University of Southern California Dissertations and Theses

Preparing for natural disasters: investigating the effects of gain-loss framing on individual choices

(USC Thesis Other)

Preparing for natural disasters: investigating the effects of gain-loss framing on individual choices

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Preparing for Natural Disasters: Investigating the Effects of Gain-Loss Framing on Individual
Choices

by

Mengtian Zhao

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
PSYCHOLOGY
August 2020

Copyright 2020 Mengtian Zhao
ii
Acknowledgements
First and foremost, I would like to thank my advisor Dr. Richard John. Throughout five
years of graduate school, Dr. John has always provided unconditional support and offered
numerous opportunities for me to explore my research interests. He shaped my interested in
behavioral decision making and uncertainty, always encourages me to work on different arrays
of projects, and strives to provide support for not just his own students, but for anyone who seeks
guidance. Most importantly, Dr. John has always been supportive and understanding when I was
experiencing the ups and downs of graduate school and in my personal life.
I would also like to extend my gratitude toward my committee members Prof. Detlof
von Winterfeldt, Prof. Antoine Bechara, Prof. John Monterosso, and Prof. Mark Lai, who took
valuable time out of their busy schedules to provide me insightful comments and guidance
throughout my dissertation. I would also like to thank my former and current lab-mates Dr.
Zhiqin Chen, Dr. Jinshu Cui, Dr. Kenneth D. Nguyen, Sarah Kusumastui, Katie Sippel and Aili
Qiao for helping me with my research, many of them continued to reach out to me even after
they had graduated and offered me valuable advice and feedbacks, which I deeply appreciate.
This acknowledgement ont be complete ithot mentioning m famil. I oldnt be
ho I am toda ithot m parents nconditional loe and spport . My father, who has a
doctorate degree in economics has always been understanding of my graduate school life since
he experienced similar struggles and challenges when he was a student. My mother, hos the
best cook ever and the nicest human being I know, has always encouraged me to take the path
less travelled. Last year she was diagnosed with cancer and my family struggled a lot, yet she is
still just as upbeat and strong-minded as before, and I cannot begin to express how proud I am of
her. My cat King has accompanied me through many late nights of working on courses and
iii
research since college, and I also thank him for the continuous emotional support. I hope as I
graduate from USC and take on a new path in life, my family will continue to enjoy our times
together for many years to come.

iv
TABLE OF CONTENTS
Acknowledgements........................................................................................................................ii
List of Tables .................................................................................................................................v
List of Figures ...............................................................................................................................vi
Abstract..........................................................................................................................................vii
Chapter 1: Overview of Natural Disaster Risk Mitigation..............................................................1
Chapter 2: Gain-Loss Framing Effects in the Context of Natural Disasters...................................6
Chapter 3: Framework for Designing Decision Vignettes.............................................................18
Chapter 4: Research Questions......................................................................................................24
Chapter 5: Effects of Framing on Individual Choices...................................................................30
Method..................................................................................................................32
Results..................................................................................................................41
Hurricane sample results......................................................................................43
Flood sample results.............................................................................................44
Earthquake sample results.....................................................................................45
Discussion ............................................................................................................47
Chapter 6: Effects of Framing and Psychological Factors on Individual Choices .......................57
Method..................................................................................................................58
Results..................................................................................................................66
Hurricane sample results......................................................................................69
Flood sample results.............................................................................................71
Earthquake sample results.....................................................................................74
Discussion ............................................................................................................76
Chapter 7: General Discussion ......................................................................................................83
References......................................................................................................................................91
Appendices...................................................................................................................................109
Appendix A: Decision Vignettes for Experiments 1...........................................109
Appendix B: Decision Vignettes for Experiment 2 ............................................118
Appendix C: Cognitive Reflection Task ............................................................134

v
List of Tables
Table 1. Comparison of Three Disasters........................................................................................28
Table 2. List of States for Each Disaster Sample.............................................................................31
Table 3. Design Layout for Experiment 1.......................................................................................33
Table 4. Loss Frame Vignette for Hurricane Condition.................................................................36
Table 5. Gain Frame Vignette for Hurricane Condition................................................................37
Table 6. Demographics Table for Experiment 1............................................................................39
Table 7. Logistic Regression Model Summary for Hurricane Sample, Experiment 1..................44
Table 8. Logistic Regression Model Summary for Flood Sample, Experiment 1.........................45
Table 9. Logistic Regression Model Summary for Earthquake Sample, Experiment 1................47
Table 10. Summary of Results for Experiment 1..........................................................................50
Table 11. Design Layout for Experiment 2...................................................................................59
Table 12. Demographics Table for Experiment 2.........................................................................65
Table 13. Logistic Regression Model for Hurricane EV Sample, Experiment 2..........................70
Table 14. Logistic Regression Model for Hurricane Probability Sample, Experiment 2..............71
Table 15. Logistic Regression Model for Flood EV Sample, Experiment 2.................................72
Table 16. Logistic Regression Model for Flood Probability Sample, Experiment 2.....................73
Table 17. Logistic Regression Model for Earthquake EV Sample, Experiment 2........................74
Table 18. Logistic Regression Model for Earthquake Probability Sample, Experiment 2............75
Table 19. Summary of Results for Experiment 2..........................................................................79
Table 20. Summary of Results for Experiment 1 and 2.................................................................85

vi
List of Figures
Figure 1. Value function from Prospect Theory...............................................................................8
Figure 2. Decision Trees for Hurricane Scenario, Experiment 1.....................................................34
Figure 3. Risk Preference Distribution for Experiment 1..............................................................43
Figure 4. Decision Trees for Hurricane Scenario, Experiment 2...................................................61
Figure 5. Risk Preference Distribution for Experiment 2..............................................................68

vii
Abstract
Promoting natural disaster preparation plays a critical role in natural disaster research. Adoption
of protective actions could effectively reduce the risk of injury and damage to property, however,
despite considerable expenditures on public education programs, numerous studies reported that
civilians still are under-prepared for natural disasters (Ballantyne, Paton, Johnston, Kozuch &
Daly, 2000; Lindell & Whitney, 2000; Paton, 2003; Tanaka, 2005; Shaw, Kurita, Nakamura,
Kodama & Colombage, 2006). Therefore, a more effective approach for encouraging protective
actions and a better understanding of the process underlying protective action decision making
would have tremendous theoretical and practical benefits. The effectiveness of using gain-loss
framing effects to nudge desired decisions has been extensively demonstrated across many
different domains, yet the application of gain-loss framing effects in natural disaster preparation
so far has only concentrated on policy-level decision making. Experiment 1 (N= 1,840) set out to
test gain-loss framing effects for homeowner decisions and compare whether gain-loss framing
effects impact risk mitigation and insurance purchase to the same magnitude. Experiment 2 (N=
1,224) explored whether gain-loss framing effects were moderated by two factors of the gamble,
perceived risk and disaster probability. Results indicated that consistent with Prospect Theory,
gain-frame messages were associated with higher risk averse tendencies for floods and
hurricanes, but not for earthquakes. High probability of incurring damage from the upcoming
disaster was found to be consistently associated with stronger risk-averse tendencies across
earthquakes, floods and hurricanes. This research noted diverse patterns of associations for
different natural disasters, providing unique contributions and implications to natural disaster
risk mitigation literature.
1

Chapter 1: Overview of Natural Disaster Risk Mitigation
The ik aociaed ih enionmenal haad depend no onl on phical
conditions and events but also on human actions...The seriousness of the consequences of any
disaster will depend also on how many people choose, or feel they have no choice but, to live and
ok in aea a highe ik.
- International Council for Science, 2008
In the United States, natural and climate-related disasters led to 138 fatalities in 2016
alone, and caused over $1 trillion in damage costs since 1980 (Smith, 2017). Even one single
severe weather event could result in massive destruction; for example, the 2011 Joplin tornado
caused 158 direct fatalities, and approximately $3 billion in economic losses (Paul, Emerson,
Brock & Csiki, 2013; Hoekstra, Klockow, Riley, Brotzge, Brooks & Erickson, 2011).
Voluminous research has been conducted on the topic of natural hazards preparation. Broadly,
three categories of preparation methods are used by civic emergency organizations: warning
messages, evacuations and adoption of protective measures. Warning messages and evacuations
can help reduce casualties, financial loss and injuries resulting from a natural disaster (Golden &
Adams, 2000), and extensive research has been conducted to evaluate the characteristics of
warning messages and evacuation decisions. Detailed discussion of warning messages and
evacuations are beyond the scope of this project; for a thorough review of warning messages, see
Tierney, Lindell, & Perry (2001), and for a detailed summary of evacuation research, see
Sorenson, Vogt and Mileti (1987). Another important component of risk management is to
encourage residents susceptible to natural disasters to adopt protective measures (such as storing
food and water or household retrofitting). Therefore, understanding how individuals make
2

decisions when facing natural disasters can help emergency planners better allocate resources
and aid in the development of more effective communication strategies.
It is often assumed that providing civilians with more detailed information about hazards
and mitigation alternatives would encourage protective action and reduce disaster-related
damages (Lindell & Whitney, 2000; Smith, 2013); however, previous literature has demonstrated
that this assumption is ill-founded in the context of natural disaster preparation (Mileti &
Fitzpatrick, 1992; Mullis, Duval and Lippa, 1990; Mullis & Duval, 1995). Paton, Smith and
Johnston (2000) surveyed residents in New Zealand that are susceptible to volcanic hazards, and
found that residents demonstrate poor knowledge of risk mitigation behaviors related to volcano
eruptions, even after multiple local campaigns about volcano hazards had been conducted.
Furthermore, the authors reported that knowledge about mitigation behavior did not correlate
with the adoption of protective actions. Similarly, many studies also reported that adoption of
protective actions remained low, despite considerable efforts on public natural disaster education
(Ballantyne, Paton, Johnston, Kozuch & Daly, 2000; McClure, Walkey & Allen, 1999; Lindell &
Whitney, 2000). Some scholars have proposed that the lack of success might be accounted for by
the attributes of recommended mitigation behaviors (Doyle, McClelland, Schulze, Elliott &
Russell, 1991; Edwards, 1993; Russell, Goltz & Bourque, 1995). For example, McIvor and Paton
(2007) surveyed homeowners in New Zealand and reported that residents are more likely to take
protective action if they believe that preparing for earthquakes would improve living conditions
and property values, reduce damage to homes, and disruption to daily life would be kept
minimal. Others suggested that cognitive bias also plays an important role in homeoners
decision to mitigate natural disaster related risks. For example, projection bias refers to a
tendency for the decision maker to anchor beliefs about her feelings in the future based on her
3

