated with response efficacy.
Unstandardized coefficient of association between preparation and self-efficacy from the
mediation analysis output is positive and significant (b=0.1095, p<0.01). Pearson’s correlation (or
standardized coefficient) is also positive and significant (r=0.205, p<0.01), showing support for the
hypothesis.
Hypothesis 8e: Preparation is positively associated with perceived probability.
There is no support for the association between preparation and perceived probability in the
mediation analysis. Pearson’s correlation (or standardized coefficient) is also non-significant
(r=0.097, p=0.167), showing no support for the hypothesis.
82
Hypothesis 8e: Preparation is positively associated with perceived severity.
There is no support for the association between preparation and perceived probability in the
mediation analysis. Pearson’s correlation (or standardized coefficient) is also non-significant (r=-
0.034, p=0.628), showing no support for the hypothesis. Table 4.7 displays summary of all
hypotheses and outcomes of the mediation analysis.
Table 4.7: Summary of research hypotheses testing.
# Hypothesis Outcome
Hypothesis 1
a) Perceived probability mediates the influence of socio-environmental factors on behavioral intent b) Perceived probability is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami c) Perceived probability is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami
a) Not supported b) Not supported c) Not supported
Hypothesis 2
a) Perceived severity mediates the influence of socio-environmental factors on behavioral intent b) Perceived severity is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami c) Perceived severity is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami
a) Not supported b) Supported c) Supported
Hypothesis 3
a) Self-efficacy mediates the influence of socio-environmental factors on behavioral intent b) Self-efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami c) Self-efficacy is negatively associated with the behavioral intent to engage in delay and milling behaviors in the event of M9 CSZ tsunami
a) Leadership: no mediation, but positive indirect effect of wave arrival and experience; Delay: no mediation, but positive indirect effect of preparation and experience b) Supported c) Not supported (relationship is positive)
Hypothesis 4
a) Response efficacy mediates the influence of socio-environmental factors on behavioral intent b) Response efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami c) Response efficacy is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami
a) Leadership: mediation exist for knowledge and preparation; Delay: no mediation b) Supported c) Not supported
Hypothesis 5
a) Tsunami-relevant knowledge efficacy is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami b) Tsunami-relevant knowledge efficacy is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami
a) Supported b) Not supported c) Supported d) Supported e) Not supported f) Not supported
83
c)Tsunami-relevant knowledge confidence is positively associated with self-efficacy d) Tsunami-relevant knowledge confidence is positively associated with response efficacy e) Tsunami-relevant knowledge confidence is positively associated with perceived probability f) Tsunami-relevant knowledge confidence is negatively associated with perceived severity
Hypothesis 6
a) Hazard experience is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami b) Hazard experience is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami c) Hazard experience is positively associated with self-efficacy d) Hazard experience is positively associated with response efficacy e) Hazard experience is positively associated with perceived probability f) Hazard experience is positively associated with perceived severity
a) Not supported b) Supported, when controlled for self-efficacy; no total effect c) Supported d) Not supported e) Not supported f) Not supported
Hypothesis 7
a) Risk proximity is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami b) Risk proximity is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami c) Risk proximity is negatively associated with self-efficacy d) Risk proximity is positively associated with response efficacy e) Risk proximity is positively associated with perceived probability f) Risk proximity is positively associated with perceived severity
a) Supported, when controlled for response efficacy and perceived probability b) Nor supported c) Supported d) Not supported e) Not supported, relationship is negative f) Supported
Hypothesis 8
a) Preparation is positively associated with the behavioral intent to evacuate outside of tsunami inundation zone in the event of M9 CSZ tsunami b) Preparation is negatively associated with the behavioral intent to engage in delay behaviors in the event of M9 CSZ tsunami c) Preparation is positively associated with self-efficacy d) Preparation is positively associated with response efficacy e) Preparation is positively associated with perceived probability f) Preparation is positively associated with perceived severity
a) Supported b) Not supported c) Supported d) Supported e) Not supported f) Not supported
4.4 Analysis of variance
To answer research question 2: what is the variation in behavioral intent among the
following demographic characteristics: gender, age, education, income, community tenure, house
ownership, household size, difficulty walking, presence of children and presence of elderly in a
household, this study employs analysis of variance. The standard test for the analysis of variance is
ANOVA test, which compares variation of means between groups of interests. ANOVA relies on
84
assumptions of normal distribution of the population. Table 4.8 displays statistics of mean, standard
deviation, skewness, kurtosis, and Shapiro-Wilk test of normality for leadership and delay dependent
variables.
Table 4.8: Descriptive statistics and Shapiro-Wilk test of normality for dependent variables
Statistics Leadership Delay
Mean 2.59 2.56
Standard deviation 0.74 0.69
Skewness -0.847 -0.925
Kurtosis -0.043 0.309
Shapiro-Wilk test 0.900*** 0.905***
N 206 206
Normality distributed data generally has skewness that is between -0.5 and +0.5 and kurtosis
of 3 (Bulmer 1979). Table 4.8 shows that leadership variable has a skewness of -0.847, which
represents a moderately negatively skewed distribution, i.e. the left tail of the distribution is longer
than the right tail. Delay variable has a skewness of -0.925, which also represents a moderately
negatively skewed distribution, i.e. the left tail of the distribution is longer than the right tail.
Leadership and delay behavior have kurtosis of 0.309 and -0.043 respectively, both deviating from
kurtosis of a normal distribution. Shapiro-Wilk’s test, which is significantly positive for both
leadership and delay behavior, indicates that there is a deviation from a normal distribution in
observations, but it does not violate the assumption of normality that ANOVA t-test relies on.
However, for purposes of robustness, this study provides output for both ANOVA t-test and
Kruskall-Wallis test.
Kruskall-Wallis test is a non-parametric test that does not assume normal distribution of
residuals. In the Kruskall-Wallis test the null hypothesis is that the medians of groups are statistically
85
the same. Therefore, rejection of the null means that there is a statistical difference between medians
of at least two group, if there are more than two groups in the analysis. In general ANOVA is
considered a robust test even if there is violation of its assumptions. This study relies on ANOVA t-
test and F-test to address the research question, however, Kruskall-Wallis Chi-squares (i.e. test of a
null hypothesis) are provided in Table 4.9 for comparison. For two variables “difficulty walking” and
“children present” there is a difference between t-test and Kruskall-Wallis tests, which indicates that
there may be some deviation from normality in the distribution of the observations.
86
Table 4.9: Results of ANOVA and Kruskall-Wallis tests
Variable
Leadership (1-5 scale) mean
Delay (1-5 scale) mean
Gender (n=206)
Female Male
t-test (F-test) Chi-square (Kruskall-Wallis)
2.71 2.42
7.895** 7.113**
2.65 2.43
5.293* 7.189**
Age (n=206)
(1) 24-40 (2) 41-64 (3) 65-95
t-test (F-test) Chi-square (Kruskall-Wallis)
Tukey Post Hoc
2.84 2.69 2.44
4.181* 10.899**
(1-3) = p<0.05
2.76 2.59 2.48 1.639 5.932 N/A
Education (n=206)
Less than Bachelor Bachelor
Graduate degree t-test (F-test)
Chi-square (Kruskall-Wallis)
2.58 2.7 2.53 0.683 2.166
2.59 2.51 2.53 0.301 0.313
Income (n=206)
Below 40,000 40,000-75,000 Above 75,000 t-test (F-test)
Chi-square (Kruskall-Wallis)
2.59 2.57 2.64 0.136 0.483
2.58 2.54 2.55 0.046 0.302
Community tenure (n=206)
0-10 years 11-30 years 31-76 years
t-test (F-test) Chi-square (Kruskall-Wallis)
2.61 2.65 2.49 0.701 1.008
2.56 2.55 2.57 0.010 0.107
Homeownership (n=206)
Primary Renter Other
t-test (F-test) Chi-square (Kruskall-Wallis)
2.60 2.55 2.84 0.446 0.653
2.55 2.60 2.41 0.297 0.966
Household size (n=206)
1 person 2 people
3 and more t-test (F-test)
Chi-square (Kruskall-Wallis) Tukey Post Hoc
2.50 2.56 2.81 2.420 5.529 N/A
2.48 2.47 2.88
5.936** 13.242***
(1-3) = p<0.01;(2-3) = p<0.01
Difficulty walking (n=206)
Yes No
t-test (F-test) Chi-square (Kruskall-Wallis)
2.39 2.67
6.100* 6.903**
2.43 2.61 2.839 5.490*
Children present (n=206)
Yes No
t-test (F-test) Chi-square (Kruskall-Wallis)
2.93 2.56
3.990* 3.986*
2.85 2.53 3.517 4.813*
Elderly present (n=206)
Yes No
t-test (F-test) Chi-square (Kruskall-Wallis)
2.56 2.61 0.146 0.301
2.53 2.57 0.135 0.011
*p<0.05; **p<0.01; ***p<0.001
87
Table 4.9 displays ANOVA analysis outcomes. There were distinct differences between
genders with respect to the mean levels of leadership behavioral intent, t=7.895, p<0.01 and of delay
behavioral intent, t=5.293, p<0.05. Females were statistically more likely to have an intent to engage
in both leadership and delay behaviors compared to men. Kruskall-Wallis test confirms the statistical
difference among the groups, Chi-square=7.113, p<0.01 for the leadership behavioral intent and
Chi-square=7.189, p<0.01 for the delay behavioral intent. There was a significant difference among
age groups in the mean levels of leadership behavioral intent, F=4.181, p<0.05, where 24-40 age
group was statistically more likely to have an intent to engage in the leadership behavior compared
to 65-95 age group. Kruskall-Wallis test confirms the statistical difference among the groups, Chi-
square=10.899, p<0.01 for the leadership behavioral intent. There was a significant difference
among household of different sizes in the mean levels of delay behavioral intent, F=5.936, p<0.01,
where a household of three and more people was statistically more likely to have an intent to engage
in the delay behavior compared to a household of 2 people and compared to a household of 1
person. Kruskall-Wallis test confirms the statistical difference among the groups, Chi-
square=13.242, p<0.001 for the delay behavioral intent. There was a significant difference between
respondents with and without walking difficulty with respect to the mean level of the leadership
behavioral intent, t=6.100, p<0.05. Kruskall-Wallis test confirms the statistical difference among the
groups, Chi-square=6.903, p<0.01 for the leadership behavioral intent. Finally, there was a
significant difference among household with and without children in the mean levels of leadership
behavioral intent, t=3.990, p<0.05, where households with children were statistically more likely to
have an intent to engage in the leadership behavior compared to households without children.
Kruskall-Wallis test confirms the statistical difference among the groups, Chi-square=3.986, p<0.05
for the leadership behavioral intent.
88
5 Discussion
The goal of this research was to understand how populations at risk of CSZ earthquakes and
tsunamis along the Oregon coast would respond in the event of the disaster and what factors may
explain variation in their behavioral intentions. To do two research questions were asked and several
hypotheses proposed, which will be evaluated in detail in this chapter, connecting the discussion
with the existing natural hazard literature and the goals of this study.
The first research question asked about variation in people’s behavioral intentions in the
event of M9 CSZ earthquake and tsunami. The second research question asked about differences in
people’s behavioral intentions with respect to gender, age, formal education, income, community
tenure, home ownership, household size, walking difficulty, presence of children, and presence of
elderly in the household.
In rapid-onset natural disasters such as earthquakes and tsunamis, evacuation from a risk
area is considered the best protective action within the first minutes of a disaster onset (Mas et al.
2012, 2013; Wei et al. 2017; Charnkol and Tanaboriboon 2006; Takabatake et al. 2017; Wei and
Lindell 2017). However, previous research demonstrates that in emergency situations people tend to
engage in a variety of behaviors that delay the start of their immediate evacuation, for example
searching for tsunami confirmation information (Lindell et al. 2015; Wei et al. 2017), looking for
family members, talking with neighbors, packing an emergency kit and other things around the
house, helping others, and waiting for assistance from emergency personnel (Shibayama et al. 2013;
Lindell and Perry 2012; Yun and Hamada 2012; Lindell et al. 2015; Murakami et al. 2012; . Jon et al.
2016). Delay and milling behaviors are unavoidable as local tsunamis can happen unexpectedly at
any hour during the day, likely catching people unprepared in different locations and separated from
family and friends.
89
Even though empirical literature that reveals the connection between behavioral intentions
and behavior in natural disasters is limited, the link is possible (Kang et al. 2007). Moreover,
understanding people’s intentions during hypothetical scenarios and having information about how
people think about their future actions provide a valuable perspective on their current level of
preparedness. This information can inform efforts to change people’s beliefs and attitudes toward
tsunami protective actions and preparedness, and other projects such as construction of evacuation
routes and shelters, improving emergency warning systems, targeting vulnerable populations, and
other programs.
Analysis of behavioral intentions’ distribution revealed the following results. The top three
behaviors that the respondents said they would very unlikely adopt in the event of M9 CSZ
earthquakes and tsunamis included continuing normal routine behavior (63.1%), waiting for family
at home (47.1%), and following friends and neighbors (43.7%). The top three behaviors that
respondents indicated they would very likely assume included evacuating to higher ground
immediately following the earthquake (51%), contacting loved ones (49.5%), and checking social
media, TV (40.3%).
On one hand, it is encouraging to see that 63.1% of respondents indicated that they were
very unlikely to continue normal routine behavior in the event of CSZ earthquake and tsunami
(21.4% responding with “unlikely”). This result conveys that respondents are aware of the risks of
the potential disaster and intend to respond to it by changing their routine behaviors. From a
methodological perspective, this outcome also suggests that respondents pay attention to survey
questions, because this is a logically consistent response. It is likely that their answers to the rest of
behavioral intentions scale reflect their true intentions. On the other hand, 15.6% of respondents (32
participants) who were either uncertain or said they were likely or very likely to continue normal
90
behavior also scored on the lower end of the spectrum on self-efficacy, which could be a reflection
of a fatalist mentality due to reduced ability to protect oneself in a risk situation. In addition, 16 of
32 participants belong to the eldest age category of 65-95, suggesting that older populations are
more likely to have an intention to continue normal behavior in an emergency situation. Lindell et al.