feelings at the moment. In the context of natural disaster mitigation, since mitigation decisions
are typically made in advance before a disaster occurs, the decision maker might underestimate
the likelihood of encountering the disaster in the future and the potential trauma the disaster can
bring, which makes the decision maker reluctant to invest in risk mitigation methods (Myer,
2006; Kunreuther, Meyer & Michel-Kerjan, 2013).
Cognitive Systems Underlying Decision Making Processes
In recent years, cognitive scientists have proposed two distinctive cognitive systems
underlying decision-making processes (Stanovich & West, 2000; Evans, 2003; Kahneman &
Egan, 2011). System 1 thinking deals with behaviors that are more instinctive and automated,
often times the thought process of System 1 is unconscious, and only the final product of System
1 thinking is reflected in behavior. For example, driving to work every day does not require
deliberation on every single step along the way; cognitive systems can quickly and automatically
retrieve previous experience to guide completion of the task (Stanovich & West, 2008).
Conversely, System 2 thinking governs thoughts that are more abstract and require more
deliberation (Baddeley, 2000). For example, if an unexpected traffic accident occurrs, System 2
thinking will step in and consider alternative plans will this delay my arrival time? should I
take a detour to avoid traffic? Psychologists who support the dual-process thinking systems
argue that System 2 thinking provides an evolutionary advantage as it can adjust to unexpected
or novel information from the environment (Koehler, 1996).
The intertwined System 1 and System 2 thinking can efficiently help navigate daily life;
however, the unique nature of natural hazards poses a unique challenge for the cognitive system.
Compared to typical day-to-day decisions, for most people natural disasters are infrequent and
unfamiliar, therefore more difficult to draw upon past experiences, thereby impeding System 1
4

thinking. What complicates the decision even more is that in the realm of natural disasters,
relying on previous experiences may lead to highly suboptimal decisions, resulting in
catastrophic consequences. In the case of Hurricane Katrina, residents reported that one of the
top reasons for not evacuating was previous experience of surviving less severe hurricanes
unharmed without evacuation (Elder, Xirasagar, Miller, Bowen, Glover & Piper, 2007).
Furthermore, decision-making related to natural hazards is complex in that engaging in
mitigation activities may impact many aspects of daily life and may involve a substantial
uncertainty. As an example, consider a family that just purchased a home in Southern California.
The new home is located at a seismic hazard zone, and the family needs to decide whether to
invest $5,000 in retrofitting their new home or not. Southern California has not incurred a severe,
large scale earthquake since the 1994 Northridge Earthquake. If the goal of this family is to
minimize future risk, retrofitting would be the ideal, utility-maximizing solution. However, the
decision becomes much more complicated when considering associated realistic uncertainties,
such as:
1. It is unclear when and where will the next big earthquake strike,
2. They may be spending money in preparing for nothing,
3. The cost of retrofitting could be invested in other ventures that might improve the
overall welfare of the family,
4. Retrofitting is only effective for a certain time period,
5. What if they invested in a project that can protect their property for 10 years and
the disaster occurred at year 11?
In decision theory, it is often assumed that a rational person would choose the optimal
option to maximize her expected interests when all the probabilities and consequences are known
5

for each available alternative (Tversky & Kahneman, 1986; Stanovich & West, 2000). However,
this assumption is rarely met in real life, and may be even less likely to hold when making
decisions related to natural hazards. Past behavioral research suggests that when facing such
difficult, high-stake decisions without sufficient information, people often resort to relying on
cognitive heuristics as shortcuts. Cognitive heuristics, as defined by Kahneman and Frederick
(2002), refer to judging a target by attribute(s) that come more readily to mind, while ignoring
other information that is more difficult to retrieve. Since heuristics require less cognitive effort,
they are quite challenging to counter (Gigerenzer & Gaissmaier, 2011). Moreover, heuristics
possess ecological validity in some cases and can be used to aid decision making. In their
pioneering book Nudge, Thaler and Snstein (2008) proposed the concept of choice
architectre, hich refers to the act of o rganizing the context in which the decision is made so
that the optimal option for the decision maker appears more appealing, thereby helping decision
makers choose better options. Using choice architecture to promote better decisions has been
studied extensively in the health domain (Marteau, Ogilvie, Roland, Suhrcke & Kelly, 2011;
Handel, 2013); however, to my knowledge, nudging individual decisions of natural disaster
mitigation using gain-loss framing has not been investigated empirically. The current project will
concentrate on using gain-loss framing effects as a way of nudging people to adopt protective
actions. The following chapter will review previous research on gain-loss framing effects, and
application of gain-loss framing effects in natural hazard preparation.

6

Chapter 2: Gain-Loss Framing Effects in the Context of Natural Disasters
Gain-Loss Framing: Definition and Origin
Gain-Loss framing effects first originated from Kahneman and Tersks concept of a
reference point in Prospect Theory (Tversky and Kahneman, 1981, 1986). A decision frame is
defined as the conception of acts, otcomes, and contingencies associated ith the decision
makers choice. Prospec t Theory provides an account for understanding decision making
processes involving risks, and postulates that for a decision under uncertainty, when potential
losses or negative consequences of a decision are emphasized (defined as a loss frame), people
tend to be risk-seeking, whereas when the potential benefits of positive consequences of a
decision are emphasized (defined as a gain frame), people tend to be risk averse (Kahneman &
Tversky, 1979).
Tversky and Kahneman (1981) tested this hypothesis with a hypothetical Asian disease
scenario, in which participants are told that a rare Asian disease is about to strike and kill 600
people in the US. Two programs have been proposed to combat the disease. If the first program
is chosen, 200 people will be saved for sure; if the second program is chosen, there is a 1/3
probability 600 people will be saved and 2/3 probability that no one will be saved. The same
problem was then described in different wording. If the third program is chosen, 400 will die for
sure; however, if the last program is chosen, theres a 1/3 probabilit that no one ill die and 2/3
probability that 600 people will die. Participants were asked to choose between two treatments.
In all four descriptions, the expected number of people who die is 400; therefore,
participants should either choose the sure thing option (200 people will be saved for sure or 400
will die for sure) in both frames or the gamble (where probabilities are involved) in both frames
if gain-loss framing effects dont impact decision making. If participants are risk averse, they
7

should pick the sure thing option in both frames; if participants are risk seeking, they should pick
the gamble in both frames. However, results showed that participants were more risk-averse
when outcomes are framed as lives saved, and more risk-seeking when outcomes are framed as
lives lost, providing preliminary empirical support for a gain-loss framing effect.
We can also evaluate the Asian Disease Problem by visiting the value function for
Prospect Theory. In Prospect Theory, the value function of a particular choice is defined as
losses and gains from the status quo. In the Asian Disease Problem discussed above, the status
quo is the implied reference point for the loss frame. As seen on Figure 1, the value function of
Prospect Theory is S-shaped: concave in the gain domain and convex in the loss domain. This
asymmetry captures the effects of gain-loss framing on peoples risk preferences. In the case of
the Asian Disease problem, reference point O represents the status-quo, where no one is harmed
and theres no otbreak of the disease. Since the nmber of people saed is 200 across all
programs, I can think of A as the amont of people saed and A as the amont of people lost.
Point B represents the perceived value of saving 200 people when presented with a gain-framed
message, and point C represents the perceived value of saving 200 people when presented with a
loss-framed message. Since the value function for the loss domain is steeper than the value
function for the gain domain, the perceived value for saving 200 people with a gain-framed
program is lower compared to the perceived value of saving 200 people with a loss-framed
program, nudging people to be risk averse when presented with gain-framed messages, and risk
seeking when presented with loss-framed messages.

8

Figure 1. Value function from Prospect Theory

Since the introduction of the Asian Disease Problem, gain-loss framing effects have
been studied and applied extensively in various domains (Camerer, 2004; List, 2004; Grinblatt &
Han, 2005), and the effects of gain-loss framing have been reported consistently in many meta-
analyses (for gain-loss framing effects in risky choices, see Levin, Gaeth, Schreiber & Lauriola,
2002, and for gain-loss framing effects in health messages, see Gallagher & Updegraff, 2011).
Decision frames can be manipulated in two ways. First, gain-loss framing can be
manipulated by shifting the default reference point. Reference point or status-quo refers to the
current state relevant to the risky event, such as current wealth, and by describing the same
identical decision problems using different frames, reference points can be shifted to make
people more risk seeking or risk-averse (Kahneman & Tversky, 1979). The Asian Disease
scenario implemented exactly this manipulation. The format of gain-loss framing by shifting
reference points involves a binary choice between a sure-thing option (for example, save 200
people for sure) and a risky gamble (for example, 1/3 probability 600 people will be saved and
2/3 probability that no one will be saved). A second means of gain-loss framing originates from
support theory, and is achieved by manipulating the salience (either positive or negative) of
9

outcomes (Tversky & Koehler, 1994). In this scenario, gain-loss framing simply refers to the
status quo, and evaluation of the decision depends on outcome description as positive or negative
from status quo. For example, in the case of the sunscreen brochure example (Detweiler et al.,
1999), participants will be told that not wearing sunscreens is associated with risk of developing
skin cancer (negative frame), and wearing sunscreen can make the skin smooth (positive frame).
Kühberger (1998)s meta -analysis collected effect sizes from over 100 empirical experiments,
and reported that gain-loss framing by shifting reference point produced reliable, significant
differences, whereas gain-loss framing by manipulating the salience of outcomes did not yield
significant effects. The proposed experiments in this paper adopt the traditional design of
manipulating the reference point and defining both a sure-thing option and a probabilistic option
under both reference points. In the next section, I review previous research findings about gain-
loss framing effects.
Gain-Loss Framing Effects at a Policy Level
Gain-loss framing effects have been adopted in many disciplines to encourage desired
behavior (Detweiler, Bedell, Salovey, Pronin and Rothman; 1999; Rothman, Martino, Bedell,
Detweiler & Salovey, 1999; Cherubini, Rumiati, Rossi, Nigro & Calabrò, 2005; Jenner, Jones,
Fletcher, Miller & Scott, 2005; Sherman, Mann & Updegraff, 2006). For example, in health
psychology, Detweiler and colleagues (1999) distributed gain-framed and loss-framed sunscreen
brochures to over 200 beach-goers, and measured their behavioral intention before and after
reading the brochure. In the gain-framed condition, participants read information sch as protect
orself from the sn and o ill help orself sta health; hereas in the loss -framed
condition, participants read information sch as the loer the SPF o se, the more o ill be
harmed b the sn. Reslts indicated that participants assigned to the gain frame brochres ere
10

significantly more likely to repeatedly apply sunscreen at the beach, request sunscreen, and use
sunscreen with SPF 15 or higher compared to participants assigned to the loss frame. In
marketing research, Buda and Zhang (2000) presented half of the participants with positively-
framed prodct description messages (test market reslts sho that 85% of the users of this
prodct ere satisfied ith its performance) and half participants ith negatiel -framed
prodct description messages (15% cstomers ere dissatisfied ith the prodct), and fond
that participants who viewed positively-framed messages showed significantly more favorable
attitudes toward the product compared to participants who viewed negatively-framed messages.
In the realm of natural disaster preparation, scholars have started extending gain-loss
framing research to encourage natural disaster mitigation as well. McClure, White and Sibley
(2009) attempted to apply gain-loss framing effects in a natural disaster preparation context.
McClure and colleagues sampled over 200 New Zealand citizens and explored the influence of
gain-loss framing on risk perception, preparation importance, and behavioral intention to prepare
using an earthquake preparation scenario. Four conditions were included in the study: 1. positive
action, positive outcome: well prepared for a major earthquake and likely to survive the event
unharmed; 2. negative action, positive outcome: poorly prepared for a major earthquake and
likely to survive the event unharmed; 3. positive action, negative outcome: well prepared for a
major earthquake and unlikely to survive the event unharmed; 4. negative action, negative
outcome: poorly prepared for a major earthquake and unlikely to survive the event unharmed.
Participants were randomly assigned to one of the four conditions, and completed assessments of
general attitudes toward earthquake preparation and attitudes toward specific earthquake
preparation actions. Results yielded significant main effects for both action and outcome valence
on risk perception and judgments of the importance of earthquake preparation. Behavioral
11

intention to prepare were elevated with negatively framed outcomes than positive outcomes, and
positively framed messages of the action elicited higher judgements of the importance of specific
preparation actions.
Building on this research, Henrich, McClure and Crozier (2015) investigated the effects
of gain-loss framing on earthquake-related risk by presenting participants with semantically
different descriptions of the identical risk (for example, 1,600 fatalities in 500 years compared to
10% probability of 1,600 fatalities in 50 years) and asked participants to rate risk perception for
each message. Results showed that descriptions using fatality frequencies over a 50-year time
frame (for example, 10% probability of 1,600 fatalities within 50 years) received the highest
level of risk perception compared to other descriptions.
In the context of natural disaster policies, Vinnell, McClure and Milfont (2017) surveyed
over 200 local citizens from an earthquake-prone city, and designed a hypothetical scenario for
earthquake preparation legislation. Participants were told that to better prepare for future
earthquakes, the local government is working on legislation that would provide city funding for
strengthening 5,300 old buildings. The study evaluated the effects of both valence frame
(positive vs. negative) and numerical format frame (frequency vs. percentage) using a 2 by 2
design. The positive vs. negative frame was manipulated by whether the buildings meet the
required standard (positive) or do not meet the standard (negative), and frequency vs. percentage
frame was manipulated by whether 700 of the buildings have been deemed resilient or 13% of
the buildings have been deemed resilient. Participants completed measurements on attitudes
toward the legislation and attitudes toward earthquakes (including how does the risk of
earthquakes compared to other risks such as traffic accidents, how much experience does the
participant have of earthquakes, how much preparation have the participant made to prepare for a
12

large scale earthquake, etc.). Results showed that people are more likely to support earthquake-
strengthening legislation when presented in frequency framing; however, negatively framed
messages only enhanced risk perception, but not support for the proposed legislation.
Intriguingly, results showed a significant interaction between positive/negative framing and
gender; female participants are more likely to support the legislation when the message is framed
negatively, while male participants are more likely to vote for the legislation when the message
is framed positively. Similarly, Hasseldine and Hite (2003) also reported a null main effect of
positive vs. negative framing and a significant interaction effect between gender and gain-loss
framing in the context of tax-law compliance. Gain-loss framing effects appear to be moderated
by gender and ethnicity, suggesting that demographic variables should be accounted for in
studies of gain-loss framing.
In a similar vein, Marti, Stauffacher, Matthes and Wiemer (2018) surveyed over 150
homeowners in Switzerland, and tested the interaction effects of induced mood, risk perception
and gain-loss framing on homeoners attitdes toard protectie measres for earthquakes
using a 2 by 2 by design. Negative mood was induced by asking participants to read a sad article
about a young father who died from flu, and positive mood was induced by asking participants to
complete a survey; participants received positive feedback on the survey regardless of their
answers. High/low risk perception was manipulated by statements of the disaster, for example,
high risk condition incldes descriptions sch as the net strong earthqake is oerde and
there are huge gaps regarding earthquake safety in Switzerland, even though the time is ripe for
the next one hereas lo risk condition incldes descriptions sch as the net earthqake is
impending and there are some gaps regarding earthquake safety in Switzerland, which might
be explained by the long absence of major earthquakes. Gain -framed messages emphasize the
13

positie aspect of taking protectie actions, sch as homeoners and state authorities can
achieve safety benefits; and loss - framed messages emphasize the negative aspect of
ithdraing from taking protectie actions, sch as homeoners and state authorities risk the
loss of lives. Reslts reported a three-way interaction effect between mood, perceived risk and
frame type and showed that homeowners presented with induced negative mood, high risk
information and gain-framed messages exhibited the strongest supportive attitudes toward taking
protective measures for earthquakes.
Limitations of Previous Research
The studies reviewed so far offer valuable insights for implementing gain-loss framing
in natural disaster risk communications; however, there remain several limitations.
First, the context of the natural disasters investigated is quite restrictive; current
literature appears to focus solely on earthquakes, while other disasters such as floods and
tornadoes also constitute important components if natural hazard preparation (Merz, Kreibich,
Schwarze & Thieken, 2010; Bodine, Kumjian, Palmer, Heinselman & Ryzhkov, 2013). Second,
most of the research manipulated gain-loss framing only by varying semantic descriptions (for
eample, proiding descriptions of o ill be healthier if o eercise more ithot concrete
quantifications of the benefit (or loss) and the action required to gain the said benefit (or avoid
loss)), rather than manipulating the reference point for identical gambles. Investigations of gain-
loss framing require manipulation of the salient reference point and by presenting a binary choice
between a riskless (sure-thing) option and a risky gamble with both a favorable and an
unfavorable outcome.
Third, previous studies are mostly concerned with risk preferences at a policy level
rather than at an individual or household level. Although policy contexts do shed light on how
14

various psychological factors influence natural disaster preparation, choosing the best protective
action policy for others does not directly translate into taking protective actions for oneself or
family when facing natural hazards. Numerous studies have demonstrated that residents in areas
prone to natural disasters exhibit reticence to adopt protective measures and poor knowledge of
effective mitigation options, even after local emergency management agencies had taken great
efforts to provide information related to protective measures (Johnston, Bebbington Chin-Diew
Lai, Houghton & Paton, 1999; Hurnen & McClure, 1997; Rüstemli & Karanci, 1999; McIvor&
Paton, 2007).
Fourth, the paradox of using loss framed messages to elicit risk-averse behaviors is yet
to be addressed. According to Prospect Theory (Kahneman & Tversky, 1979), individuals would
be more risk-seeking when presented with loss framed messages and more risk-averse when
presented with gain-framed messages. Yet in the context of natural disaster mitigation,
emergency management organizations often use loss-framed messages (for example, turn
arond dont dron ) to encourage people to take desired actions. The motivation for using loss-
framed messages is clearly to elicit strong emotional responses, thereby inducing changes in
behavior, similar to using fear appeals to illicit attitude change in health behavior (Rogers, 1975).
The limitation of this approach has also been identified in fear appeal literature; Janis &
Feshbach (1954) pointed out that higher levels of fear might produce defensive reactions,
rendering fear appeals ineffective. Based on empirical findings of gain-loss framing, a more
effective approach for encouraging protective actions would be to use gain-framed messages that
emphasize the positive consequences of taking protective actions. Lastly, current literature
targets only the effect of gain-loss framing on mitigation behavior; however, in the context of
natural disasters, insurance is also a critical means of preparation that is readily available for
15

most natural disaster. In the following section, I discuss previous work related to mitigation
behaviors and insurance for natural hazards.
Mitigation Behavior vs. Insurance
Since mitigation behavior is also referred as protective actions, protective measures,
precautionary behaviors, etc., to ensure parsimony of discussion, I define mitigation behavior as
any behavior recommended by emergency management authorities (such as the local and federal
government, National Weather Service, FEMA, etc.) that could mitigate risks associated with
natural hazards, excluding evacuations. The decision process of evacuation is complex in nature,
as it involves many aspects of the household, and under certain extreme weather conditions,
evacuation is mandatory. Therefore, evacuation decision making is beyond the scope of the
current project (for an excellent review on evacuation decisions, see Dash and Gladwin, 2007).
In the natural disaster preparation literature, the primary focus is concerned with encouraging
and improving mitigation behavior. The previous section reviewed the effects of gain-loss
framing on mitigation behavior; this section will elaborate on the effects of gain-loss framing in
insurance purchasing.
Although insurance plays a vital role in ensuring financial protection in the aftermath of
natural disasters in developed countries such as the US (Michel-Kerjan& Kunreuther, 2011), the
uncertainty associated with natural disaster insurance makes it challenging for people to adopt.
First, the process of both the insurance company and insured individual(s) gathering information
about the opposite party is quite taxing. Second, the uncertainty surrounding the impact of a
natural hazard makes it challenging for civilians to choose among the many available insurance
plans. Third, homeowners may resort to taking mental shortcuts and relying on heuristics when
making an insurance purchase (Raschky & Weck-Hannemann, 2007; Picard, 2008). For
16