(2016) reported that during 2011 earthquakes in Hitachi, Japan and Christchurch, New Zealand
older residents were more likely to continue normal activities or stay behind to protect their
property.
From a tsunami response perspective, it is encouraging to see that 22.8% of respondents
answered that they were very unlikely and 24.3% said that they were unlikely to wait for tsunami
confirmation. In local tsunami events, the earthquake itself is the only necessary indicator of an
upcoming tsunami that the population at risk must follow (Fraser et al. 2013, Bernard 2005). Waiting
for and relying on additional tsunami warnings from any sources is discouraged because channels of
communication are likely disabled during an earthquake. Waiting reduces time available for
evacuation. Therefore, it is also troubling to see that 29.6% of respondents indicated that they would
likely wait for tsunami confirmation, while 14.1% said they would be very likely to do so. These
results could be an outcome of an unawareness on the part of survey participants that behavioral
questions referred to a local tsunami event. Also, respondents may not know the difference between
local and distant tsunami events and the warning signs for each type of event, even though 35.4% of
respondents reported that they were confident and 30.6% said they were very confident in knowing
the difference between local and distant tsunamis. It is possible that confidence in knowing
information does not match with actual knowledge. Residents of the Oregon coast are familiar with
distant tsunami warnings because of the relative frequency of distant tsunami threats (e.g. latest
tsunami watch was announced in January 2018 as a result of 7.9 M earthquake in the Gulf of
91
Alaska). There could be an expectation among coastal populations that a warning for local tsunamis
would come via the same channels. The fact that 35.0% of survey respondents indicated that they
would likely and 40.3% would very likely to check social media and TV in the event of the Cascadia
earthquake and tsunami suggests that respondents are not sure about the extent of damages during
the CSZ scenario, during which it is highly unlikely that phones and electrical services would be
functioning. Finally, 63.1% of respondents indicated that they were either very unlikely or unlikely to
panic in the event of the disaster. This result supports empirical findings in hazard literature that
argue that a decision-making process in emergency situations more often involves a reflective
process of interpreting available information and evaluating alternatives for actions. Inefficient and
tragic outcomes are generally a result of inadequate information and resources than faults in
cognition (Lindell and Perry 2012).
This study did not explicitly focus on the analysis of people’s behavioral intentions or
preparation for impacts of the earthquake. However, results indicated that 80.6% of respondents
would either likely or very likely follow earthquake behavioral guidelines of sheltering under the
table. Although covering under a table inside is a recommended action in earthquake emergencies,
effectiveness of protective actions depends on a seismic performance of particular types of
buildings. While post-earthquake assessments show that many people die while attempting to escape
outside, running outside could be the only chance of surviving if one is located in a non-seismically
resistant building. Previous research on behavioral responses to earthquakes revealed a range of
behaviors that people tend to engage in, suggesting that availability and access to information on the
safest earthquake behavior by specific locations is important (Alexander and Magni 2013; Goltz et al.
1992; Lindell et al. 2016).
92
There are contradicting results within behavioral intentions outcomes, which could be
indicative of a poor question construction and the uncertainty in intentions on the part of survey
participants. The most popular behaviors respondents indicated they would adopt included
evacuating to higher ground immediately following the earthquake (very likely - 51%, likely –
22.8%), contacting loved ones (very likely - 49.5%, likely – 31.6%), and checking social media, TV
(very likely - 40.3%; likely – 40.3%).
Factor analysis of the behavioral intent scale revealed two distinct behavioral indices: the
leadership behavioral intent, which included two behavioral intentions of evacuating to higher
ground immediately and taking the lead to evacuate and encourage others; and the delay behavioral
intent, which included intentions of checking social media, TV, contacting loved ones, and collecting
documents and other items. Conceptually, the two indices represent different types of behavior, but
not mutually exclusive behaviors of evacuation vs. non-evacuation. The leadership index captures
behavioral intent of immediate evacuation following earthquake shaking, while the delay index
represents intentions of waiting for tsunami confirmation, desires to reunite with family and friends
and to help others before evacuating.
ANOVA results demonstrated that on average women indicate a higher likelihood of having
an intention to adopt leadership behavior compared to men, while also indicating a higher likelihood
of having an intention to engage in delay behaviors. The reason that females score higher than males
on both behavioral intentions could be because females are more aware of risks and the need to act
proactively, but they also feel the importance of protecting their families. In this study female gender
is also associated with higher level of perceived severity (r=0.171, p < 0.05) and lower level of self-
efficacy (r = -0.142, p < 0.05). Being female is the only demographic characteristic that is
consistently shown to be associated with higher levels of risk perception and the adoption of
93
protective actions in the natural hazard literature (Lindell et al. 2015; Terpstra and Lindell 2012;
Barberi et al. 2008; Fothergill 1996; Griffin et al. 1999; Lindell and Hwang 2008; Peacock et al. 2005).
According to Rivers’ (1982) argument on sex discrimination in disasters, it is worth paying
attention to how gender differences in social role expectations result in different levels of exposure
to risk and mortality rates during disaster events between females and males. For instance, the
reason that females scored higher on both leadership and delay behavioral intentions scales could be
indicative of the fact that females are more likely to be aware of official emergency guidelines,
because they feel responsible for safety of their children and other family members (that being a
stronger social role for females than males). At the same time, females may also feel a stronger need
to protect and help others, mobilizing instincts of social attachment, leading to delay in evacuation
(Shapira et al. 2018).
An ANOVA test also showed that there were distinct differences between survey
respondents who live in households with and without children under the age of 10 with respect to
the mean level of likelihood of having an intention to adopt leadership behavior. At the same time
individuals living in households with three or more people scored higher on the mean level of
likelihood of adopting delay behaviors. There is a strong correlation between household size and
presence of children in a household (r = 0.634, p < 0.01). It is possible that individuals in
households with children are more aware of recommended protective responses (similar to the
argument with respect to female gender), therefore they are more inclined to evacuate immediately
to ensure safety of children. At the same time, respondents in households with higher numbers of
people might also want to account for their cohabitants before evacuating, having an intent to
engage in delay behaviors. In addition, Kruskall-Wallis test shows statistically significant difference
between survey respondents who live in households with and without children under the age of 10
94
with respect to the mean level of likelihood of having an intention to adopt leadership behavior,
suggesting that individuals living in households with children scored higher on the mean level of
likelihood of delay behaviors as well, suggesting that perhaps parents have intentions to find and
account for their children before evacuating to high ground.
Mean differences in the levels of behavioral likelihood with respect to age were consistent
with previous empirical findings in the disaster literature. The highest mortality rates in natural
disasters are generally observed among older populations (Murakami et al. 2012; Rofi et al. 2006; Sun
et al. 2013; Yun and Hamada 2015). In this study, there was a significant difference in mean levels of
the leadership behavioral intent likelihood among age groups, where the 24-40 age group was more
likely to have an intent to engage in the leadership behavior compared to the 65-95 age group. Older
respondents also demonstrated a lower level of confidence in tsunami-relevant knowledge (r = -
0.165, p < 0.05), a lower self-efficacy (r = -0.268, p < 0.001), and reported fewer items from the list
of preparatory measures compared to younger participants (r = -0.186, p < 0.01). Low levels of self-
efficacy and preparation can induce avoidance behaviors (e.g. intention to not evacuate), which puts
older residents at higher risks of injuries during a disaster (Abraham et al., 1994; Rippetoe and
Rogers, 1987; Van der Velde and Van der Pligt, 1991). Age also displays positive and significant
correlation with walking difficulty (r = 0.376, p < 0.01). Respondents reporting a walking difficulty
demonstrate a lower mean level of likelihood to have the leadership behavioral intent, compared to
respondents who report having no walking difficulties. It is reasonable to conclude that older
respondents are more likely to have walking difficulties compared to younger respondents and they
are also less likely to have an intention to evacuate immediately after the earthquake. This finding
identifies a specific vulnerable population group with elevated risks to the effects of a CSZ
earthquake and tsunami.
95
This research also examined a series of hypotheses designed to understand factors that
influence behavioral intentions and channels through which these elements operate. More
specifically, it is argued that perceptions and attitudes translate socio-environmental observable
factors into behavioral intentions, serving as mediators between outside influences and behavioral
intentions. Four cognitive constructs (perceived probability, perceived severity of disaster outcomes,
self-efficacy to perform protective actions, and efficacy of response actions) were evaluated. Four
socio-environmental factors, that were hypothesized to influence behavioral intentions, included
tsunami wave arrival time or risk proximity, confidence in knowing tsunami-relevant information,
preparedness measures already taken, and previous experience with natural hazards. Mediation
analysis utilized bootstrapping confidence intervals for significance analysis of indirect effects
advocated by Hayes (2013) as the primary method of hypotheses’ assessment.
Risk perception (perceived severity and perceived probability) has been shown to be one of
the most important cognitive factors that influences people’s behavior in different types of disaster
situations, including earthquakes and tsunamis (Fraser et al. 2016; Lindell and Whitney 2000; Lindell
et al. 2009; Spittal et al. 2008; Tekeli-Yesil et al. 2010; Wei et al. 2017; Apatu et al. 2013; Lindell and
Perry 2000, 2012). Previous findings attest that higher level of risk perception motivates people to
adopt protective actions and be more proactive in terms of preparation for and in reaction to risks
of natural hazards. On the other hand, low risk perception may lead to delay and avoidance
behaviors. Outcomes of the mediation analysis partially support these arguments.
Perceived severity (i.e. perceptions about dangers to health, property and financial resources)
displayed a positive significant association with the leadership behavioral intention and a negative
significant association with the delay behavioral intention. This outcome suggests that a higher level
of perception that CSZ earthquakes and tsunamis would result in negative consequences to health
96
and property, is associated with higher likelihood of having intentions of immediate evacuation, and
lower likelihood of having intentions to engage in delay and milling behaviors. However, the analysis
did not show support for theoretical claims of a mediating role of risk perception. Neither perceived
probability nor perceived severity met either Baron and Kenny’s (1986) criteria for mediation or
Hayes’s (2013) direct test of indirect effects. Certainly, other studies have previously found that
direction and degree of relationship between risk perception and behavioral intent and behavior can
be inconsistent (e.g. Lindell and Prater 2000; Lindell and Whitney 2000; Paton et al. 2000; Perry and
Lindell 2008). Different research designs and measurements of risk perception can contribute to the
variation in outcomes, suggesting that risk perception is an important but not a sufficient factor that
explains households’ adoption of protective behaviors (Lindell and Whitney 2000).
Even though risk perception did not display a mediating role, wave arrival time or risk
proximity revealed to have an influence on both perceived probability and perceived severity. Wave
arrival time showed a negative significant association with perceived severity, meaning that
individuals living in households further away from the coastline (i.e. threats of physical risks are
lower) reported lower perceived severity, aligning with arguments of previous research findings
(Apatu et al. 2013, 2016; Lindell et al. 2015; Charnkol and Tanaboriboon, 2006; Wei et al. 2017; Arias
et al. 2017). Pearson’s correlation between wave arrival time and perceived severity was also negative
and significant (r = -0.374, p < 0.001). Interestingly, association between wave arrival time and
perceived likelihood (i.e. perception of the probability of CSZ earthquake and tsunami occurrence in
the next 50 years) was positive and significant in the mediation analysis, meaning that people living
in households further away from the coastline reported having higher perceived likelihood of CSZ
earthquake and tsunami happening in the next 50 years. This association is likely to be driven by a
different factor than the threat of risk. Perception of likelihood of a disaster event occurrence may
97
depend more on people’s access to information and trust in information sources about the disaster
event. In general, it is interesting to observe that out of four socio-environmental variables only
wave arrival time showed a significant association with perceived severity and perceived probability.
Certainly, there are other factors that could influence the level of risk perception, but this analysis
revealed that the level of objective risk had the greatest association with the subjective level of risk,
perhaps suggesting that respondents approach the assessment of threats based on objective
evaluations of available information.
Cognitive theories of behavior argue that as a psychological construct self-efficacy can
mediate the effects of outside observable variables on behavioral intent and behavior. In addition,
studies of natural disasters showed that self-efficacy has an influence on the adoption of protective
behaviors (Johnston et al. 2005). Mediation analysis did not show support for the theoretical claim of
the mediating role of self-efficacy between socio-environmental factors and either leadership nor
delay behavioral intentions. However, analysis of indirect effects according to Hayes’ (2013) method
revealed a significant indirect influence of wave arrival and experience on the leadership behavioral
intent. Even though there was no initial impact of wave arrival and experience on behavioral intent
to mediate, self-efficacy still served as a path through which these two factors were associated with
the leadership behavior. Greater wave arrival time was associated with higher level of self-efficacy,
meaning individuals living in households that are located further away from the coastline had a
higher level of confidence in their abilities to protect themselves from risks. In turn, higher self-
efficacy was associated with higher likelihood of leadership behavioral intent, supporting previous
research findings and assumptions of behavioral theories. Also, a greater number of disaster events
experienced was associated with a higher level of self-efficacy, which, in turn, positively influenced
the likelihood of leadership behavioral intent.
98
Similarly, preparation and experience had an indirect significant influence on the delay
behavioral intent through self-efficacy, which, in turn, was positively associated with the likelihood
of delay behavioral intent. This is an interesting outcome that implies preparation and experience are
important factors that positively influence one’s attitudes toward their abilities to protect themselves
in the event of CSZ earthquakes and tsunamis. At the same time, an increase in the level of self-
efficacy could create an effect of overconfidence. This attitude may result from a perception of
being sufficiently prepared (Lindell and Whitney 2000). Also, past experiences with disasters that did
not result in severe damages or a lack of personal experience with disasters (especially low-
probability, high impact) may lead to overconfidence (Eiser et al. 2012). That can potentially lead to
engaging in delay behaviors, because of one’s belief that they can still survive the tsunami even if
they take time to help family and friends before evacuating. In addition, in distant tsunami events,
previously experienced tsunami warnings that never manifested themselves into an actual tsunami
(i.e. false tsunami alarms), may create hesitation and mistrust in the population to react to warnings
in the future (Simmons and Sutter 2009). In the latter situation, as a part of public education, it is
important to emphasize that there is a difference between local and distant tsunami events. In the
event of a distant tsunami there is rarely a felt earthquake, therefore centrally issued warnings is the
indication for the public of an upcoming tsunami. In the event of a local tsunami, an earthquake is
inevitable, and it is the only necessary sign of an upcoming tsunami.
Response efficacy is the only cognitive variable that demonstrated a mediating effect
between tsunami-relevant knowledge confidence and preparation variables and leadership behavioral
intent according to Baron and Kenny’s (1986) criteria for mediation. It suggests that knowledge
confidence and preparation had a direct influence on the leadership behavioral intent with some of
the influence minimized or mediated by the response efficacy. It is encouraging to see that
99
respondents who claimed confidence in tsunami-relevant knowledge also reported a higher level of
response efficacy, suggesting a link between knowledge and a belief in the importance of protective
actions. Similarly, a higher number of preparedness measures was associated with a higher level of
response efficacy, implying that adoption of preparatory measures facilitates people’s favorable
evaluation of preparatory and tsunami protective actions. In turn, response efficacy displayed a
positive significant association with the leadership behavioral intent when controlling for all four
socio-environmental variables. Response efficacy did not play a mediating role between socio-
environmental and delay behavioral intent, nor did it show a direct influence on the delay behavioral
intent, suggesting that the delay behavioral intent construct is not a conceptual opposite to the
leadership behavioral construct.
Previous findings suggest that greater experience is positively related with an adoption of
recommended protective actions (Lindell et al. 2016); an increase in likelihood of evacuation to
higher ground (Charnkol and Tanaboriboon 2016); and with a recognition of the need to evacuate
quickly (Suppasri et al. 2013). This study did not find a strong support for these claims. Experience
did not show a significant association with leadership behavioral intent. This outcome likely reflects
how the experience variable was operationalized. Because of the rarity of local tsunamis on the west
coast of North America, to increase variation in responses the survey asked individuals to indicate
their experience with any natural disasters in the past. However, past experiences with any disaster
may not necessarily prepare individuals for a specific event such as a local tsunami. General disaster
experience can help people to be more psychologically prepared, but it does not necessarily lead to
the acquisition of specialized information. This assumption is supported in research by Goltz et al.
(1992) who analyzed evacuation behaviors in the U.S. 1987 Los Angeles County earthquake; by
Lindell et al. (2016) in the study of 2011 earthquake in Christchurch, New Zealand and Hitachi,
100
Japan; and by Shapira et al. (2018) in the evaluation of behavioral intentions in earthquake threats in
Israel. In all three cases, due to the lack of previous experience with large scale earthquakes, studies
did not find strong associations between previous experience and adoption of preparedness and
protective actions. Yet, in this study, experience showed an indirect influence on the leadership and
the delay behavioral intent though self-efficacy as discussed earlier. In addition, experience was
negatively significantly associated with the delay behavioral intent, when controlled for self-efficacy.
This outcome supports previous research findings, suggesting that the greater number of disasters
experienced in the past is associated with a decrease in the intent to adopt delay behaviors.
Confidence in tsunami-relevant knowledge demonstrated a significant positive association
with leadership behavioral intent independently and when controlling for each of the four cognitive
constructs. This finding aligns with the previously reported importance of knowledge in supporting
effective reactions in emergency situations. The knowledge confidence variable is not the most
precise assessment of people’s knowledge, because it captures one’s confidence rather than actual
knowledge. However, the connection between knowledge confidence and evacuation intention
suggests that people have access to information that motivates them to have the leadership
behavioral intent.
Similar to the knowledge construct, preparedness shows a significant positive association
with the leadership behavioral intent independently and when controlling for each of the four
cognitive constructs, suggesting the utility of engaging in preparedness actions. Higher number of
preparatory measures also showed association with higher level of self-efficacy and response
efficacy, instilling in people confidence to engage in protective actions. Preparation also revealed an
indirect positive significant association with the delay behavioral intent through self-efficacy,
101
suggesting that preparation could possibly lead to overconfidence in one’s skills, knowledge, and
ability to respond to disasters (Lindell and Whitney 2000).
Tsunami arrival time did not show a significant association with the leadership behavioral
intent. When controlled for response efficacy and perceived probability, wave arrival showed a
negative significant influence on the leadership behavioral intent, suggesting that individuals living in
households located further away from the coastline are less likely to exhibit leadership behavioral
intent. It is possible that individuals perceived themselves to be in no rush to evacuate to a high
ground immediately because of the closer location to safety. However, when controlled for self-
efficacy, wave arrival time was positively significantly associated with leadership behavioral intent,
implying that respondents perceived themselves to have the ability to do so due to the shorter
distance that they need to cover to get themselves to safety.
102
6 Limitations and suggestions for future research
This research examined individual perceptions and behavioral intentions of the Oregon
coastal population under the threat of a M9 CSZ earthquake and tsunami. The goal of this study was
to understand how individuals at risk along the Oregon coast would respond in the event of the
tsunami and what factors may explain variations in their behavioral intentions. A structured, self-
administered household survey instrument was used to obtain information from a randomly selected
sample of households in a purposively targeted community of Seaside, Oregon. There are some
limitations associated with the implemented research design and methodology that could potentially
limit the quality of findings and the ability to address study goals, even though it is difficult to assess
the true extent of the limitations.
Generalizability of results to a broader Oregon coastal population can be questionable due to
the nature of the sampling design. First, non-probability sampling was used by purposively selecting
Seaside, Oregon as a population of interest. In the methods section I discussed how Seaside is
representative of other coastal communities vulnerable to CSZ earthquakes and tsunamis in terms of
number of individuals and businesses exposed to tsunami risks. However, one must be cautious
when considering application of findings outside of Seaside. Generalizability of results could also be
improved by sampling a greater number of individuals not exposed to a high level of tsunami risk.
There are communities in Oregon that are located high above sea level and are not exposed to
tsunami dangers. In addition, in this study, in order to increase the response rate, we targeted
households that are primary residency for their inhabitants. However, it is possible that seasonal
residents have different levels of risk perceptions and knowledge about tsunamis and tsunami
protective behaviors compared to primary residents. There is an opportunity for future studies,
given the appropriate timing of data collection, to assess opinions of coastal seasonal residents.
103
I can say with more confidence that results of this study are generalizable to the population
of Seaside, from which a probability sample was drawn. Yet, there could be issues of external and
internal validity related to the response rate of 22.2%. Some may argue that 22.2% is a low response
rate. However, mail-based household survey response rates are sensitive to locations and, in general,
are declining (Allen et al. 2015). Some researchers are switching to different modes of response
solicitation. Email-based surveys are becoming more popular. In a comparative analysis of multiple
modes of survey distribution, a sample that received a mail questionnaire displayed a 21.1% response
rate, while a sample that received a questionnaire via internet yielded a 42.9% response rate
(Ansolabehere and Schaffner 2014). In the study of hurricane Lili evacuation response rate equaled
to 24.6% (Kang et al. 2007).
However, low response rate does not automatically invalidate the results (Newman 2009;
Curtin et al. 2000; Keeter et al. 2000). Unless all participants in the sample respond to the survey
(100% response rate), there is a possibility of having a nonresponse bias in the sample. Yet, it is
almost impossible to conduct unbiased research. The importance is to be cognizant of possible
biases and try to minimize them. Low response rate could lead to survey level non-response bias,
meaning that the answers obtained from the respondents may not be representative of the views of
the individuals in the original sample, because those selected into the sample purposefully did not
return surveys. In practice, it is impossible to determine whether the data is missing completely at
random, because that would require comparing observed values of Y to missing values of Y that do
not exist (Lance and Vandenberg 2009). There is a possibility of self-selection of respondents in this
study because of the questionnaire subject. Only those concerned about the risks of CSZ
earthquakes and tsunamis answered the survey. It is possible to minimize the risk of a non-response
bias by comparing the survey sample to U.S. Census. However, one cannot compare sample and
104
population on the main variables of interest such as risk perceptions and behavioral intentions. As
such, I demonstrated the representativeness of the sample to the population of Seaside on main
demographic characteristics. The survey sample under-represented younger individuals, so the data
was weighted to give more value to observations from younger participants.
Low response rate could also result in a nonresponse bias as a function of a sample size.
That problem can lead to type II error: a failure to reject the null, affecting internal validity of the
results. Even when there may be a non-zero effect in the population, the sample may be too small to
yield statistically significant results. While the response rate affects the ability to generalize the results
to the population, sample size affects the certainty of the conclusions – that is, the ability to estimate
the true effect. The working sample of 206 in this study is considered an acceptable sample size
within survey research design.
Internal validity of the results could also be affected by the validity of survey construct
measurements. In this study, there are several variables that could be improved to ensure more valid
outcomes. When thinking about construct validity, one must consider whether a variable is
measuring what it is supposed to measure according to theory and whether the questionnaire
wording and formatting elicits appropriate responses. This study followed Dillman et al. (2014)
questionnaire design, relying on previous research for question design to make them as clear as
possible. Yet, there is always a possibility that questions could be misinterpreted, although pre-
testing of the instrument was conducted. There is especially a higher risk of questions’
misinterpretation when trying to assess people’s perceptions because of vagueness of the concepts.
As an alternative way of question construction, behavioral intentions could be assessed by asking
respondents to rank the likelihood of adopting each behavior, to avoid the possibility of respondents
indicating the same likelihood of engaging in conceptually different behaviors.
105
In terms of concepts’ accuracy in representing theory, response efficacy measurements could
be improved to better assess association with evacuation behavioral intention. Instead of asking
about one’s perception of efficacy of preparedness, it is better to assess one’s perceptions of efficacy
of protecting themselves in an earthquake and tsunami. It is important to tie perception
measurements more precisely to the behavior that the cognitive concepts are trying predict. In
addition, experience with any natural hazards maybe not be the best proxy for understanding how
experience influences intentions of a specific behavior such as evacuation in tsunamis. Yet, in this
study, the experience construct was difficult to modify to measure experience in terms of experience
with tsunami disaster events because of the rarity of tsunamis on the Oregon coast.
In terms of timing of the survey distribution, it is important to consider a possibility of
unexpected events that could ‘artificially’ alter people’s perceptions from a normal level. Risk
perceptions are easily influenced by emergence of new information that specifically targets certain
images related to risk. For example, Dunn et al. (2016) in a survey of residents in Washington,
Oregon, and California about perceptions of earthquake hazards found that the New Yorker’s “The
Really Big One” article about threats of CSZ earthquakes and tsunamis, and the film “San Andreas”
about earthquake threats in California had increased people’s risk perceptions of earthquakes,
suggesting that mass media have an influence on people’s perceptions. Increase in awareness of risks
because of media campaigns or hazard events around the globe is beneficial from the disaster
preparedness perspective, but from the research design perspective these types of events may create
an above normal level of risk perceptions for a short period of time, compromising internal validity
because of possible inconsistency of results.
Also, it could be useful to consider the role of group dynamics and social context more
explicitly is shaping people’s perceptions and behavior. Solberg et al. (2010) argue that the limitation
106
of social psychological studies is that they have a narrow focus on the primacy of individual
cognitive processes in explaining how people interpret information and act to protect themselves
from risks, without proper regard for influence of wider social contexts, community and
neighborhoods. They argue that because of the lack of such considerations, ‘one size fits all’
preparedness programs generally do not motivate people to prepare. While this study did not
thoroughly consider cultural approaches to understand people’s perceptions and behavioral
intentions, important considerations were made to capture certain aspects of place uniqueness. For
instance, experience variables had absorbed some of the context related variation. In addition,
demographic variables and analysis of variance were employed specifically to identify groups that are
more susceptible to risks and difficulties of performing protective actions. Applying cultural
paradigms in research on natural disaster is highly important and can yield significant results,
because people’s attitudes and perceptions are shaped by fundamental beliefs and norms, according
to theories of behavior. However, in light of limited time and considerations of the survey length,
cultural questions were not the main focus in this study.
As an opportunity for future research, soliciting knowledge and risk perceptions of CSZ
earthquakes and tsunami from tourist populations could give an important insight into the level of
preparedness of coastal communities for the disaster. Tourists are generally the most vulnerable
population group in a disaster context, because they tend to be less familiar with areas they visit,
including evacuation routes and safety procedures compare to the resident population (Arce et al.
2017). According to Raskin and Wang (2016) the population of Seaside can double and triple during
summer months because of the influx of tourists. Targeting tourists in Seaside could be a way to
compare different sub-groups allowing to examine how cultural differences between populations
influence outcomes, behaviors, and perceptions. Yet, tourists are difficult to target as study
107
participants. Different research designs rather than self-administered mail survey would yield better
results for understanding tourists’ behavior.
In terms of data gathering techniques, other approaches could be implemented to reach
residents of coastal communities as well, besides household self-administered surveys. As this study
revealed, people were not motivated to respond to the questionnaire. Therefore, perhaps in-person
interviews or onsite surveys could solicit more responses. With in-person data gathering techniques
it is also possible to reach more targeted audiences such as low-income neighborhoods or younger
populations that are generally under-represented in household survey samples, as was indicated in
this study.