example, amnesia bias (the tendency of making decisions based only on recent experiences) and
optimism bias (the tendency to underestimate the probability of a natural hazard or consequent
financial losses) may cause residents who just experienced a natural disaster without incurring
any financial loss to not renew an existing insurance policy (Meyer, 2006; Meyer & Kunreuther,
2017). Kunreuther (1996) characterized the challenges of mitigating natural disaster losses using
insurance the term natural disaster sndrome, hich refers to the combination of residents
limited interest in mitigating risk, and the high financial costs incurred by insurers and the
federal government after a natural hazard strikes.
While I am unaware of any investigation of gain-loss framing effects on insurance
decisions in the context of natural disasters, the effects of gain-loss framing have been evaluated
in other areas. In marketing, Johnson, Hershey, Meszaros and Kunreuther (1993) examined gain-
loss framing effects in the context of auto insurance. Respondents were told that they need to
pick an auto insurance for a new $12,000 car. In the deductible condition, which emphasizes the
loss frame, participants were told that the policy has a deductible amount of $600 which will be
subtracted from the total claims. If you make any claims, the insurance company will
compensate the total amount claimed minus the $600 deductible. Participants were asked
whether they would pay $1,000 for one year of this insurance plan. In the rebate condition which
emphasizes the gain frame, participants were told that a rebate of $600 minus any claims paid
will be returned at the end of the insured year. If no claim is made, the company will give back
$600 at the end of the year. If claims were filed, $600 minus the amount claimed will paid.
Participants were asked whether they would pay $1,600 for one year of this insurance plan.
Results showed a significant difference 68% respondents were willing to pay $1,600 for the
insurance in gain-framed condition, whereas only 44% respondents were willing to pay $1,000
17

for the insurance in loss-framed condition. A replication experiment using auto insurance
scenarios revealed the same finding, respondents assigned to the gain-framed condition were
willing to pay 32% more for full coverage of the auto insurance plan compared to respondents
assigned to the loss-framed condition.
Brown, Kling, Mullainathan and Wrobel (2008) evaluated the effects of gain-loss
framing on purchasing life annuities. Despite the large welfare gains (Brown, 2007), very few
individuals made the decision to purchase late life insurance. The author described the insurance
plan in a gain frame b sing positie ords sch as inestments and earnings and b
emphasizing value of the account. The loss frame was captred b ords sch as pament and
spend, and the consmption aspect of the insrance plan is stressed , while the acconts ale
is avoided. Survey results collected from over 1,000 respondents showed a significantly strong
preference for a life annuity plan when presented with a gain frame, and a strong preference for
other alternative products when presented with a loss frame. The robust effects of gain-loss
framing have been replicated in other studies related to insurance purchasing as well (Wiener,
Gentry & Miller, 1986; Burton, 1990; Agnew, Anderson, Gerlach& Szykman, 2008). Another
goal of the current project is to compare whether gain-loss framing effects may impact mitigation
behavior and insurance purchases differently. In the next chapter, I review two important
frameworks for modelling protective action decision making processes.

18

Chapter 3: Framework for Designing Decision Vignettes
Protection Motivation Theory Framework
Protection Motivation Theory (PMT) was first introduced by Rogers (1975) to explain
how fear appeals impact health attitudes and behaviors, and has been widely applied in the
context of health behavior research. Fear appeals are defined as a means of communicating a
threat releant an indiidals ell -being, and typically include recommended measures that
could alleviate the threat. For example, in the scenario of communicating the risk of an
upcoming tornado, a fear appeal approach would inform residents that a tornado is projected to
pass the area, failure to take precautions would result in severe injury or even death, and
recommend that residents take shelter. PMT proposed that after exposure to fear appeals, two
cognitive mediating processes are involved. First, threat appraisal is activated to evaluate
severity of the threat and how likely the threat is to cause damage. Secondly, coping appraisal
follows up by assessing the affected indiidals capabilit and costs of coping ith the threat.
After both appraisal processes are completed, the decision of whether to adopt protective actions
or not will be made. It is important to note that in PMT, protective motivation is defined as the
intention to take actions. Intention to take precautionary actions does not equate to actual
changes in behavior. Previous meta-analyses in the health behavior literature found that PMT
predicts concurrent health-related behavior, and both the threat appraisal and coping appraisal
processes predict health-related behavioral intentions. (Floyd, Prentice ‐Dunn & Rogers, 2000;
Milne, Sheeran, & Orbell, 2000).
PMT is useful in disaster preparation research in two ways. First, both appraisal
processes in PMT are extremely relevant in the discussion of natural disaster preparation. The
threat appraisal process contains two subcomponents, perceived probability and perceived
19

severity of the threat (Rogers, 1975), which are also the components of measuring risk
perception in natural disaster research (Dash & Gladwin, 2007). The coping appraisal process
can be utilized to capture the perceived efficacy of taking precautionary measures (Grothmann &
Reusswig, 2006). Second, PMT provides a concise model for describing the decision-making
process of natural disaster preparation. Therefore, scholars have also extended the PMT
framework to evaluate changes in adopting precautionary measures. Mullis, Duval and Lippa
(1990) surveyed over 100 homeowners in California over a 5-week period, and investigated the
relationship between both threat appraisal, coping appraisal and earthquake preparedness
behavior. Results indicated that homeowners are more likely to take precautions when perceived
probability and perceived severity of the earthquake are high, when perceived effectiveness of
precautionary measures is high, and when perceived capacity of preparedness is high. Similarly,
Grothmann and Reusswig (2006) added a socio-psychological component to the original PMT
model and tested the extended framework by surveying residents from flood-prone areas.
Consistent ith Mllis and Lippa (1990)s findings, the athors reported that both threat
appraisal processes are critical in predicting protective responses to flood risks.
However, the PMT framework has limitations when applied to natural disasters. For
example, the coping appraisal process in PMT primarily concerns self-efficacy, which measures
the indiidals knoledge and skill reqired for taking actions (Rogers, 1975) . However, in
preparing for natural disasters, both the cost and time required to take actions are also important
(Lindell & Perry, 2012). In the next section, I introduce a more recent framework specifically
tailored to the context of natural disaster preparation.
PADM Framework
The Protective Action Decision Model (PADM) was introduced as a multi-stage
20

framework for explaining the decision-making process of initiating protective action in response
to an environmental hazard (Lindell & Perry, 1992, 2003, 2012). According to the PADM,
environmental cues (such as the sight of a funnel cloud, or hearing howling winds), social cues
(for example, information from authorities or peers about protective actions), and other socially
transmitted warning messages (such as information from media sources) constitute perceptions
of the threat. After being exposed to the natural disaster through these environmental and social
cues, civilians at risk go through a sequence of pre-decisional processes related to core
perceptions of the threat (including perception of the disaster, perception of protective action,
and perception of social stakeholders), and finally, protective action decision making.
During the pre-decisional processes, people receive, heed and comprehend the
information received from environmental and social cues. After the pre-decisional process, core
perceptions of the threat are engaged. In this stage, three important evaluations of the threat are
formed. First, perception of the natural disaster is formed. Based on the social and environmental
cues received, people perceive the risk associated with the threat and estimate expected potential
personal impact caused by the threat. This perception may be a measure of risk perception, and
can be operationalized as perceived probability and/or severity of the disaster (Mileti & Peek,
2000; Mileti & Sorensen, 1987). The second core perception formed is an evaluation of
protective actions. After receiving information about recommended protective actions from
social networks, civilians at risk evaluate attributes of the protective action, including efficacy of
the adjustment in protecting people and property, cost and knowledge required for the
adjustment, time and effort requirements, etc. (Mulilis & Duval, 1995; Davis, 1989; Russell,
Goltz & Bourque, 1995). The last core perception formed is an evaluation of the social
stakeholders involved. In the context of natural disasters, social stakeholders are defined as
21

authorities (government), emergency management agencies (such as National weather service),
watchdogs (media, environmental groups), employers and households (Pijawka & Mushkatel,
1991; Lang & Hallman, 2005). Social stakeholders are important to protective action decision
making because people perceive them to have varied levels of trustworthiness based on different
characteristics of the stakeholders, and trust in stakeholders may influence adoption of protective
action (Mulilis & Duval, 1997).
It is important to note that formation of the three perceptions do not necessarily follow a
sequential order in PADM; rather, they are formed within the same psychological stage after
exposure to the threat. After the pre-decisional processes are completed and the core perceptions
of the threat have been formed, the cognitive process of deciding whether to take protective
actions or not is initiated. The PADM postulates that intention to take protective action is a
function of attitudes toward the protective action and normative influences to conduct the action,
while evaluation of the protective action is motivated by perceived risk of the threat.
Furthermore, the final behavioral response and associated situational factors (for example,
whether the planned protective action was impeded or carried out smoothly) will serve as cues
for future decision making, thereby completing a feedback loop between previous experiences
and intention to take protective actions in the future (Lindell & Perry, 2012).
The PADM framework has been applied in a variety of natural disaster scenarios, such
as hurricanes (Huang, Lindell, Prater, Wu & Siebeneck, 2012; Lindell, Lu & Prater, 2005) and
earthquakes (Lindell & Prater, 2002). Lindell and Hwang (2008) collected data from over 300
households in Texas and tested part of the PADM in the context of flood and hurricane risks, and
found a significant association between perceived risk and adoption of protective actions for both
disasters. Interestingly, previous personal experience with the disaster also consistently predicted
22

risk perception. Terpstra and Lindell (2013) surveyed over 1,000 residents from flood-prone
areas in the Netherlands, and reported that respondents who rated the efficacy of hazard
adjustments to be high or viewed the recommended adjustments to be useful for other purposes
are significantly more likely to adopt protective actions. Furthermore, risk perception
significantly predicted intention of taking protective actions. Lindell and Perry (2000) conducted
a meta-analysis using 23 studies to examine the relationship between various attributes of the
disaster and adoption of protective actions in the context of seismic risk, and reported a
consistent positive relationship between risk perception and adoption of protective actions.
Furthermore, social influences from friends, relatives, neighbors, coworkers and mass media
have all been found to significantly correlate with adoption of protective actions, supporting the
PADMs theor that perception o f social stakeholders predict protective actions.
In previous sections, I introduced two of the most widely applied theoretical frameworks
in natural disaster preparation research. The PADM is similar to the PMT framework with the
addition of components that are unique to the context of natural disaster to better capture
decision making processes. Therefore, in this project, I only rely on the PADM framework as the
theoretical basis for designing decision vignettes. Decision vignettes that incorporate key
components of the PADM framework have been commonly used in natural disaster mitigation
research to isolate and study different aspects of the PADM, without waiting for actual disasters
that prohibit manipulation of key variables. The paradigm of using realistic, hypothetical
vignettes has been found to be effective in identifying relationships between risk perception and
behavioral intention (Huang, Lindell, Prater, Wu & Siebeneck, 2012; Lindell, Lu & Prater, 2005;
Lindell and Hwang 2008; Zhao, Rosoff & John, 2019). For example, Zhao, Rosoff and John
(2019) implemented a three-stage ignette design to test the seqential stages of eposre risk
23

perception protectie action sing a hpothetical nes report abot an pcoming seere
tornado. Results indicated that risk perception significantly predicted intention to take protective
actions, spporting the PADMs hpothesis. It is clear that realistic decision vignettes could be
used to test the effects of gain-loss framing on peoples decis ion-making when preparing for an
upcoming natural disaster. In the next chapter, I summarize the research questions and goals of
this project.

24

Chapter 4: Research Questions
Investigating Gain-Loss Framing Effects at Individual Level
Although gain-loss framing effects have been replicated across many domains
(Kühberger, 1998; Gallagher& Updegraff, 2011), research on gain-loss framing to encourage
mitigation behaviors limited. A major limitation of the current literature is that experiments have
mostly focused on making decisions related to natural disaster policy decisions affecting others
(McClure et al., 1999, 2009; Vinnell et al., 2017), with no empirical evidence that findings from
policy scenarios translate to individual-level mitigation decisions. Moreover, current research has
focused exclusively on earthquake preparation, yet floods account for nearly two-thirds of the
presidential disaster declarations from 1953 to 2010 (Michel-Kerjan & Kunreuther, 2011), and
tornadoes remain one of the most severe natural disasters in the U.S. with respect to injuries and
financial losses (Boruff, Easoz, Jones, Landry, Mitchem & Cutter, 2003; Ashley, 2007).
In this project, I aim to contribute to existing literature in two ways. First, my
experiments are designed to assess generalizability across multiple disaster scenarios, including
earthquakes, floods and hurricanes. Second, experiments are designed around individual
household mitigation decision scenarios that assess realistic behavioral intentions for natural
disaster mitigation alternatives. Moreover, considering the lack of research on gain-loss framing
of natural disaster insurance purchase decisions, both structural mitigation and insurance
purchase decisions will be examined. By constructing analogous gamble options for structural
mitigation and insurance, a direct comparison of mitigation intentions in for both con be made,
as well as exploring whether decision context (structural mitigation vs. insurance) moderates
gain-loss framing effects.

25

Effects of Perceived Risk and Disaster Probability on Gain-Loss Framing
In addition to comparing structural mitigation and insurance purchase, another key
aspect of this project is to explore whether gain-loss framing is moderated by perceived risk,
specifically, whether gain-loss framing effects are robust across manipulations of risk level
(expected value of the gamble) and probability of the identified natural disaster inflicting damage
within the specified time horizon.
Perceived risk is an integral component of gambles that affects decisions among different
alternatives. In their pioneering work, Pollatsek and Tversky (1970) proposed three important
assumptions in studying models of perceived risk. First, risk is a property of options that impact
the decision-makers choice among alternatives. Second, options of a gamble can be
meaningfully ordered in terms of their riskiness. Third, the risk associated with a certain
alternative is related to the dispersion of its outcomes. Echoing these assumptions, previous
literature has also demonstrated that decision makers are consistently able to order gambles with
respect to riskiness (Keller, Sarin and Weber, 1986; Weber & Bottom, 1990).
A rich body of literature on modeling perceived risk has provided abundant evidence
supporting the importance of EV in conceptualizing perceived risk. Pollatsek and Tversky (1970)
proposed a scale that combines EV and variance of monetary gambles to depict perceived
riskiness of the gamble. Building on their work, Coombs and Bowen (1971) reported that
perceived risk is indeed affected by EV and variance of the gamble, and tested applying various
transformations on the gamble to investigate whether perceived risk changed accordingly in an
effort to provide a more comprehensive model. Luce (1980) and Weber (1984) explored the
additive and multiplicative effect of gambles and derived a model for perceived risk based on
integrating power transformations of EV for gains and losses separately. Jia and colleagues (Jia
26

& Dyer, 1997; Jia, Dyer & Butler, 1999) aimed to unify various findings from previous research
and proposed a two-dimensional model with mean of a gamble and standard risk as parameters
and EV as the reference point to captre peoples perceptions of risk . They concluded that the
model demonstrated an appealing connection between perceived risk and risk preference of the
decision maker, and that EV can sere as a conenient and probabilisticall appealing
reference point. Taken together, previous scholars have demonstrated that EV is an integral
factor affecting choices among gambles. Since EV can be a useful tool for decision makers to
evaluate the desirability of each option when facing choices under risk (Raiffa, & Schlaifer,
1970; Gintis, 2000), I aim to empirically test whether EV serves as a moderator for individual
level gain-loss framing effect in the context of natural disasters. Due to the exploratory nature of
the analysis, I propose no specific hypothesis regarding the direction of effect.
For decisions related to natural disaster preparations, another important factor that could
affect risk propensities is the probability of the upcoming disaster. Since EV is calculated as the
probability weighted average value over all possible outcomes of a particular alternative,
intuitively, variations in probability of incurring damage from the disaster might also induce
variations in people s attitudes toward risk mitigation methods. When probability of incurring
damage from a hazard is high, the decision maker might perceive the risk associated with not
taking protective actions to be high. Conversely, when the probability of incurring damage from
the hazard is low, the decision maker might view the risk as low and have a more optimistic
outlook on the situation.
Slovic (1967) tested the effect of the magnitude of potential losses and their probabilities
of occurring with a sample of college students by asking them to rate the inherent risk associated
with each of the presented gambles, and found that the level of perceived risk was primarily
27

determined by the probability of negative outcomes (losing). Similarly, Weber, Anderson and
Birnbaum (1992) implemented a set of monetar lotteries to test respondents risk attitdes and
perception of attractiveness of the gambles, and classified respondents into two categories based
on their responses. The first category contained respondents who reported a linear relationship
between perceived risk and attractiveness of gambles (the greater the risk, the less attractive the
gambles are), and the second category consisted of respondents who did not exhibit any pattern
or relationship between perceived risk and attractiveness of gambles. Results indicated that the
two categories of participants differed in the degree to which their risk judgments were affected
by the probability of negative outcomes; specifically, respondents who reported a linear
relationship between perceived risk and attractiveness of gambles were more impacted by the
probability of negative outcomes. No significant difference was observed between the two
groups regarding the attractiveness of the gambles. Miller and Mulligan (2002) found that when
participants believe they can obtain positive outcomes, they are more likely to make risky
decisions. When participants beliee the cant do mch to inflence the otcomes of their
decisions, there more likel to aoid risk decisions .
Descriptive models of perceived risk have also suggested that people assign different
weights to positive and negative outcomes, and these weights could affect perceived riskiness of
choice alternatives (Weber & Bottom, 1989; Weber & Bottom, 1990). Shead and Hodgins (2009)
tested college stdents risk propensit ies using amount-adjusting gambles, and found that
respondents who discount the probability of gains at lower rates tend to discount probability of
losses at higher rates. These results motivated the current project to conduct an exploratory
analysis of the impact on mitigation decision making of the probability of a negative outcome
(i.e., probability of incurring damage from the identified natural disaster over the specified time
28

horizon). To the athors knoledge, there has yet to be an investigation on the effect of disaster
probability on at-risk residents risk attitudes. This project aims to provide the first attempt at
empirically testing whether the gain-loss framing effect is robust over disaster probability
manipulations for individual mitigation decisions. Since the manipulation of disaster probability
is exploratory in nature, I propose no specific hypothesis regarding the direction of effect.
Diverse Natural Disaster Contexts
Previous literature in natural disaster preparation mostly focused on a particular disaster
(Lindell & Perry, 2000; Peacock, 2003; Grothmann & Reusswig, 2006), yet natural disasters
differ on many characteristics that might influence risk attitudes and mitigation decision making.
Using data from the USGS (United States Geological Survey) and NOAA (National Oceanic &
Atmospheric Administration)s ebsite ( https://www.usgs.gov/natural-hazards/earthquake-
hazards/earthquakes; https://www.usgs.gov/mission-areas/water-resources/science/usgs-flood-
information; https://www.nhc.noaa.gov/), I compare and contrast three most common natural
disasters in the US (earthquakes, floods and hurricanes) on three dimensions. As shown in Table
1, earthquakes differ drastically from both floods and hurricanes from the perspectives of
duration, prediction method and frequency. Compared to floods and hurricanes, earthquakes are
comparatively shorter in duration, cannot be reliably predicted, and large-scale earthquakes are
relatively rare. In this dissertation, I aim to compare and contrast mitigation decision vignettes
across all three natural disasters, and to explore whether gain-loss framing effects differ across
these three natural disaster contexts.
Table 1. Comparison of Three Disasters

Duration Prediction Method Frequency
Earthquakes Seconds or
less than 1
None Magnitude 7 and above happen
less than once per year in the US;
29

minute most earthquakes are so small they
are never felt
Floods Days or
weeks
The Global Flood
Detection System
On average 2 major floods per
month during flood season
Hurricanes Hours Satellites, reconnaissance
aircraft, Ships, buoys,
radar, and other land-
based platforms
On average 2.5 major hurricanes
(Category 3 or greater) happen
each hurricane season

To summarize, the current project plans to answer three major research questions. First,
do gain-loss framing effects impact individual-level decision making in the context of natural
disaster preparation? Second, I aim to investigate whether gain-loss framing effects impact both
risk mitigation and insurance purchase, and compare responses of both risk mitigation and
insurance purchase to test whether at-risk residents exhibit preference for either protective action
method. Lastly, I explore the impact of EV and disaster probability on risk attitudes in the
context of individual mitigation decision making, and investigate whether these two factors
moderate gain-loss framing effects. The first two research questions are addressed in Experiment
1, and the last research question is explored in Experiment 2. In addition to the three main
research questions, analyses were also conducted to examine how disaster mitigation decisions
are influenced by demographic variables, objective numeracy and risk propensities. In the next
two chapters, I present the methods and results for both experiments.