Future research efforts may consider using Structural Equation Modeling (SEM) in testing
for mediation effects. SEM uses a path diagram and system of linked regression-style equations to
capture relationships among observed and unobserved variables. Unlike a hierarchical regression
approach to testing mediation that requires making an assumption about dependent and
independent variables, SEM does not make such assumptions and can tests multiple directional
relationships between different variables at the same time, which may exist in cross-sectional designs
(Gunzler et al. 2013). More importantly, however, a hierarchical regression approach applied in this
study assumes no measurement errors in variables to achieve unbiased estimates of mediation
effects. Mediators often are latent constructs of perceptions and attitudes and measurement errors
are almost unavoidable (Baron and Kenny 1986). It is likely that there are measurements errors
within cognitive constructs in this study. SEM is claimed to provide a better tool for estimating
latent variables with several indicators and control for measurement errors in the process of
examination of relationships among variables (Preacher and Hayes 2004; Cheung and Lau 2008).
108
7 Key findings and policy recommendations
The state of Oregon in general and Oregon coastal communities in particular face high risks
of inevitable future Cascadia Subduction Zone earthquakes and tsunamis, as advocated in the 2013
Oregon Resilience Plan. According to Secretary of State Audit Report of January 2018, Oregon’s
preparedness level for the disaster event is poor. According to the Oregon Resilience Plan, calculated
impacts of CSZ earthquakes and tsunami include fatalities ranging from 1,250 to more than 10,000,
tens of thousands damaged buildings and displaced households, and more than $30 billion in direct
and indirect economic losses. Estimated recovery time to 90% of state’s pre-earthquake operational
levels is 1-3 years. To reduce post-disaster recovery time and improve state resilience, investments
are necessary in community-, infrastructural-, institutional-, and individual-level preparedness.
This study has been designed to understand individual-level preparedness, perceptions of
risks, and behavioral intentions of residents in Oregon coastal communities in the face of tsunami
risks. Because of the variety of possible behaviors that people can adopt in emergency situations,
measuring individual perceptions and behaviors is an integral part of preparedness and risk
reduction efforts to ensure efficient disaster response (Dunn et al. 2016). Knowledge about people’s
intentions, factors that influence intentions, and how intentions vary across different populations
can aid emergency personnel in targeting areas that need more attention. This information can be
used in developing public education and awareness programs to change people’s behavioral
intentions to align with recommended tsunami preparedness and response.
The key findings of this research include information about people’s behavioral intentions,
about how behavioral intentions are connected with people’s perceptions of risks and their abilities
to reduce those risks, about people’s current state of preparedness, and how people’s intentions vary
depending on their demographic characteristics.
109
Behavioral intentions
The top three behaviors that the respondents said they would very unlikely adopt in the
event of M9 CSZ earthquakes and tsunamis included continuing normal routine behavior (63.1%),
waiting for family at home (47.1%), and following friends and neighbors (43.7%). The top three
behaviors that respondents indicated they would very likely assume included evacuating to higher
ground immediately following the earthquake (51%), contacting loved ones (49.5%), and checking
social media, TV (40.3%). From a policy and disaster preparedness perspectives, it is encouraging to
see that more than half of the respondents indicated that they would evacuate to the high ground
immediately following the earthquake. Yet, there are those who have different behavioral intentions,
driven by a variety factors. The presence of variation in behavioral intentions creates a necessity to
design policies and programs to encourage people to avoid having a delay to the start of their
evacuation in a case of a local tsunami.
Preparation and information dissemination
The research revealed several factors that had an influence on behavioral intentions.
Research shows that hazard knowledge and availability of information about risks can motivate
people to prepare for disasters and lead to effective evacuation response (Walshe and Nunn 2012;
Dudley et al. 2011; Liu and Jiao, 2018; Gregg et al. 2006). As results of this study indicate, there are
individuals who are not fully confident in their knowledge of natural warning signs of tsunamis,
tsunami evacuation routes, tsunami safe areas in their community, and other tsunami related
information. For example, survey participants show the least confidence in knowing government
recommendations about how to prepare for tsunamis (26.2% are not confident and 21.4% are
somewhat confident) and tsunami safe places closest to their homes (17% are not confident and
11.2% are somewhat confident). A policy recommendation is to design programs that convey
110
information about tsunami evacuation areas, shortest routes to reach them, and the importance of
starting evacuation on their own immediately after an earthquake without waiting for other tsunami
warning signals should be a priority in preparing the public for successful tsunami response
(Takabatake et al. 2018; Hall et al. 2017).
Higher levels of preparation and tsunami-relevant knowledge also show a direct positive
association with the leadership behavioral intent, suggesting the importance of education and
preparation as a method to influence people’s behavioral intentions toward immediate evacuation
during tsunamis. While knowledge is important, it may be difficult to disseminate information to the
public in ways they will use. People are not active in attending public educational meetings. Survey
results indicate that 47.5% of respondents have not attended a meeting or have received information
about earthquakes and tsunamis in the past two years. A policy recommendation is to, use a diverse
set of channels to deliver information. One way to disseminate information about preparedness for
tsunamis is through education in public schools and encouraging children to discuss the information
with their parents. Mimura et al. (2011) showed a successful outcome of schools’ tsunami awareness
programs when during the 2011 great east Japan tsunami, about 3,000 school children were able to
successfully evacuate relying on ground shaking as a warning sign and prior to instructions by
authorities and the alert system.
Education of children (in concert with their parents) should include information about
appropriate behavior in the event of a local tsunami, evacuating to high ground immediately after
the earthquakes without waiting for parents, no matter where they find themselves when an
earthquake strikes. At the same time, it is important to deliver a message to parents that searching
for their children after the earthquake minimizes chances of survival in a tsunami for everyone,
including other people in a community because a chaotic movement of people creates traffic jams
111
and delays movement of people to the higher ground. In addition, emphasizing earthquake and
tsunami awareness in schools and appealing to younger generation to be the leaders in creating and
promoting a “culture of preparedness”, similar to places like Japan or Chile that experience
earthquakes and tsunamis on a regular basis, can help shift the whole community toward greater
awareness. With assistance of social media, broader audiences can be captured with tsunami
educational and preparedness messages and photos, shifting people’s thinking about CSZ
earthquakes and tsunamis from inevitable uncontrollable disaster to a state of preparedness and
awareness.
Informational messages and factors that influence intentions
Content of messages is as important as means of delivering messages. Certain messages are
better able to change peoples’ intentions than others (Dunn et al. 2016). This study revealed that
there are two kinds of messages that needs to be emphasized that could have an influence on
people’s behavior. One should be focused on perceptions of risk, because the analysis showed that
higher level of risk perception is associated with higher likelihood of having leadership behavioral
intention. And second message should emphasize importance of preparation to increase people’s
confidence in their abilities to protect themselves and confidence in protective actions. Peacock et
al. (2003) argues that public participation in awareness programs and increasing people’s cognitive
reactions to risk exposure are effective strategies to motivate people to adopt preparedness activities.
They also suggest that risk communication programs should emphasize repeated messages about
possible personal consequences and impacts of the disaster to increase the frequency of thought and
discussion about the hazard. Results of this study support the fact that higher level of risk
perception is associated with an intent for leadership behavior, suggesting that people who perceive
greater risk are more aware about the hazard. However, in addition, results of this research also
112
show the importance of promoting a message of how to increase one’s ability in protecting
themselves in disasters and emphasizing that preparation and specific actions during a tsunami can
make a difference in a successful outcome. Increasing the frequency and the degree of risk messages,
i.e. influencing public perceptions of risk, may be a catalyst for motivating a greater number of
people to put pressure on their elected officials to implement policies that contribute to
communities’ resilience in the face of risk from CSZ earthquakes and tsunamis.
Risk perception, self-efficacy, and response efficacy
The study showed that higher risk perceptions motivate people to adopt leadership
behaviors, while lower risk perceptions motivate people to engage in delay behaviors. Wave arrival
time, or the proxy for physical risk proximity was the only variable associated with perceived
severity. Wave arrival, together with experience, also displayed an indirect positive influence on the
leadership behavior via self-efficacy, which, in return, had a positive association with the leadership
behavior. This outcome suggests that confidence in one’s abilities to protect themself in disaster
situations, is an important factor that influences people’s intentions to adopt leadership behavior.
However, higher self-efficacy, especially in association with higher levels of preparation and
experience, is shown to have a positive influence on delay behavioral intentions, implying a certain
level of overconfidence among those that either had previous experience and believe that they are
able to protect themselves in tsunamis or among individuals who believe that they are well-prepared
to be able to afford delaying their evacuation and perhaps assisting others. While it is important to
increase individuals’ self-efficacy because it motivates people to adopt leadership behavior,
preparedness efforts must also convey the message of immediate evacuation as detrimental to one’s
safety in tsunamis, no matter how much one feels prepared for it.
113
Response efficacy played a mediating role between socio-environmental variables of
tsunami-relevant knowledge confidence and preparation and the leadership behavioral intent,
suggesting that an increase in one’s confidence about tsunami knowledge also improves one’s view
of the efficacy of preparing for earthquakes and tsunamis, which in turn promotes intentions to
engage in leadership behaviors. This result suggests that increasing people’s confidence in their
knowledge about tsunamis can lead to positive outcomes. It also suggests that promoting the
message that preparation and protective actions can make a difference in disaster survival is as
important as conveying the message of potential negative consequences of a tsunami. That is,
conveying positive information about the importance of preparation and options for survival may be
as effective as trying to scare people into making preparatory decisions.
Vulnerable populations
Another key finding includes the fact that there are vulnerable groups that may need greater
assistance during the disaster event due to various situational barriers and may require greater
outreach efforts for information dissemination because of other barriers that prevent them from
easily acquiring knowledge. As the results of the survey show, older residents require greater
attention, especially considering their large representation in the population along the coast. Older
residents tend to have lower self-efficacy, lower confidence in tsunami-relevant knowledge, more
likely to have intentions to continue normal routine behavior, and less likely to engage in the
leadership behavior in the tsunami situation. These attitudes can arise from the fact that older
residents also tend to have mobility issues, suggesting that they may not expect to be able to
evacuate from the risk zone prior to the onset of the tsunami. Oregon coastal communities are
home to many retirees and assisted living facilities, making tsunami preparedness particularly
challenging, because of the central role that physical mobility plays in successful outcomes.
114
Evacuation models developed by Wang et al. (2015) suggest that tsunami mortality rate is sensitive to
the variations in walking speed of the evacuee population. A policy recommendation is to pay more
attention to vulnerable groups, specifically older populations and people with mobility problems,
when creating educational programs, improving evacuation routes and roads, designing future plans
for cities, and other tsunami preparedness actions.
115
8 Conclusion
Main goals of this research were to provide information on public awareness of the Cascadia
Subduction Zone earthquakes and tsunamis, people’s perceptions of risk and behavioral intentions,
and factors that influence attitudes and intentions, to expand previous research on human behavior
in a rapid onset natural hazard context and to suggests ways in which people’s intentions can be
influenced to motivate preparedness for the anticipated CSZ earthquakes and tsunamis in Oregon.
To explore these goals this research applied a survey questionnaire to a random sample of
households in the coastal town of Seaside, Oregon. The survey questionnaire was designed based on
theories of cognition and previous research findings in the natural disaster literature. The research
framework developed in this study examined associations between socio-environmental factors,
variables of cognition, and behavioral intentions. Results of mediation analysis and analysis of
variance supported previous research findings, showing that risk perception had a positive
association with evacuation behavioral intentions. Self-efficacy and response efficacy also displayed
an influence on behavioral intentions directly and served as a moderator between socio-
environmental factors and behavioral intentions. In terms of public disaster preparedness and
recommendations for emergency planning, respondents showed relatively high awareness of risks of
CSZ earthquakes and tsunamis, with 72.8% of respondents having perceptions that M9 Cascadia
earthquake and tsunami would likely lead to injuries and 85.4% of respondents believing that M9
CSZ earthquake and tsunami would create a life-threatening situation. However, the study also
found that respondents reported low confidence in their knowledge of government
recommendations about how to prepare for tsunamis and about tsunami safe locations in their
community. In addition, almost half of the respondents did not attended tsunami informational
meetings and did not make a communication plan with family or friends. Educational outreach,
116
especially appealing to school students and their parents as a possible effective channel of
information dissemination, was proposed as one of the most salient recommendations to improve
emergency preparedness and shift Oregon coastal communities’ culture toward the “culture of
preparedness”.
117
Bibliography
Abraham, C. S., Sheeran, P., Abrams, D., & Spears, R. (1994). Exploring teenagers' adaptive and maladaptive thinking in relation to the threat of HIV infection. Psychology and Health, 9(4), 253-272.
Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological bulletin, 84(5), 888.
Ajzen, I., & Fishbein, M. (2005). The influence of attitudes on behavior. The handbook of attitudes, 173(221), 31.
Aizen, I. (2010). Behavioral interventions based on the theory of planned behavior. Alexander, D., & Magni, M. (2013). Mortality in the l'Aquila (central Italy) earthquake of 6 April
2009. PLoS currents, 5. Allen, J., Liptak, K., Guo, J., & Worasinchai, L. (2015). Mail survey response rates in academic
research: A multi-disciplinary trend analysis. In Proceedings of the 1999 Academy of Marketing Science (AMS) Annual Conference(pp. 250-254). Springer, Cham.
Ansolabehere, S., & Schaffner, B. F. (2014). Does survey mode still matter? Findings from a 2010 multi-mode comparison. Political Analysis, 22(3), 285-303.
Armitage, C. J., & Christian, J. (2003). From attitudes to behaviour: Basic and applied research on the theory of planned behaviour. Current Psychology, 22(3), 187-195.
Apatu, E. J., Gregg, C. E., Richards, K., Sorensen, B. V., & Wang, L. (2013). Factors Affecting Household Adoption of an Evacuation Plan in American Samoa after the 2009 Earthquake and Tsunami. Hawai'i Journal of Medicine & Public Health, 72(8), 267.
Apatu, E. J., Gregg, C. E., Wood, N. J., & Wang, L. (2016). Household evacuation characteristics in American Samoa during the 2009 Samoa Islands tsunami. Disasters, 40(4), 779-798.
Arias, J. P., Bronfman, N. C., Cisternas, P. C., & Repetto, P. B. (2017). Hazard proximity and risk perception of tsunamis in coastal cities: Are people able to identify their risk?. PLoS one, 12(10), e0186455.
Arce, R. S. C., Onuki, M., Esteban, M., & Shibayama, T. (2017). Risk awareness and intended tsunami evacuation behaviour of international tourists in Kamakura City, Japan. International Journal of Disaster Risk Reduction, 23, 178-192.
Baker, E. J. (1991). Hurricane evacuation behavior. International Journal of Mass Emergencies and Disasters, 9(2), 287–310.
Barberi, F., Davis, M. S., Isaia, R., Nave, R., & Ricci, T. (2008). Volcanic risk perception in the Vesuvius population. Journal of volcanology and geothermal research, 172(3-4), 244-258.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of personality and social psychology, 51(6), 1173.
Basolo, V., Steinberg, L. J., Burby, R. J., Levine, J., Cruz, A. M., & Huang, C. (2009). The effects of confidence in government and information on perceived and actual preparedness for disasters. Environment and Behavior, 41(3), 338-364.
Becker, J. S., Paton, D., Johnston, D. M., Ronan, K. R., & McClure, J. (2017). The role of prior experience in informing and motivating earthquake preparedness. International Journal of Disaster Risk Reduction, 22, 179-193.
Bell, K. P., Huppert, D., & Johnson, R. L. (2003). Willingness to pay for local coho salmon enhancement in coastal communities. Marine Resource Economics, 18(1), 15-31.
118
Bernard, E. N. (Ed.). (2005). Developing tsunami-resilient communities: the national tsunami hazard mitigation program. Springer Science & Business Media.
Bird, D. K. (2009). The use of questionnaires for acquiring information on public perception of natural hazards and risk mitigation-a review of current knowledge and practice. Natural Hazards and Earth System Sciences, 9(4), 1307.
Bird, D., & Dominey-Howes, D. (2008). Testing the use of a ‘questionnaire survey instrument to investigate public perceptions of tsunami hazard and risk in Sydney, Australia. Natural Hazards, 45(1), 99-122.
Blumer, H. (1955). Attitudes and the social act. Social problems, 3(2), 59-65. Bollen, K. A., & Stine, R. (1990). Direct and indirect effects: Classical and bootstrap estimates of
variability. Sociological methodology, 115-140. Bourque, L. B., Regan, R., Kelley, M. M., Wood, M. M., Kano, M., & Mileti, D. S. (2013). An
examination of the effect of perceived risk on preparedness behavior. Environment and Behavior, 45(5), 615–649.
Bright, A. D., & Manfredo, M. J. (1996). A conceptual model of attitudes toward natural resource issues: a case study of wolf reintroduction. Human Dimensions of Wildlife, 1(1), 1-21.
Bulmer, M. G. (1979). Principles of statistics. Courier Corporation. Burby, R. J., Steinberg, L. J., & Basolo, V. (2003). The tenure trap: The vulnerability of renters to
joint natural and technological disasters. Urban Affairs Review, 39(1), 32-58. Burton, I., R.W. Kates, G.F. White (1993). The environment as hazard. Guilford Press. Campbell, D. T. (1963). Social attitudes and other acquired behavioral dispositions. In S. Koch
(Ed.), Psychology: A study of a science (Vol. 6, pp. 94–172). New York: McGraw-Hill. Castro, C. P., Sarmiento, J. P., Edwards, R., Hoberman, G., & Wyndham, K. (2017). Disaster risk
perception in urban contexts and for people with disabilities: case study on the city of Iquique (Chile). Natural hazards, 86(1), 411-436.
Charnkol, T., & Tanaboriboon, Y. (2006). Tsunami evacuation behavior analysis: One step of transportation disaster response. IATSS research, 30(2), 83-96.
Cheung, G. W., & Lau, R. S. (2008). Testing mediation and suppression effects of latent variables: Bootstrapping with structural equation models. Organizational research methods, 11(2), 296-325.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, S95–S120.
Corey, S. M. (1937). Professed attitudes and actual behavior. Journal of educational psychology, 28(4), 271. Couling, M. (2014). Tsunami risk perception and preparedness on the east coast of New Zealand
during the 2009 Samoan Tsunami warning. Natural hazards, 71(1), 973-986. Crossett, K. M. (2005). Population trends along the coastal United States: 1980-2008. Government Printing
Office. Curtin, R., Presser, S., & Singer, E. (2000). The effects of response rate changes on the index of
consumer sentiment. Public Opinion Quarterly, 64(4), 413-428. Deutscher, I. (1966). Words and deeds: Social science and social policy. Social Problems, 13, 235–
254. Dash, N., & Gladwin, H. (2007). Evacuation decision making and behavioral responses: Individual
and household. Natural Hazards Review, 8(3), 69-77. DeVellis, Robert F. (2016) Scale Development: Theory and Applications. Sage publications. Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, phone, mail, and mixed-mode surveys: the
tailored design method. John Wiley & Sons.
119
DiStefano, C., Zhu, M., & Mindrila, D. (2009). Understanding and using factor scores: Considerations for the applied researcher. Practical Assessment, Research & Evaluation, 14(20), 1-11.
DLCD (2016) Tsunami Hazard Summary Drabek, T. E. (2012). Human system responses to disaster: An inventory of sociological findings. Springer
Science & Business Media. Douglas, M., & Wildavsky, A. (1982). Risk and culture. Berkeley. University of California Press, 272, 10-
15. Dudley, W. C., Whitney, R., Faasisila, J., Fonolua, S., Jowitt, A., & Chan-Kau, M. (2011). Learning
from the victims: new physical and social science information about tsunamis from victims of the September 29, 2009 event in Samoa and American Samoa. Earth-science reviews, 107(1), 201-206.
Dunn, P. T., Ahn, A. Y., Bostrom, A., & Vidale, J. E. (2016). Perceptions of earthquake early warnings on the US West Coast. International Journal of Disaster Risk Reduction, 20, 112-122.
Duval, T. S., & Mulilis, J. P. (1999). A Person‐Relative‐to‐Event (PrE) approach to negative threat appeals and earthquake preparedness: A field study. Journal of applied social psychology, 29(3), 495-516.
Eiser, J. R., Bostrom, A., Burton, I., Johnston, D. M., McClure, J., Paton, D., ... & White, M. P. (2012). Risk interpretation and action: A conceptual framework for responses to natural hazards. International Journal of Disaster Risk Reduction, 1, 5-16.
Elliott, J. R., Haney, T. J., & Sams-Abiodun, P. (2010). Limits to social capital: Comparing network assistance in two New Orleans neighborhoods devastated by Hurricane Katrina. The Sociological Quarterly, 51(4), 624–648.
EM-DAT. International Disaster Database. http://www.emdat.be/ Esteban, M., Tsimopoulou, V., Mikami, T., Yun, N. Y., Suppasri, A., & Shibayama, T. (2013).
Recent tsunamis events and preparedness: Development of tsunami awareness in Indonesia, Chile and Japan. International Journal of Disaster Risk Reduction, 5, 84-97.
Emerson, R. M. (1976). Social exchange theory. Annual Review of Sociology, 2(1), 335–362. Faas, A. J., & Barrios, R. E. (2015). Applied anthropology of risk, hazards, and disasters. Human
Organization, (2015), 287–295. Fekete, A., Tzavella, K., Armas, I., Binner, J., Garschagen, M., Giupponi, C., ... & Serre, D. (2015).
Critical data source; tool or even infrastructure? Challenges of geographic information systems and remote sensing for disaster risk governance. ISPRS International Journal of Geo-Information, 4(4), 1848-1869.
Festinger, L. (1964). Behavioral support for opinion change. The Public Opinion Quarterly, 28, 404–417.
Field, A. (2009). Discovering statistics using SPSS. Sage publications. Fincham, J. E. (2008). Response rates and responsiveness for surveys, standards, and the Journal.
American journal of pharmaceutical education, 72(2), 43. Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S., & Combs, B. (1978). How safe is safe enough? A
psychometric study of attitudes towards technological risks and benefits. Policy Sciences, 9(2), 127–152.
Fishbein, M., & Ajzen, I. (1975). Belief attitude, intention and behaviour: An introduction to theory and research. Reading, MA: Addison-Wesley
120
Floyd, D. L., Prentice‐Dunn, S., & Rogers, R. W. (2000). A meta‐analysis of research on protection motivation theory. Journal of applied social psychology, 30(2), 407-429.
Fothergill, A. (1996). Gender, risk, and disaster. International Journal of Mass Emergencies and Disasters, 14(1), 33-56.
Fraser, S., Leonard, G., & Johnston, D. M. (2013). Intended evacuation behaviour in a local earthquake and tsunami at Napier, New Zealand. GNS Science.
Fraser, S., Matsuo, I., Leonard, G. S., & Murakami, H. (2012). Tsunami evacuation: Lessons from the Great East Japan earthquake and tsunami of March 11th, 2011 (GNS Science Report 2012/17).
Fraser, S. A., Doyle, E. E., Wright, K. C., Potter, S. H., McClure, J., Johnston, D. M., ... & Johal, S. (2016). Tsunami response behaviour during and following two local-source earthquakes in Wellington, New Zealand. International Journal of Disaster Risk Reduction, 16, 123-133.
Fritz, C. E., & Williams, H. B. (1957). The human being in disasters: A research perspective. The Annals of the American Academy of Political and Social Science, 309(1), 42-51.
Gaillard, J. C., Clavé, E., Vibert, O., Denain, J. C., Efendi, Y., Grancher, D., ... & Setiawan, R. (2008). Ethnic groups’ response to the 26 December 2004 earthquake and tsunami in Aceh, Indonesia. Natural Hazards, 47(1), 17-38.
Garber-Yonts, B., Kerkvliet, J., & Johnson, R. (2004). Public values for biodiversity conservation policies in the Oregon Coast Range. Forest Science, 50(5), 589-602.
Gierlach, E., Belsher, B. E., & Beutler, L. E. (2010). Cross‐cultural differences in risk perceptions of disasters. Risk Analysis, 30(10), 1539-1549.
Goldfinger, C., Nelson, C. H., Morey, A. E., Johnson, J. E., Patton, J. R., Karabanov, E., ... & Enkin, R. J. (2012). Turbidite event history: Methods and implications for Holocene paleoseismicity of the Cascadia subduction zone. US Geological Survey Professional Paper, 1661, 170.
Goltz, J. D., Russell, L. A., & Bourque, L. B. (1992). Initial behavioral response to a rapid onset disaster: a case study of the October 1, 1987, Whittier Narrows earthquake. International Journal of Mass Emergencies and Disasters, 10(1), 43-69.
Goszczynska, M., Tyszka, T., & Slovlc, P. (1991). Risk perception in Poland: A comparison with three other countries. Journal of Behavioral Decision Making, 4(3), 179-193.
Gregg, C. E., Houghton, B. F., Paton, D., Swanson, D. A., & Johnston, D. M. (2004). Community preparedness for lava flows from Mauna Loa and Hualālai volcanoes, Kona, Hawai ‘i. Bulletin of Volcanology, 66(6), 531-540.
Gregg, C. E., Houghton, B. F., Paton, D., Lachman, R., Lachman, J., Johnston, D. M., & Wongbusarakum, S. (2006). Natural warning signs of tsunamis: human sensory experience and response to the 2004 great Sumatra earthquake and tsunami in Thailand. Earthquake Spectra, 22(S3), 671-691.
Gregg, C. E., Houghton, B. F., Paton, D., Johnston, D. M., Swanson, D. A., & Yanagi, B. S. (2007). Tsunami warnings: understanding in Hawai ‘i. Natural Hazards, 40(1), 71-87.
Griffin, R. J., Dunwoody, S., & Neuwirth, K. (1999). Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors. Environmental research, 80(2), S230-S245.
Grothmann, T., & Reusswig, F. (2006). People at risk of flooding: why some residents take precautionary action while others do not. Natural hazards, 38(1), 101-120.
Groves, R. M. (2004). Survey errors and survey costs (Vol. 536). John Wiley & Sons.
121
Gunzler, D., Chen, T., Wu, P., & Zhang, H. (2013). Introduction to mediation analysis with structural equation modeling. Shanghai archives of psychiatry, 25(6), 390.
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2006). Multivariate data analysis: Pearson Prentice Hall Upper Saddle River. NJ.
Hall, T. E., & Slothower, M. (2009). Cognitive factors affecting homeowners' reactions to defensible space in the Oregon Coast Range. Society and Natural Resources, 22(2), 95-110.
Hall, S., Pettersson, J., Meservy, W., Harris, R., Agustinawati, D., Olson, J., & McFarlane, A. (2017). Awareness of tsunami natural warning signs and intended evacuation behaviors in Java, Indonesia. Natural Hazards, 89(1), 473-496.
Hayes, A. F. (2013). An introduction to mediation, moderation, and conditional process analysis. New York: The Guilford Press.
Hayes, A. F., & Preacher, K. J. (2014). Statistical mediation analysis with a multicategorical independent variable. British Journal of Mathematical and Statistical Psychology, 67(3), 451-470.
Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion; psychological studies of opinion change.
Huang, S. K., Lindell, M. K., & Prater, C. S. (2016). Multistage model of hurricane evacuation decision: empirical study of hurricanes Katrina and Rita. Natural Hazards Review, 18(3), 05016008.
Huang, S. K., Lindell, M. K., & Prater, C. S. (2016). Who leaves and who stays? A review and statistical meta-analysis of hurricane evacuation studies. Environment and Behavior, 48(8), 991-1029.
Huang, S. K., Lindell, M. K., Prater, C. S., Wu, H. C., & Siebeneck, L. K. (2012). Household evacuation decision making in response to Hurricane Ike. Natural Hazards Review, 13(4), 283-296.