30

Chapter 5: Effects of Framing on Individual Choices
Previous literature has demonstrated that gain-loss framing can be used to encourage the
occurrence of desired behaviors across various research fields (Camerer, 2004; List, 2004;
Grinblatt & Han, 2005; Levin, Gaeth, Schreiber & Lauriola, 2002), including natural disaster
preparation (McClure, White & Sibley, 2009; Henrich, McClure & Crozier, 2015; Vinnell,
McClure & Milfont, 2017). However, previous research in this realm only focused on
manipulating gain-loss framing for policy decisions affecting others. Although policy selection
decision scenarios are convenient to use and can bring useful insights, without evidence from
empirical studies, I cannot confidently infer that choices in policy-selection tasks necessarily are
consistent with individual choices of whether to take protective actions for a natural disaster or
not. The current research bridges this gap by quantitatively evaluating the effects of gain-loss
framing for individual mitigation decisions in the context of three natural disasters: earthquakes,
floods and hurricanes. Based on previous literature from Prospect Theory (Kahneman &
Tversky, 1979), I postulate that participants assigned to loss frame vignettes will tend to be risk-
seeking, whereas participants assigned to gain frame vignettes will tend to be risk averse. I make
no specific hypothesis regarding the difference between risk mitigation and insurance purchase
decisions in the context of natural disasters due to lack of previous literature.
Since the goal of this research is assess the extent to which gain-loss framing effects
nudge people toward choosing the more risk averse option (structural mitigation or insurance
purchase in the context of natural disaster preparation), my target population consists of those
who may face natural disaster risks in the future and may need to make individual household
decisions related to natural disaster preparation. To ensure responses were collected from a
representative target population, participants were screened based on their current geolocation.
31

Numerous studies have emphasized that previous experience or knowledge of the topic being
investigated is critical and can influence the degree of gain-loss framing effects (Hasseldine &
Hite, 2003; Krishnamurthy, Carter & Blair, 2001; Simmons, Nelson, Galak & Frederick, 2010);
therefore, for each of the three natural disasters, only residents currently living in risk-prone
states are eligible to participate. For hurricanes, the NOAA (National Oceanic & Atmospheric
Administration) identified the following 19 hurricane-prone states along the south-Atlantic
seaboard and Gulf of Mexico: Texas, Louisiana, Mississippi, Alabama, Florida, Georgia, South
Carolina, North Carolina, Virginia, Maryland, Delaware, New Jersey, Pennsylvania, New York,
Connecticut, Rhode Island, Massachusetts, New Hampshire and Maine (Landsea, 2018; Jarrell,
Mayfield, Rappaport & Landsea, 2001). For floods, the NOAA identified the following 14 states
most prone to flooding: Texas, Alabama, Georgia, Washington, North Carolina, Connecticut,
Maryland, Massachusetts, Louisiana, South Carolina, New York, Florida, California, and New
Jersey (Dahl, Spanger-Siegfried, Caldas & Udvardy, 2017; Garfield, 2018). For earthquakes, the
USGS (United States Geological Survey) identified that California is most vulnerable for seismic
hazards (Petersen, Moschetti, Powers, Mueller, Haller, Frankel, ..., 2014). Based on the NOAA
and USGSs reports, several states are at risk for both hurricanes and floods. To avoid duplicate
respondents for hurricanes and floods, states were partitioned so that no state was sampled for
more than one disaster. Table 2 presents the specific states implemented for screening
respondents for each of the three natural disaster samples.
Table 2. List of States for Each Disaster Sample
Natural Disaster States
Floods Connecticut, Maryland, Massachusetts, New Jersey, New York,
Washington
32

Hurricanes Alabama, Florida, Georgia, Louisiana, Mississippi, North Carolina,
South Carolina, Texas, Virginia
Earthquakes California

In addition to screening by geolocation, since the vignettes in the current research were
designed using a homeowner scenario, participants were screened by home ownership to make
sure the vignettes were realistic and relatable. Participants were asked about their
homeownership history prior to participating in the survey, and only participants who had been
homeowners in the past, participants who were currently homeowners, or participants who plan
to own homes in the next five years were eligible to participate in the study.
Method
Design Overview
In Experiment 1, two variables were manipulated: message frame (gain vs. loss) and
preparation context (risk mitigation vs. insurance purchase). Gain-loss framing is manipulated by
shifting the reference point, and risk mitigation versus insurance purchase is manipulated by
changing the scenario descriptions. Risk mitigation methods described in the decision vignettes
ere gathered from the NOAAs recommendations ( https://www.noaa.gov/topic-tags/natural-
disasters). Participants were randomly assigned into one of the four conditions demonstrated
below in Table 3, starting with a decision vignette related to natural disaster preparation,
followed by one attention check question asking which natural disaster was described in the
vignette, and concluded with assessments of demographic variables and objective numeracy.
Table 3 presents the design layout for Experiment 1.

33

Table 3. Design Layout for Experiment 1
Risk Mitigation*Gain Frame Risk Mitigation*Loss Frame
Insurance Purchase*Gain Frame Insurance Purchase*Loss Frame

Procedure
Manipulation of gain-loss framing is achieved by shifting the reference point while
keeping the expected value constant at -1000 (spending or losing $1,000) across four conditions.
For risk mitigation, the loss frame (which is more intuitive to consider) involves a choice
between spending $1,000 on installing storm shutters to prepare for the upcoming hurricane
season, or to pick the gamble in hich theres a 90% probabilit of incrring no damage from
hurricanes and a 10% probability of incurring damages worth of $10,000. The gain frame used a
slightly different description where the decision maker must put afront a $10,000 security deposit
in escrow, and the participant can either choose to use $1,000 of the $10,000 deposit to purchase
storm shutters and receive the remaining $9,000 back for sure, or to not invest in preparing for
the upcoming hurricane season, in which theres a 90% probabilit of not incrring an damage
and receive the entire deposit back, and a 10% probability of incurring damage and no deposit is
returned. In the loss frame condition, the reference point is the status quo of maintaining
everything as is, therefore any amount invested on preparing for hurricanes would be perceived
as a loss, whereas in the gain frame the reference point is having already spent a $10,000 deposit,
and any amount received back from the deposit would be perceived as a gain. For insurance
investment, the same manipulation was implemented and only the descriptions regarding
hurricane shutters were changed to hurricane insurance. As an illustration, Figure 2 shows the
four decision trees for each of the condition for the hurricane sample.

34

Figure 2. Decision trees for hurricane scenario, Experiment 1

35

Decision Vignettes
The decision vignette implemented in Experiment 1 described a hypothetical scenario of
selling a home due to job re-location. In the case of hurricanes, since the property is located in a
natural disaster-prone area and recent forecasts predicted an upcoming hazard on the way, the
homeowner in this scenario faces the choice of whether or not to invest in storm shutters or
hurricane insurance. In loss frame conditions, participants faced two options, first, to spend a
certain amount (on either retrofitting the property to be hazard-proof or purchasing an insurance
for the hazard) on preparing for the upcoming hurricane season (which is the risk averse sure
thing option), or to not spend any money and take their chances (which is the risk seeking
gamble option). If the choose the gamble option, theres a probabilit that nothing happened to
the property, and theres a chance that the haard cased damages to the property and they would
incur a considerable amount of repair costs. In the loss frame, the status quo is maintaining the
current state as is, therefore any amount the participant decides to spend on preparing for
hurricanes would be perceived as a loss. Table 4 presents an example of the loss frame decision
vignette in the context of hurricanes for risk mitigation.

36

Table 4. Loss Frame Vignette for Hurricane Condition
Loss Frame
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation
of the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the
Gulf of Mexico and Western Atlantic and an increase of hurricanes in the Pacific. NWS
predicted theres a 10% chance a hrricane ill hit or cit in the pcoming hrricane
season. Since your home is located on high ground, your home will be safe from flooding
during hurricane season. Homes vulnerable to hurricane winds typically have storm shutters,
which guarantee your home will not incur wind damage from a hurricane. (Note that the
shutters will not affect the sale price of the home.) Unfortunately, your home does not have
storm shutters. You have two options:

You may install storm shutters for a total cost of $1,000, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 90% chance that no hrricane dring the 2020 hrricane season cases ind
damage to your home and you will lose nothing.
2. Theres a 10% chance that a hrricane dring the 2020 hrricane season damages or
home; the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1000 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage
to my home or pay $10,000 to repair my home before it can be sold.

The gain frame used the same scenario with the addition of one detail: a new state
regulation requires the seller to put a deposit in an escrow account to pay for any potential
damages incurred before transferring the property to the new owner. In the gain frame,
participants faced two options, first, to invest a certain amount from the escrow account (for
either retrofitting the property to be hazard-proof or purchasing an insurance for the hazard) and
receive the remaining deposit back for sure (which is the risk averse sure-thing option), or to not
spend any money and take their chances (which is the risk seeking gamble option). If they
37

choose the gamble option, theres a probabilit t hat nothing happened to the property and the
deposit ill be retrned in fll, and theres a chance that the haard cased damages to the
property and none of the deposit will be returned. In the gain frame, the status quo is having
already spent $10,000 in security deposit, therefore any amount returned back from the deposit
would be perceived as a gain. Table 5 presents an example of the gain frame decision vignette in
the context of hurricanes for risk mitigation. Full descriptions of all decision vignettes are
summarized in Appendix A. To aoid the order of options cloding respondents choices, the
order in which the two options appeared were randomized in all surveys.
Table 5. Gain Frame Vignette for Hurricane Condition
Gain Frame
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation
of the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the
Gulf of Mexico and Western Atlantic and an increase of hurricanes in the Pacific. NWS
predicted theres a 10% chance a hrricane ill hit or cit in the pcoming hrricane season
(June 1st to Nov 30th). State real estate regulations require that the seller deposit $10,000 in
escrow to pay for any damages incurred from the upcoming hurricane season. The deposit will
be retrned to o in fll after or propert has been transferred to the ne oners name
and your property remains in good condition after hurricane season officially ends on Nov
30th, 2020. Since your home is located on high ground, your home will be safe from flooding
during hurricane season. Homes vulnerable to hurricane winds typically have storm shutters,
which guarantee your home will not incur wind damage from a hurricane. (Note that the
shutters will not affect the sale price of the home.) Unfortunately, your home does not have
storm shutters. You have two options:

You may install the storm shutters using part of the $10,000 deposit; you will receive the
remaining $9,000 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 90% chance that no hrricane cases ind damage to or propert dring the
2020 hurricane season, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that a hrricane does case ind damage to or propert dring the
2020 hurricane season, and none of the $10,000 deposit is returned to you.

38

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $9000 of the deposit back after
November 30, regardless any hurricane activity.
● I choose to not install the storm shutters, and receive the entire $10,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is
wind damage to my home before November 30.

Participants
Respondents were recruited from Amazon Mechanical Turk, and each worker received
$0.55 for participating in the survey. The survey took participants on average 8 minutes to
complete. Rouse (2015) found that when attention check questions are used, Turk workers
provided more reliable scores, therefore one attention check question was included in each
survey to filter out respondents who are not paying attention or participants who are responding
randomly. Participants who failed the attention check question were excluded from analyses.
Three different samples were collected for hurricane, flood and earthquake respectively. A power
analysis (Cohen, 1988) with power equal to or larger than 0.80 and sample size set to be
sufficient enough to obtain a moderate effect size (d = 0.50) revealed that for a 2 by 2 factorial
design, for each sample, each condition must have at least 50 participants (d = 0.527). Therefore,
the target population for each sample was set as 200 participants.
Table 6 summarizes the full demographic information of the three samples for
Experiment 1. For the hurricane sample, a total of 608 participants (152 participants in each
condition) who currently live in one of the hurricane-prone states identified by the NOAA
(Landsea, 2018; Jarrell, Mayfield, Rappaport & Landsea, 2001) were recruited. For flood
sample, a total of 620 participants (155 participants in each condition) who currently live in one
of the flood-prone states identified by the NOAA (Dahl, Spanger-Siegfried, Caldas & Udvardy,
2017; Garfield, 2018). For earthquake sample, a total of 612 participants (153 in each condition)
39

who currently live in California (Petersen et al., 2014) were recruited. In all three samples, < 1%
participants were dropped due to failure to answer attention check question correctly.
Across all three samples, the majority of participants have previous experience with the
disaster (with the earthquake sample reporting the highest percentage), slightly more than half
are female, about 2/3 currently own their home (the rest were previous home owners or soon to
be homeowners) and received some college education or college degree. The hurricane sample
are overall more politically conservative compare to the earthquake and flood sample, and the
earthquake sample has fewer current homeowners compared to the other two samples.
Table 6. Demographics Table for Experiment 1
Characteristic (%) Hurricane (N=608) Flood (N=620) Earthquake(N=612)
Response Time
Average (minutes) 6.97 8.78 8.53
IQR (minutes) 4.22 3.64 4.18
Previous Experience with the disaster
Yes 86.6 68.3 95.8
Gender
Female 55.6 54.4 56.6
Age (years)
Mean 40.6 (SD = 13.6) 39 (SD = 12) 38 (SD = 12)
Median 37 36 35
Education
High school or less 11.3 9.0 6.0
Some college or college degree 68.9 69.7 77
Masters degree or aboe 19.8 20.7 16.5
Prefer not to answer 0 0.6 0.5
Own or rent current residence
Own 70.4 67.0 60.9
Household Annual Income
< $10,000 to $39,999 32.6 20.0 24.9
40

$40,000 to $79,999 43.1 39.6 38.6
$80,000 to $150,000 or more 22.7 38.8 34.5
Prefer not to answer 1.6 1.6 2.0
Political Orientation
Liberal or leaning toward liberal 35.1 44.3 45.2
Neutral 22.0 27.6 26.2
Conservative or leaning toward
conservative
42.9 28.1 28.6

Measures
Objective Numeracy Since the decision problems used in this research all involve understanding
numerical values and probabilities, objective numeracy was measured. Foundational numeracy
skills are necessary for comprehending the risks associated with decisions, and previous research
in medical decision making showed that low numeracy skills could impede comprehension of
health statistics (Schwartz, Woloshin, Black & Welch, 1997; Woloshin, Schwartz, Byram,
Fischhoff& Welch, 2000; Schapira, Davids, McAuliffe & Nattinger, 2004). Objective numeracy
was measured using a 7-item Cognitive Reflection Test (CRT7; Toplak, West & Stanovich,
2014). Respondents answered seven open ended questions related to the construct of numeracy.
For example, a bat and a ball cost $1.10 in total; the bat costs a dollar more than the ball. How
much does the ball cost? Full description of the CRT7 can be found in Appendix C. Score for the
CRT7 was computed by first coding whether each answer is correct or incorrect and then
summing up the scores. Each correct answer was coded as 1 and each incorrect answer was
coded as 0, making the total range of scores between 0-7. Cronbachs alpha for the CRT7 was
0.72 in the original std (Toplak, West & Stanoich, 2014), and Cronbachs alpha for the
current study was estimated as 0.84, 0.80 and 0.78 for the earthquake, flood and hurricane
samples, respectively. CRT7 correlates well with other objective measures of numeracy and was
41

a significantly better predictor compared to either measures of intelligence or measures of
executive functioning for rational thinking tasks. The measure of objective numeracy was
included as covariates to account for group differences.
Data Analyses
A 2 by 2 (message frame by preparation context) logistic regression model with
demographic variables included as predictors was used to evaluate the effects of gain-loss
framing and mitigation versus insurance purchase on risk preferences. Gain-loss framing effects
and mitigation vs. insurance purchase were coded as -0.5 and 0.5 using contrast coding, and the
odds ratios obtained are therefore the average effects of both groups. The same data analyses
procedure and coding were used for all three samples in Experiment 1.
Based on previous literature from Prospect Theory (Kahneman & Tversky, 1979), I
postulate that participants assigned to loss frame vignettes tend to be risk-seeking, whereas
participants assigned to gain frame vignettes will tend to be risk averse. I make no specific
hypothesis regarding the difference between risk mitigation and insurance purchase decisions in
the context of natural disasters due to lack of previous literature.
Results
Figure 3 summarizes the percentage of participants choosing risk averse versus risk
seeking option for each of the four conditions for all three samples. For hurricanes, for
participants assigned to risk mitigation vignettes, 72.4% picked the risk averse option when the
vignette is in gain frame, whereas 65.1% chose the risk averse option when the vignette is in loss
frame. For participants assigned to hurricane insurance purchase vignettes, 68.4% picked the risk
averse option when the vignette is in gain frame, whereas 60.0% chose the risk averse option
when the vignette is in loss frame. For both risk mitigation and insurance purchase, distributions
42

of risk averse tendencies were consistent with my hypothesis that gain-frame messages are more
likely associated with risk averse preferences.
For floods, 87.7% participants who were assigned to risk mitigation vignettes picked the
risk averse option when the vignette is in gain frame, whereas 73.4% chose the risk averse option
when the vignette is in loss frame. For participants assigned to flood insurance purchase
vignettes, 77.7% picked the risk averse option when the vignette is in gain frame, whereas 75.0%
chose the risk averse option when the vignette is in loss frame. For both risk mitigation and
insurance purchase, distributions of risk averse tendencies were consistent with my hypothesis
that gain-frame messages are more likely associated with risk averse preferences.
For earthquakes, of all the participants assigned to risk mitigation vignettes, 65% picked
the risk averse option when the vignette is in gain frame, whereas 71.9% chose the risk averse
option when the vignette is in loss frame, contrary to my hypothesis that participants will more
likely choose the risk averse option when presented with gain-frame messages. For participants
assigned to earthquake insurance purchase vignettes, 64.6% picked the risk averse option when
the vignette is in gain frame, whereas 50.9% chose the risk averse option when the vignette is in
loss frame, consistent with the hypothesis that gain-frame messages are more likely associated
with risk averse preferences. In the following section, I discuss and present results from logistic
regression models.

43

Figure 3. Risk Preference Distribution for Experiment 1

Hurricane Sample, Experiment 1 Flood Sample, Experiment 1

Earthquake Sample, Experiment 1
Hurricane Sample Results
Table 7 presents the logistic regression model for the hurricane sample. First, focusing
on the effects of gain-loss framing and mitigation versus insurance, logistic regression revealed a
significant main effect for gain-loss framing (p = 0.01, = 0.42, 95% CI: 0.47 -0.92), indicating
participants are more likely to choose risk averse option when they were assigned to gain-frame
vignettes. The odds ratio for gain-loss framing is 1.52, suggesting that the odds of participants in
the gain frame condition being risk averse are about one and a half times the odds of participants
in the loss framing being risk averse. A significant main effect of mitigation versus hurricane
insurance was also detected (p = 0.04, = 0.33, 95% CI: 0.51 -1.02), indicating participants are
more likely to choose risk averse option when they were assigned to risk mitigation vignettes
44

than hurricane insurance vignettes. The odds ratio for risk mitigation versus hurricane insurance
is 1.78, suggesting that the odds of participants in the risk mitigation condition choosing the risk
averse option is 1.78 times the odds of participants assigned to hurricane insurance condition
selecting the risk averse option. No significant interaction effect between mitigation versus
insurance and gain-loss framing was detected.
Demographic variables (including previous experience with earthquakes, gender, age,
education, political orientation, own or rent current home, and household income) and an
individual difference variable (objective numeracy) were included as predictors to examine
potential impacts on risk preferences. None of the demographic variables was found to be a
significant predictor. No significant relationship from objective numeracy was detected.
Table 7. Logistic Regression Model Summary for Hurricane Sample, Experiment 1
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 0.42 0.18 0.01* 1.52 (0.47, 0.92)
Risk Mitigation vs. Insurance 0.33 0.17 0.04* 1.78 (0.51, 1.02)
Framing*Context Interaction 0.15 0.35 0.68 0.87 (0.43, 1.73)
Previous experience -0.49 0.27 0.15 0.67 (0.39, 1.15)
Gender 0.35 0.18 0.09 1.21 (0.98, 1.66)
Age -0.01 0.07 0.73 0.99 (0.99, 1.01)
Education 0.04 0.07 0.56 1.04 (0.91, 1.19)
Own or Rent 0.05 0.21 0.81 1.05 (0.70, 1.58)
Household Income 0.03 0.03 0.28 1.03 (0.97, 1.10)
Political Orientation -0.02 0.05 0.64 0.98 (0.89, 1.08)
Objective Numeracy 0.48 0.46 0.70 0.35 (0.67, 1.01)
*: p-value is significant at 0.05 level.
Flood Sample Results
Table 8 presents the logistic regression model summary for flood sample. First, focusing
on the effects of gain-loss framing and mitigation versus insurance, logistic regression revealed a
significant main effect for gain-loss framing (p = 0.002, = 0.65, 95% CI: 1.28 -2.90), indicating
participants are more likely to choose risk averse option when they were presented with gain-
45