Hurlbert, J. S., Haines, V. A., & Beggs, J. J. (2000). Core networks and tie activation: What kinds of routine networks allocate resources in nonroutine situations? American Sociological Review, 598–618.
Iemura, H., Takahashi, Y., Harris Pradono, M., Sukamdo, P., & Kurniawan, R. (2006). Earthquake and tsunami questionnaires in Banda Aceh and surrounding areas. Disaster Prevention and Management: An International Journal, 15(1), 21-30.
Imamura, F. (2009). Dissemination of Information and Evacuation Procedures in the 2004–2007 tsunamis, including the 2004 Indian Ocean. Journal of Earthquake and Tsunami, 3(02), 59-65.
Jaeger, C. C., Webler, T., Rosa, E. A., & Renn, O. (2013). Risk, uncertainty and rational action. Routledge.
Janis, I. L. (1967). Effects of fear arousal on attitude change: Recent developments in theory and experimental research. Advances in experimental social psychology, 3, 166-224.
Janis, I. L., & Leventhal, H. (1968). Human reactions to stress. Handbook of personality theory and research, 1041-1085.
Johnston, D. M., Bebbington Chin-Diew Lai, M. S., Houghton, B. F., & Paton, D. (1999). Volcanic hazard perceptions: comparative shifts in knowledge and risk. Disaster Prevention and Management: An International Journal, 8(2), 118-126.
Johnston, D., Paton, D., Crawford, G. L., Ronan, K., Houghton, B., & Bürgelt, P. (2005). Measuring tsunami preparedness in coastal Washington, United States. Natural Hazards, 35(1), 173-184.
Jon, I., Lindell, M. K., Prater, C. S., Huang, S. K., Wu, H. C., Johnston, D. M., ... & McClure, J. (2016). Behavioral Response in the Immediate Aftermath of Shaking: Earthquakes in
122
Christchurch and Wellington, New Zealand, and Hitachi, Japan. International journal of environmental research and public health, 13(11), 1137.
Jones, E. C., & Faas, A. J. (2017). Social Network Analysis of Disaster Response, Recovery, and Adaptation. Butterworth-Heinemann.
Jones, E. C., Faas, A. J., Murphy, A. D., Tobin, G. A., Whiteford, L. M., & McCarty, C. (2013). Cross-cultural and site-based influences on demographic, well-being, and social network predictors of risk perception in hazard and disaster settings in Ecuador and Mexico. Human Nature, 24(1), 5–32.
Kailes, J. I., & Enders, A. (2007). Moving beyond “special needs” A function-based framework for emergency management and planning. Journal of Disability Policy Studies, 17(4), 230-237.
Kang, J. E., Lindell, M. K., & Prater, C. S. (2007). Hurricane evacuation expectations and actual behavior in Hurricane Lili. Journal of Applied Social Psychology, 37(4), 887-903.
Kaniasty, K., & Norris, F. H. (1993). A test of the social support deterioration model in the context of natural disaster. Journal of Personality and Social Psychology, 64(3), 395.
Kaniasty, K., & Norris, F. H. (2004). Social support in the aftermath of disasters, catastrophes, and acts of terrorism: Altruistic, overwhelmed, uncertain, antagonistic, and patriotic communities. Bioterrorism: Psychological and Public Health Interventions, 3, 200–229.
Kapucu, N. (2008). Collaborative emergency management: better community organising, better public preparedness and response. Disasters, 32(2), 239–262.
Keeter, S., Miller, C., Kohut, A., Groves, R. M., & Presser, S. (2000). Consequences of reducing nonresponse in a national telephone survey. Public opinion quarterly, 64(2), 125-148.
Knocke, E. T., & Kolivras, K. N. (2007). Flash flood awareness in southwest Virginia. Risk analysis, 27(1), 155-169.
Kurita, T., Arakida, M., & COLOMBAGE, S. R. (2007). Regional characteristics of tsunami risk perception among the tsunami affected countries in the Indian Ocean. Journal of Natural Disaster Science, 29(1), 29-38.
Lachman, R., Tatsuoka, M., & Bonk, W. J. (1961). Human behavior during the tsunami of May 1960. Science, 133(3462), 1405-1409.
Lance, C. E., & Vandenberg, R. J. (Eds.). (2009). Statistical and methodological myths and urban legends: Doctrine, verity and fable in the organizational and social sciences. Taylor & Francis.
LaPiere, R. T. (1934). Attitudes vs. actions. Social forces, 13(2), 230-237.
Lee, J. E., & Lemyre, L. (2009). A social‐cognitive perspective of terrorism risk perception and individual response in Canada. Risk analysis, 29(9), 1265-1280.
Lee, Y. J., Tung, C. M., & Lin, S. C. (2018). Attitudes to climate change, perceptions of disaster risk, and mitigation and adaptation behavior in Yunlin County, Taiwan. Environmental Science and Pollution Research, 1-11.
Lin, S., Shaw, D., & Ho, M. C. (2008). Why are flood and landslide victims less willing to take mitigation measures than the public?. Natural Hazards, 44(2), 305-314.
Lindell, M. K., & Hwang, S. N. (2008). Households' perceived personal risk and responses in a multi hazard environment. Risk Analysis, 28(2), 539-556.
Lindell, M. K., & Perry, R. W. (1992). Behavioral foundations of community emergency planning. Hemisphere Publishing Corp.
Lindell, M. K., & Perry, R. W. (2000). Household adjustment to earthquake hazard: A review of research. Environment and behavior, 32(4), 461-501.
123
Lindell, M. K., & Perry, R. W. (2003). Communicating environmental risk in multiethnic communities (Vol. 7). Sage Publications.
Lindell, M. K., & Perry, R. W. (2012). The protective action decision model: theoretical modifications and additional evidence. Risk Analysis, 32(4), 616-632.
Lindell, M. K., & Prater, C. S. (2000). Household adoption of seismic hazard adjustments: A comparison of residents in two states. International Journal of Mass Emergencies and Disasters, 18, 317–338.
Lindell, M. K., & Prater, C. S. (2002). Risk area residents’ perceptions and adoption of seismic hazard adjustments. Journal of Applied Social Psychology, 32(11), 2377-2392.
Lindell, M. K., & Prater, C. S. (2010). Tsunami preparedness on the Oregon and Washington coast: Recommendations for research. Natural Hazards Review, 11(2), 69-81.
Lindell, M. K., Arlikatti, S., & Prater, C. S. (2009). Why people do what they do to protect against earthquake risk: Perceptions of hazard adjustment attributes. Risk Analysis, 29(8), 1072-1088.
Lindell, M. K., Prater, C. S., Gregg, C. E., Apatu, E. J., Huang, S. K., & Wu, H. C. (2015). Households' immediate responses to the 2009 American Samoa Earthquake and Tsunami. International journal of disaster risk reduction, 12, 328-340.
Lindell, M. K., Prater, C. S., Wu, H. C., Huang, S. K., Johnston, D. M., Becker, J. S., & Shiroshita, H. (2016). Immediate behavioural responses to earthquakes in Christchurch, New Zealand, and Hitachi, Japan. Disasters, 40(1), 85-111.
Lindell, M. K., & Whitney, D. J. (2000). Correlates of household seismic hazard adjustment adoption. Risk analysis, 20(1), 13-26.
Liu, T., & Jiao, H. (2018). How does information affect fire risk reduction behaviors? Mediating effects of cognitive processes and subjective knowledge. Natural Hazards, 1-23.
MacKinnon, D. P., & Dwyer, J. H. (1993). Estimating mediated effects in prevention studies. Evaluation review, 17(2), 144-158.
MacKinnon, D. P., Warsi, G., & Dwyer, J. H. (1995). A simulation study of mediated effect measures. Multivariate behavioral research, 30(1), 41-62.
Mas, E., Adriano, B., & Koshimura, S. (2013). An integrated simulation of tsunami hazard and human evacuation in La Punta, Peru. Journal of Disaster Research, 8(2), 285-295.
Mas, E., Suppasri, A., Imamura, F., & Koshimura, S. (2012). Agent-based simulation of the 2011 great east japan earthquake/tsunami evacuation: An integrated model of tsunami inundation and evacuation. Journal of Natural Disaster Science, 34(1), 41-57.
Mawson, A. R. (2005). Understanding mass panic and other collective responses to threat and disaster. Psychiatry: Interpersonal and biological processes, 68(2), 95-113.
McAdoo, B. G., Dengler, L., Prasetya, G., & Titov, V. (2006). Smong: How an oral history saved thousands on Indonesia’s Simeulue Island during the December 2004 and March 2005 tsunamis. Earthquake Spectra, 22(S3), 661-669.
McCarty, C., Killworth, P. D., & Rennell, J. (2007). Impact of methods for reducing respondent burden on personal network structural measures. Social Networks, 29(2), 300–315.
McCaughey, J. W., Mundir, I., Daly, P., Mahdi, S., & Patt, A. (2017). Trust and distrust of tsunami vertical evacuation buildings: Extending protection motivation theory to examine choices under social influence. International Journal of Disaster Risk Reduction, 24, 462-473.
McGuire, W. J. (1968). Personality and susceptibility to social influence. Handbook of personality theory and research, 2, 1130-1187.
124
Meyer, M. A., Mitchell, B., Purdum, J. C., Breen, K., & Iles, R. L. (2018). Previous Hurricane Evacuation Decisions and Future Evacuation Intentions among Residents of Southeast Louisiana. International Journal of Disaster Risk Reduction.
Mileti, D. (1999). Disasters by design: A reassessment of natural hazards in the United States. Joseph Henry Press.
Mileti, D. S., & Fitzpatrick, C. (1992). The causal sequence of risk communication in the Parkfield earthquake prediction experiment. Risk Analysis, 12(3), 393-400.
Milne, S., Sheeran, P., & Orbell, S. (2000). Prediction and intervention in health‐related behavior: A
meta‐analytic review of protection motivation theory. Journal of Applied Social Psychology, 30(1), 106-143.
Mimura, N., Yasuhara, K., Kawagoe, S., Yokoki, H., & Kazama, S. (2011). Damage from the Great East Japan Earthquake and Tsunami-a quick report. Mitigation and Adaptation Strategies for Global Change, 16(7), 803-818.
Montano, D. E., & Kasprzyk, D. (2015). Theory of reasoned action, theory of planned behavior, and the integrated behavioral model. Health behavior: Theory, research and practice, 95-124.
Mostafizi, A., Wang, H., Cox, D., Cramer, L. A., & Dong, S. (2017). Agent-based tsunami evacuation modeling of unplanned network disruptions for evidence-driven resource allocation and retrofitting strategies. Natural Hazards, 88(3), 1347-1372.
Mulilis, J. P., Duval, T. S., & Rogers, R. (2003). The Effect of a Swarm of Local Tornados on
Tornado Preparedness: A Quasi‐Comparable Cohort Investigation. Journal of Applied Social Psychology, 33(8), 1716-1725.
Mulilis, J. P., & Lippa, R. (1990). Behavioral change in earthquake preparedness due to negative threat appeals: A test of protection motivation theory. Journal of Applied Social Psychology, 20(8), 619-638.
Murakami, H., Takimoto, K., & Pomonis, A. (2012). Tsunami Evacuation Process and Human Loss Distribution in the 2011 Great East Japan Earthquake-A Case Study of Natori City, Miyagi Prefecture. In 15th World Conference on Earthquake Engineering (pp. 1-10).
Murakami, H., Yanagihara, S., Goto, Y., Mikami, T., Sato, S., & Wakihama, T. (2014). Study on casualty and tsunami evacuation behavior in Ishinomaki city—Questionnaire survey for the 2011 Great East Japan Earthquake—. In Proc. 10th US National Conf. on Earthq. Engr., USB publication Paper (No. 1280).
National Geophysical Data Center / World Data Service (NGDC/WDS): Global Historical Tsunami Database. National Geophysical Data Center, NOAA. doi:10.7289/V5PN93H7 [January 23, 2018]
Natural Hazards Mitigation Plan (2015) Coastal County. State of Oregon. Nelson, K. C., Monroe, M. C., Johnson, J. F., & Bowers, A. (2005). Living with fire: homeowner
assessment of landscape values and defensible space in Minnesota and Florida, USA. International Journal of Wildland Fire, 13(4), 413-425.
Newman, D. A. (2009). Missing data techniques and low response rates. Statistical and methodological myths and urban legends: Doctrine, verity and fable in the organizational and social sciences, 7.
Norris, F. H., Stevens, S. P., Pfefferbaum, B., Wyche, K. F., & Pfefferbaum, R. L. (2008). Community resilience as a metaphor, theory, set of capacities, and strategy for disaster readiness. American journal of community psychology, 41(1-2), 127-150.
125
Okumura, Y., Harada, K., & Kawata, Y. (2011). Evacuation Behavior in the 29 September 2009 Samoa Islands Region Earthquake Tsunami. Journal of Earthquake and Tsunami, 5(03), 217-229.
O'Sullivan, T. L., Kuziemsky, C. E., Toal-Sullivan, D., & Corneil, W. (2013). Unraveling the complexities of disaster management: A framework for critical social infrastructure to promote population health and resilience. Social Science & Medicine, 93, 238-246.
Paton, D. (2003). Disaster preparedness: a social-cognitive perspective. Disaster Prevention and Management: An International Journal, 12(3), 210-216.
Paton, D., Smith, L., & Johnston, D. M. (2000). Volcanic hazards: risk perception and preparedness. New Zealand Journal of Psychology, 29(2), 86.
Paton, D., Houghton, B. F., Gregg, C. E., Gill, D. A., Ritchie, L. A., McIvor, D., ... & Johnston, D. M. (2008). Managing tsunami risk in coastal communities: Identifying predictors of preparedness. Australian Journal of Emergency Management, The, 23(1), 4.
Peacock, W. G. (2003). Hurricane mitigation status and factors influencing mitigation status among Florida’s single-family homeowners. Natural Hazards Review, 4(3), 149-158.
Peacock, W. G., Brody, S. D., & Highfield, W. (2005). Hurricane risk perceptions among Florida's single family homeowners. Landscape and Urban Planning, 73(2-3), 120-135.
Perry, R. W. (1979). Evacuation decision-making in natural disasters. Mass Emergencies, 4(1), 25–38. Perry, R. W., & Lindell, M. K. (2008). Volcanic risk perception and adjustment in a multi-hazard
environment. Journal of Volcanology and Geothermal Research, 172(3-4), 170-178. Potthoff, R. F., Tudor, G. E., Pieper, K. S., & Hasselblad, V. (2006). Can one assess whether missing
data are missing at random in medical studies?. Statistical methods in medical research, 15(3), 213-234.
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior research methods, instruments, & computers, 36(4), 717-731.
Priest, G. R., Witter, R. C., Zhang, Y. J., Wang, K., Goldfinger, C., Stimely, L. L., ... & Smith, R. L. (2013). Tsunami inundation scenarios for Oregon. Oregon Department of Geology Mineral Industries Open-File Report O-13, 19.
Priest, G. R., Stimely, L. L., Wood, N. J., Madin, I. P., & Watzig, R. J. (2016). Beat-the-wave evacuation mapping for tsunami hazards in Seaside, Oregon, USA. Natural Hazards, 80(2), 1031-1056.
Quarantelli, E. L. (1980). Evacuation Behavior and Problems: Findings and Implications from the Research Literature. OHIO STATE UNIV COLUMBUS DISASTER RESEARCH CENTER.
Raskin, J., & Wang, Y. (2016). Fifty-Year Resilience Strategies for Coastal Communities at Risk for Tsunamis. Natural Hazards Review, 18(1), B4016003.
Riad, J. K., Norris, F. H., & Ruback, R. B. (1999). Predicting evacuation in two major disasters: risk perception, social influence, and access to resources. Journal of Applied Social Psychology, 29(5), 918–934.
Rippetoe, P. A., & Rogers, R. W. (1987). Effects of components of protection-motivation theory on adaptive and maladaptive coping with a health threat. Journal of personality and social psychology, 52(3), 596.
Rivers, J. P. W. (1982). Women and children last: an essay on sex discrimination in disasters. Disasters, 6(4), 256-267.
Rivers, L., Arvai, J., & Slovic, P. (2010). Beyond a simple case of black and white: Searching for the
white male effect in the African‐American community. Risk analysis, 30(1), 65-77.
126
Rofi, A., Doocy, S., & Robinson, C. (2006). Tsunami mortality and displacement in Aceh province, Indonesia. Disasters, 30(3), 340-350.
Rogers, G. O. (1982). Social status and perceived risk: Some social processes and risk perception. In 10th World Congress of Sociology.
Ruiter, R. A., Abraham, C., & Kok, G. (2001). Scary warnings and rational precautions: A review of the psychology of fear appeals. Psychology and Health, 16(6), 613-630.
Russell, L. A., Goltz, J. D., & Bourque, L. B. (1995). Preparedness and hazard mitigation actions before and after two earthquakes. Environment and behavior, 27(6), 744-770.
Said, A., Ahmadun, F. L. R., Rodzi Mahmud, A., & Abas, F. (2011). Community preparedness for tsunami disaster: a case study. Disaster Prevention and Management: An International Journal, 20(3), 266-280.
Schultz, M. T., & Smith, E. R. (2016). Assessing the Resilience of Coastal Systems: A Probabilistic Approach. Journal of Coastal Research, 32(5), 1032-1050.
Shapira, S., Aharonson-Daniel, L., & Bar-Dayan, Y. (2018). Anticipated behavioral response patterns to an earthquake: The role of personal and household characteristics, risk perception, previous experience and preparedness. International Journal of Disaster Risk Reduction.
Shibayama, T., Esteban, M., Nistor, I., Takagi, H., Thao, N. D., Matsumaru, R., ... & Ohira, K. (2013). Classification of tsunami and evacuation areas. Natural hazards, 67(2), 365-386.
Siegrist, M., & Gutscher, H. (2008). Natural hazards and motivation for mitigation behavior: People cannot predict the affect evoked by a severe flood. Risk Analysis, 28(3), 771-778.
Simmons, K. M., & Sutter, D. (2009). False alarms, tornado warnings, and tornado casualties. Weather, Climate, and Society, 1(1), 38-53.
Sjöberg, L., Moen, B. E., & Rundmo, T. (2004). Explaining risk perception. An evaluation of the psychometric paradigm in risk perception research, 10(2), 665-612.
Sleeter, B. M., Wood, N. J., Soulard, C. E., & Wilson, T. S. (2017). Projecting community changes in hazard exposure to support long-term risk reduction: A case study of tsunami hazards in the US Pacific Northwest. International Journal of Disaster Risk Reduction, 22, 10-22.
Slovic, P. (1992). Perception of risk: Reflections on the psychometric paradigm. Retrieved from https://philpapers.org/rec/SLOPOR
Slovic, P., Fischhoff, B., & Lichtenstein, S. (1980). Facts and fears: Understanding perceived risk. Societal Risk Assessment: How Safe Is Safe Enough, 4, 181–214.
Slovic, P., & Weber, E. U. (2002). Perception of risk posed by extreme events. Retrieved from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2293086
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. Sociological methodology, 13, 290-312.
Spence et al. (1989) Earthquakes and Volcanoes USGS https://earthquake.usgs.gov/learn/topics/measure.php
Spittal, M. J., McClure, J., Siegert, R. J., & Walkey, F. H. (2008). Predictors of two types of earthquake preparation: survival activities and mitigation activities. Environment and Behavior, 40(6), 798-817.
Steel, B. S., Pierce, J. C., Warner, R. L., & Lovrich, N. P. (2015). Environmental value considerations in public attitudes about alternative energy development in Oregon and Washington. Environmental management, 55(3), 634-645.
Stone, C. A., & Sobel, M. E. (1990). The robustness of estimates of total indirect effects in covariance structure models estimated by maximum. Psychometrika, 55(2), 337-352.
127
SUN, Y., YAMORI, K., TANISAWA, R., & KONDO, S. (2013). Consciousness of disaster risk and tsunami evacuation: a questionnaire survey in Okitsu, Kochi Prefecture. Journal of Natural Disaster Science, 34(2), 127-141.
Sun, Y., Nakai, F., Yamori, K., & Hatayama, M. (2017). Tsunami evacuation behavior of coastal residents in Kochi Prefecture during the 2014 Iyonada Earthquake. Natural hazards, 85(1), 283-299.
Suppasri, A., Shuto, N., Imamura, F., Koshimura, S., Mas, E., & Yalciner, A. C. (2013). Lessons learned from the 2011 Great East Japan tsunami: Performance of tsunami countermeasures, coastal buildings, and tsunami evacuation in Japan. Pure and Applied Geophysics, 170(6-8), 993-1018.
Takabatake, T., Shibayama, T., Esteban, M., Ishii, H., & Hamano, G. (2017). Simulated tsunami evacuation behavior of local residents and visitors in Kamakura, Japan. International Journal of Disaster Risk Reduction, 23, 1-14.
Takabatake, T., Shibayama, T., Esteban, M., & Ishii, H. (2018) Advanced casualty estimation based on tsunami evacuation intended behavior: case study at Yuigahama Beach, Kamakura, Japan. Natural Hazards, 1-26.
Tekeli‐Yeşil, S., Dedeoğlu, N., Braun‐Fahrlaender, C., & Tanner, M. (2010). Factors motivating individuals to take precautionary action for an expected earthquake in Istanbul. Risk analysis, 30(8), 1181-1195.
Terpstra, T., & Lindell, M. K. (2013). Citizens’ perceptions of flood hazard adjustments: an application of the protective action decision model. Environment and Behavior, 45(8), 993-1018.
Tierney, K. J. (1994). Sociology's unique contributions to the study of risk. University of Delaware Disaster Research Center.
Tobin, G. A., Whiteford, L. M., Jones, E. C., Murphy, A. D., Garren, S. J., & Padros, C. V. (2011). The role of individual well-being in risk perception and evacuation for chronic vs. acute natural hazards in Mexico. Applied Geography, 31(2), 700–711.
Toledo, T., Marom, I., Grimberg, E., & Bekhor, S. (2018). Analysis of Evacuation Behavior in a Wildfire Event. International Journal of Disaster Risk Reduction.
Turner, B. A. (1979). The social aetiology of disasters. Disasters, 3(1), 53–59. Ulleberg, P., & Rundmo, T. (2003). Personality, attitudes and risk perception as predictors of risky
driving behaviour among young drivers. Safety science, 41(5), 427-443. Van der Velde, F. W., & Van der Pligt, J. (1991). AIDS-related health behavior: Coping, protection
motivation, and previous behavior. Journal of behavioral medicine, 14(5), 429-451. Vaske, J. J. (2008). Survey research and analysis: Applications in parks, recreation and human dimensions.
Venture Publishing. Vogt, W. P., & Johnson, R. B. (2011). Dictionary of statistics & methodology: a nontechnical guide for the social
sciences: a nontechnical guide for the social sciences. Sage. Wachinger, G., Renn, O., Begg, C., & Kuhlicke, C. (2013). The risk perception paradox—
implications for governance and communication of natural hazards. Risk analysis, 33(6), 1049-1065
Walshe, R. A., & Nunn, P. D. (2012). Integration of indigenous knowledge and disaster risk reduction: A case study from Baie Martelli, Pentecost Island, Vanuatu. International Journal of Disaster Risk Science, 3(4), 185-194.
128
Walton, D., & Lamb, S. (2009). An experimental investigation of post-earthquake travel behaviours: the effects of severity and initial location. International Journal of Emergency Management, 6(1), 14-32.
Wang, H., Mostafizi, A., Cramer, L. A., Cox, D., & Park, H. (2016). An agent-based model of a multimodal near-field tsunami evacuation: decision-making and life safety. Transportation Research Part C: Emerging Technologies, 64, 86-100.
Wei, H. L., & Lindell, M. K. (2017). Washington households' expected responses to lahar threat from Mt. Rainier. International Journal of Disaster Risk Reduction, 22, 77-94.
Wei, H. L., Wu, H. C., Lindell, M. K., Prater, C. S., Shiroshita, H., Johnston, D. M., & Becker, J. S. (2017). Assessment of households’ responses to the tsunami threat: A comparative study of Japan and New Zealand. International Journal of Disaster Risk Reduction.
Wenger, D. E. (1985). Mass media and disasters. Retrieved from http://udspace.udel.edu/handle/19716/474
Whitehead, J. C., Edwards, B., Van Willigen, M., Maiolo, J. R., Wilson, K., & Smith, K. T. (2000). Heading for higher ground: factors affecting real and hypothetical hurricane evacuation behavior. Global Environmental Change Part B: Environmental Hazards, 2(4), 133-142.
Wilson, S. G., & Fischetti, T. R. (2010). Coastline Population Trends in the United States 1960 to 2008 (pp. 1-28). Washington, DC: US Department of Commerce, Economics and Statistics Administration, US Census Bureau.
Wirtz, P. W., Rohrbeck, C. A., & Burns, K. M. (2017). Anxiety effects on disaster precautionary behaviors: A multi-path cognitive model. Journal of health psychology, 1359105317720277.
Wolters, E. A., Steel, B. S., Weston, D., & Brunson, M. (2017). Determinants of residential Firewise behaviors in Central Oregon. The Social Science Journal, 54(2), 168-178.
Wood, N. (2007). Variations in city exposure and sensitivity to tsunami hazards in Oregon. Geological Survey (US). Retrieved from https://pubs.er.usgs.gov/publication/sir20075283
Wood, N. J., Burton, C. G., & Cutter, S. L. (2010). Community variations in social vulnerability to Cascadia-related tsunamis in the US Pacific Northwest. Natural Hazards, 52(2), 369-389.
Wood, N. J., Jones, J., Spielman, S., & Schmidtlein, M. C. (2015). Community clusters of tsunami vulnerability in the US Pacific Northwest. Proceedings of the National Academy of Sciences, 112(17), 5354-5359.
Wood, N. J., & Schmidtlein, M. C. (2013). Community variations in population exposure to near-field tsunami hazards as a function of pedestrian travel time to safety. Natural hazards, 65(3), 1603-1628.
Xu, D., Peng, L., Liu, S., Su, C., Wang, X., & Chen, T. (2017). Influences of Sense of Place on Farming Households’ Relocation Willingness in Areas Threatened by Geological Disasters: Evidence from China. International Journal of Disaster Risk Science, 8(1), 16-32.
Yogaswara, H., & Yulianto, E. (2008). Smong, Local Knowledge of Tsunami among the Simeulue Community, Nangroe Aceh Darusallam. Jakarta Tsunami Information Centre, Jakarta.
Yun, N. Y., & Hamada, M. (2012). Evacuation behaviors in the 2011 Great East Japan earthquake. Journal of Disaster Research, 7(7), 458-467.
Yun, N. Y., & Hamada, M. (2015). Evacuation behavior and fatality rate during the 2011 Tohoku-Oki earthquake and tsunami. Earthquake Spectra, 31(3), 1237-1265.
Zanocco, C. M., Song, G., & Jones, M. D. (2017). Fracking Bad Guys: Narrative Character Affect in Public Opinion about Hydraulic Fracturing.
130
Appendix A Survey Cover Letter
Dear <<GivenName>> <<Surname>> (otherwise insert <<resident>>), As you may be aware, the Oregon coast is located in the zone of a potential earthquake. Its source is the Cascadia Subduction Zone (CSZ). An earthquake along this zone is believed to occur on average once every ~240 to ~530 years, varying by the location and the degree. The last Cascadia earthquake happened in 1700. During a Cascadia earthquake the Oregon coast can experience a magnitude of 8.0 to 9.0 earthquake followed immediately by tsunamis.
We are asking for your help in improving our understanding of perceptions of risk and potential behavior in the event of Cascadia earthquakes and tsunamis among coastal residents. The ultimate goal of our study is to enhance community preparedness for such an event. Your household is one of small number that has been randomly selected to help in this study. This project is being conducted by Oregon State University with funding from the National Science Foundation. We would appreciate if you take about 15-25 minutes to respond to the enclosed questionnaire and return it in the prepaid postage envelope provided. We ask that a respondent 18 years of age and older who has had the most recent birthday complete the survey (if unavailable, an adult with the next most recent birthday). Your participation in this study is voluntary and you may refuse to answer any question(s) for any reason. Your household location will be combined with others to understand the difference in perceptions based on distance from the ocean and location in the community. Your responses will be reported in a summary format that does not identify individuals or households. The ID number on your questionnaire is used only to ensure that we can check your household off the mailing list when the survey is completed. You have an opportunity to take the online version of the survey by following this link <<QualtricsURL>>. We recommend taking the online survey on a computer instead of a phone. Your name will not be included in the reports and your responses will be kept confidential to the extent permitted by the technology and the law. If you do not want to participate in the study, please return the uncompleted survey in the enclosed envelope. If you would like a copy of the results, include a note with an address and ‘survey results requested’. Do not put this information on the survey. If you have any questions about the survey, please contact us at (541)737-5382 or by email at [email protected]. If you have any questions about your rights as a participant in this research project, please contact the Oregon State University Institutional Review Board (IRB) office at (541)737-8008 or by email at [email protected]
Thank you for your help. We hope you enjoy completing the survey. Respectfully,
signature signature
Lori A. Cramer, Ph.D. Alexandra Buylova Associate Professor of Sociology Ph.D. Candidate School of Public Policy Public Policy Graduate School Oregon State University Oregon State University [email protected] [email protected]
131
Appendix B Survey Questionnaire
Cascadia Earthquakes and Tsunamis in Oregon: Perceptions and Behavioral Intentions of Local Residents
Fall 2017
Please return survey to:
Cascadia Perception Survey
School of Public Policy
Bexell Hall Oregon State University
Corvallis, Oregon 97331
541-737-5382
ID # ______________________
[for mailing purposes only]
132
Q1. How likely or unlikely is a M9 Cascadia earthquake and tsunami to happen in the next 50 years? (Circle one number)
1. Very unlikely 2. Unlikely 3. Neither 4. Likely 5. Very likely
Q2. How much do you trust experts, if at all, to accurately estimate the likelihood of a M9 Cascadia earthquake and tsunami happening in the next 50 years? (Circle one number)
1. Don’t trust at all 2. Trust a little 3. Somewhat trust 4. Trust 5. Trust a lot
Q3. Have you ever experienced a natural disaster such as flood, earthquake, wildfire, or other types of natural disasters? (Circle one number)
1. No (Pass to Q4.) 2. Not sure (Pass to Q4.) 3. Yes (Answer Q3.1. and Q3.2.)
Q3.1. Which of these natural disasters have you experienced? (Check all that apply)
Hurricane
Wildfire
Landslide
Flood
Tornado
Earthquake
Tsunami
Dam Failure
Other (specify)
Q3.2. Was your life or lives of your family members in danger during any of those events? (Circle one number)
1. No 2. Yes
SECTION 1
This section will ask you about your knowledge, perceptions, and preparation for Cascadia
earthquakes and tsunamis. As you are answering the questions, we ask you to consider the
case of 9.0 magnitude (M9) Cascadia earthquake that triggers a tsunami wave affecting
the entire Oregon coast. Please remember that ‘you’ in this questionnaire refers to you
133
Q4. Have you ever experienced a natural disaster where you needed to evacuate? (Circle one number)
1. No. 2. Yes. What kind of disaster was it and where? (Briefly describe)
Q5. How confident are you about each of the following statements? (Circle one number for each statement)
Statement Not at all confident
A little confident
Somewhat confident
Confident Very
confident
1. I know the difference between local and distant tsunami events
1 2 3 4 5
2. I understand natural warning signs of tsunamis
1 2 3 4 5
3. I understand emergency warning messages of tsunamis
1 2 3 4 5
4. I know where to get the information about preparation for tsunamis
1 2 3 4 5
5. I know government recommendations on how to prepare for tsunamis
1 2 3 4 5
6. I know what to do when there is a warning about a tsunami
1 2 3 4 5
7. I know tsunami evacuation routes from my home
1 2 3 4 5
8. I know a tsunami safe place close to my home
1 2 3 4 5
134
Q6. If a M9 Cascadia earthquake and tsunami were to happen tomorrow, how likely or unlikely are they to result in the following? (Circle one number for each statement)
Statement Very
unlikely Unlikely Neither Likely
Very likely
1. Lead to injuries for you or your family
1 2 3 4 5
2. Create a life-threatening situation for you or your family
1 2 3 4 5
3. Severely damage or destroy your home
1 2 3 4 5
4. Create a severe financial burden for you or your family
1 2 3 4 5
5. Destroy or severely damage roads, homes, etc. in your town
1 2 3 4 5
6. Create a life-threatening situation for people in your town
1 2 3 4 5
Q7. Given your personal skills and resources, how confident are you that you could effectively protect yourself and reduce the chance of severe injuries in the event of a M9 Cascadia earthquake and tsunami? (Circle one number)
1. Not at all confident 2. A little confident 3. Somewhat confident 4. Confident 5. Very confident
Q8. Do you agree or disagree that people could increase their chances of surviving a M9 Cascadia earthquake and tsunami by making various emergency plans or preparatory actions? (Circle one number)
1. Strongly disagree 2. Disagree 3. Neither 4. Agree 5. Strongly agree
135
Q9. Check all the measures that you have taken in the past 2 years, if any, that can assist you in the event of Cascadia earthquakes and tsunamis. (Check all that apply)
Attended a meeting or received written information on how to be better prepared for Cascadia earthquakes and tsunamis.
Prepared an ‘emergency supply kit’ (stored extra food, water, batteries, etc.).
Discussed the topic of preparedness for Cascadia earthquakes and tsunamis with others in the community.
Developed an ‘emergency plan’ with family or friends in order to decide what everyone would do in the event of Cascadia earthquakes and tsunamis.
Developed a communication plan (e.g. a plan that establishes how to get in touch with each other if meeting at home is not possible and phones do not work).
Attended First Aid or Cardio-Pulmonary Resuscitation (CPR) training.
Identified an emergency contact person outside of the Northwest.
Other (Specify) Q10. If you were at home and you feel the ground shaking, would you try to evacuate to a safe or high ground to escape a tsunami? (Circle one number)
1. No, (Please explain) 2. Not sure, (Please explain) 3. Yes
Q11. Is your home located in Cascadia tsunamis inundation area? (Circle one number)
1. No 2. Not sure 3. Yes
SECTION 2
This section will ask you about your potential behavior and factors related to how you
would respond in the event of a M9 or greater Cascadia earthquake and tsunami.
136
Q12. How likely or unlikely are you to adopt each of the following behaviors if you were at your home in the event of a M9 Cascadia earthquake and tsunami? (Circle one number for each statement)
Behavior Very
unlikely Unlikely Neither Likely
Very likely
1. Continue normal routine behavior 1 2 3 4 5
2. Follow earthquake emergency guidelines (e.g. shelter under a table if inside)
1 2 3 4 5
3. Evacuate to higher ground immediately after ground shaking
1 2 3 4 5
4. Wait to hear from an authority figure/sirens/warning for tsunami confirmation
1 2 3 4 5
5. Check social media, radio, TV for additional information about a tsunami
1 2 3 4 5
6. Contact loved ones (phone or text family or friends to see what they are doing and if they are safe)
1 2 3 4 5
7. Collect important documents and other items
1 2 3 4 5
8. Travel to gather family or friends (e.g. kids at school, parents or grandparents at their house)
1 2 3 4 5
9. Go towards the ocean to see if a tsunami is coming
1 2 3 4 5
10. Talk with neighbors 1 2 3 4 5
11. Take the lead to evacuate and encourage people to leave with me
1 2 3 4 5
12. Wait for my family at home before evacuating
1 2 3 4 5
13. Wait to see if my friends or neighbors are evacuating, then I would follow them
1 2 3 4 5
14. Try to help and rescue people 1 2 3 4 5
137
15. Panic 1 2 3 4 5
Q13. If you were at your home, how confident are you that you could escape to a tsunami safe zone in the event of a M9 Cascadia earthquake and tsunami? (Circle one number)
1. Not at all confident 2. A little confident 3. Somewhat confident 4. Confident 5. Very confident 6. Not relevant (my home is not in an inundation zone)
Q14. If you were at home, what is the most likely mode you would try to use to reach the safe zone? (Circle one number)
1. Drive 2. By bicycle 3. By foot 4. Other (specify) 5. Not relevant (my home is not in an inundation zone)
Q15. If you were to use the evacuation mode of your choice, how long would it take you to
reach a tsunami safe zone from your home? (Circle one number)
1. Less than 15 minutes 2. Between 15 and 30 minutes 3. Between 30 and 45 minutes 4. Between 45 minutes and 1 hour 5. Between 1 and 1.5 hours 6. Between 1.5 and 2 hours 7. Not relevant (my home is not in an inundation zone)
Q16. The next two questions ask you about your personal networks. Exploring personal contacts will help us better understand people’s potential movements on the ground in a case of emergency and better plan for a real event.
Q16.1. Imagine that you are at home and you feel an earthquake and/or you learn of a public warning of a tsunami. How many people in your community would you try to immediately contact to check on their safety, if any at all? (Write a number)
Q16.2. How many of these contacts would you try to reach in person to assist them in their mobility or evacuation? (Write a number)
Q17. Approximately, how many times in the past 5 years have you either participated in tsunami evacuation drills in your community or walked your neighborhood evacuation route? (Circle one number)
1. Never 2. 1-2 times
138
3. 3-5 times 4. More than 5 times
Q18. How effective are tsunami evacuation drills (e.g. “walking the evacuation route”) in preparing for evacuation in a real event? (Circle one number)
1. Not at all effective 2. A little effective 3. Somewhat effective 4. Effective 5. Very effective
Q19. What is your assessment of the general level of community preparedness for Cascadia earthquakes and tsunamis in your local area? (Circle one answer)
1. Lacking 2. Poor 3. Adequate 4. Good 5. Excellent 6. I don’t know
Q20. How effective have each of the following individuals or groups been in earthquakes and tsunamis preparedness in your community? (Circle one number for each statement; check “Not familiar” if you haven’t heard of a group or an individual participating in preparedness activities)
Organization Not
familiar Not at all effective
A little effective
Somewhat effective
Effective Very
effective
1. American Red Cross
1 2 3 4 5
2. University or research institutions
1 2 3 4 5
3. Local government 1 2 3 4 5
4. County government
1 2 3 4 5
5. Tribal authorities or governments
1 2 3 4 5
6. State of Oregon Department of Emergency Management
1 2 3 4 5
7. Oregon Parks and Recreation Department
1 2 3 4 5
139
8. Local police department
1 2 3 4 5
9. Local fire department
1 2 3 4 5
10. Governor of Oregon
1 2 3 4 5
11. US Coast Guard 1 2 3 4 5
12. National Oceanic and Atmospheric Administration
1 2 3 4 5
13. Federal Emergency Management Agency (FEMA)
1 2 3 4 5
Q21. What is your sex? (Circle one number)
1. Female 2. Male 3. Other
Q22. What is your current age in years? (Write the response) Q23. Because of a physical, mental, or emotional condition, do you have serious difficulty: (Check one answer for each statement)
1. Concentrating, remembering, or making decisions Yes No 2. Walking or climbing stairs Yes No 3. Dressing or bathing Yes No
Q24. How many people, besides you, are living or staying at this address right now? (Write the response) Q25. What are their ages? (Provide your best estimates) Q26. What is your ownership status of this residence? (Circle one number)
1. My family or I are the primary owners 2. I am a renter/occupier (with or without payment of rent)
SECTION 3
To best evaluate our survey results we need some information about your
background. Remember that all responses will be confidential.
140
3. Other (Specify) Q27. Is the address where you received this survey your: (Circle one number)
1. Primary residence 2. Part-time residence/vacation home 3. Other (Specify)
Q28. Approximately how many months or years have you lived in this community (not necessarily at this address)? (Circle one number)
1. Less than a year 2. A year and more. How many years? (Write the response) years
Q29. Approximately how many months or years have you lived at this address? (Circle one number)
1. Less than a year 2. A year and more. How many years? (Write the response) years
Q30. In the past 12 months have you been involved in any emergency response organizations or organizations related to earthquakes and tsunamis preparedness (e.g. Red Cross, CERTs)? (Circle one number)
1. No 2. Yes. If yes, what organization(s) have you been involved with? (Write the response)
Q31. What is the highest year of formal education that you have completed? (Circle one number)
1. Less than high school diploma 2. High school diploma or GED equivalent 3. Vocational, trade, or business school 4. Some college
5. Associate's degree 6. Bachelor’s degree 7. Graduate degree or other professional degree
Q32. Which category best describes your household income before taxes in 2016? (Circle one number) 1. Less than $10,000 2. $10,000-$14,999 3. $15,000-$19,999 4. $20,000-$29,999 5. $30,000-$39,999
6. $40,000-$49,000 7. $50,000-$59,999 8. $60,000-$74,999 9. $75,000 or higher
141
If you want to add explanations for your responses and/or if you have any comments, questions, suggestions regarding this questionnaire, please write those below.
We appreciate you taking the time to fill out this survey. Please return this completed survey as soon as possible in the enclosed postal paid envelope. If you have any additional questions, please feel free to contact Dr. Lori Cramer at 541-737-5382 or [email protected]. For information regarding your rights as a participant, contact the Institutional Review Board at [email protected] or 541-737-3467.