framed vignettes. The odds ratio for gain-loss framing is 1.93, suggesting that the odds of
participants in the gain frame condition being risk averse over the odds of participants in the loss
framing being risk averse is 1.93. No significant effects of mitigation versus insurance or
interaction effect were detected.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
and an individual difference variable (objective numeracy) were included as predictors to
examine potential impacts on risk preferences. Household annual income was found to be a
significant predictor for risk preference (p = 0.02, = 0.09, 95% CI: 1.01 -1.17) with an odds
ratio of 1.10, suggesting that as household income increased another level, the odds of
respondents with higher household income being risk averse are 1.10 times the odds of
participants with lower household income being risk averse. No significant relationship from
objective numeracy was detected.
Table 8. Logistic Regression Model Summary for Flood Sample, Experiment 1
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 0.66 0.21 0.002* 1.93 (1.28, 2.90)
Risk Mitigation vs. Insurance 0.28 0.22 0.18 1.32 (0.88, 1.99)
Framing*Context Interaction 0.77 0.42 0.10 1.67 (0.96, 2.91)
Previous experience -0.25 0.22 0.26 0.78 (0.51, 1.20)
Gender -0.25 0.21 0.23 0.78 (0.52, 1.17)
Age 0.01 0.01 0.36 1.01 (0.99, 1.03)
Education 0.07 0.09 0.44 1.07 (0.90, 1.27)
Own or Rent 0.36 0.24 0.13 1.43 (0.90, 2.26)
Household Income 0.09 0.04 0.02* 1.10 (1.01,1.17)
Political Orientation -0.17 0.06 0.10 0.85 (0.75, 1.96)
Objective Numeracy 0.66 0.09 0.32 0.94 (0.75, 1.78)
*: p-value is significant at 0.05 level.
Earthquake Sample Results
Table 9 presents the summary of logistic regression model for the earthquake sample.
46

First, focusing on the effects of gain-loss framing and mitigation versus insurance, logistic
regression revealed a significant main effect for mitigation versus insurance (p = 0.006, = 0.49,
95% CI: 0.44-0.87), indicating participants are more likely to choose risk averse option when
they were presented with risk mitigation scenarios. The odds ratio for mitigation versus
insurance is 1.56, suggesting that the odds of participants assigned to risk mitigation condition
being risk averse over the odds of participants being risk averse in the insurance purchase
condition is 1.56. Logistic regression analysis also revealed a significant interaction effect
between gain-loss framing and risk mitigation versus insurance (p = 0.003, = 1.05, 95% CI:
1.43-3.67). The odds ratio for the interaction effect is 2.67, suggesting that the odds of
participants being risk averse under gain frame for insurance purchase condition are higher
compared to the odds of participants being risk averse under loss frame for risk mitigation
condition. The significant interaction effect suggests that mitigation versus insurance moderates
the gain-loss framing effect, and for earthquakes gain-loss framing works more effectively in
nudging people to choose risk averse option if the decision context is insurance purchase. No
significant main effect of gain-loss framing was detected.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
and an individual difference variable (objective numeracy) were included as predictors to
examine potential impacts on risk preferences. Whether participants own or rent their current
resident was found to be a significant predictor for risk preference (p = 0.004, = 0.54, 95% CI:
0.40-0.87) with an odds ratio of 1.45, suggesting that the odds of renters choosing the risk averse
option are 1.45 times the odds of homeowners choosing the risk averse option. No significant
relationship from objective numeracy was detected.
47

Table 9. Logistic Regression Model Summary for Earthquake Sample, Experiment 1
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. loss Framing 0.14 0.15 0.44 1.15 (0.81, 1.62)
Risk Mitigation vs. Insurance 0.49 0.18 0.006* 1.56 (0.44, 0.87)
Framing*Context Interaction 1.05 0.36 0.003* 2.67 (1.43, 3.67)
Previous experience -0.40 0.42 0.34 0.67 (0.29, 1.52)
Gender 0.08 0.17 0.65 1.08 (0.77, 1.52)
Age -0.01 0.01 0.21 0.99 (0.98, 1.05)
Education -0.04 0.08 0.61 0.96 (0.83, 1.12)
Own or Rent 0.54 0.17 0.004* 1.45 (0.40, 0.87)
Household Income 0.05 0.03 0.21 1.05 (0.99, 1.11)
Political Orientation 0.09 0.05 0.11 1.09 (0.98, 1.21)
Objective Numeracy 0.11 0.64 0.80 0.89 (0.19, 0.68)
*: p-value is significant at 0.05 level.
Discussion
The current experiment investigated the effects of message frame (gain vs. loss) and
preparation context (risk mitigation vs. insurance purchase). To the athors knoledge, this
research took the first step in both evaluating gain-loss framing effects for individual mitigation
decisions and directly comparing risk mitigation versus insurance purchase. By using relatable
and realistic decision vignettes, the current experiment empirically tested how framing effects
impact risk preferences for three types of natural disasters, respectively. Results from the current
experiment offer a set of implications for risk mitigation research and natural disaster related
policies, and also suggest several future research directions.
Table 10 provides a summary of main and interaction effects from Experiment 1. With
respect to gain-loss framing effects, gain-loss framing yielded significant main effects for both
floods and hurricanes, but gain-loss framing did not impact risk averse tendencies for
earthquakes. Respondents living in flood-prone states and hurricane-prone states are more likely
to be risk averse when presented with gain-frame descriptions, however gain-loss framing failed
to elicit significant changes in risk preference for the earthquake sample. The significant results
48

from flood and hurricane samples are consistent with previous literature (Spence & Pidgeon,
2010; Marti, Stauffacher, Matthes & Wiemer, 2018). For example, Marti and colleagues (2018)
evaluated the relationship between gain-loss framing effect and homeoners attitdes toa rd
earthquake protective action programs. In this study, gain-loss framing effects were manipulated
by changing word descriptions rather than shifting the reference point, and results showed
homeowners are most supportive toward proposed earthquake precautions when the message
was presented using gain frames. Similarly, Spence and Pidgeon (2010) examined the effects of
gain-loss framing on attitudes toward climate change mitigation, and reported that gain-frame
messages significantly increased positive and supportive attitude toward climate change
mitigation proposals.
Rothman, Bartels, Wlaschin and Salovey (2006) offered a theoretical foundation for the
effectiveness of gain-framed messages in promoting risk averse preferences. In their review of
the health literature, Rothman and colleagues found that the impacts of gain-loss framing largely
depends on peoples cognitie constral of the behaior. When the behaior or decision at hand
is perceived as having a low probability of leading to unpleasant outcome, information presented
using gain-frames are effective and advantageous in eliciting desired behaviors. By this
definition, in relation to the current research, for both risk mitigation and insurance purchase
contexts, risk averse options offer guaranteed protection of property at a small cost, whereas for
risk seeking options theres a possibilit of incurring additional costs to repair damage to the
property. Therefore, risk averse options would be deemed as having a low probability of bad
outcomes, and using gain-frames to promote protective actions is befitting.
With respect to risk mitigation versus insurance purchase, Table 10 indicates that
participants at risk for hurricanes and earthquakes preferred to structurally mitigate than to
49

purchase insurance, but this preference was not evident for residents at risk for floods.
Specifically, respondents living in states that are at risk for hurricanes and earthquakes are more
likely to select risk averse options when presented with risk mitigation options than insurance
purchase options, yet no significant difference was observed for the flood sample. These results
suggest that the preference of mitigating risk through structural retrofitting or purchasing
insurance might be context dependent. The indifference toward flood insurance from respondents
living in flood-prone states could potentially be explained by the lack of low cost and convenient
insurance plans (Kunreuther, 1996; Browne & Hoyt, 2000). Previous research in the realm of
flood insurance have proposed various methods for designing an affordable and reliable
insurance policy; however, private insurance agencies are not motivated to offer competitive
flood coverage due to low market penetration, making the National Flood Insurance Program
(NFIP) established by the U.S. government the only long-lasting widely available insurance plan
for more than 40 years (Browne & Hoyt, 2000; Michel-Kerjan, 2010; Michel-Kerjan &
Kunreuther, 2011). However, NFIP is only made available if the community where the decision
maker currently resides in agree to adopt required flood mitigation and land use measures
(Browne & Hoyt, 2000; Michel-Kerjan & Kunreuther, 2011), whereas for earthquakes and
hurricanes, numerous insurance plans are available on the market without the requirement of
community participation (Grace, Klein, & Liu, 2005; Kunreuther & Kleffner, 1992). This finding
points to the need for flood insurance education campaigns and the importance of providing
more diversified flood insurance options.
Regarding the interaction between gain-loss framing and preparation context, Table 10
indicates that only the earthquake context yielded a significant interaction between framing and
risk mitigation versus insurance purchase, however no main effects of framing and risk
50

mitigation versus insurance were detected. Reslts demonstrated that participants hore prone
to seismic risks are more likely to be risk averse when presented with earthquake insurance
options using gain-frames than loss frames. Intriguingly, the direction of effect for gain-loss
framing is consistent with Prospect Theory for earthquake insurance purchases, but opposite for
risk mitigation. For earthquake risk mitigation decisions, respondents are more likely to be risk
averse under the loss frame than the gain frame. For earthquake insurance purchase decisions,
respondents tend to be risk averse under the gain frame than the loss frame. Since gain loss
framing effects were detected in predicted direction for hurricanes and floods in both mitigation
and insurance investment context and for earthquakes only in insurance investment context,
results indicate that earthquake mitigation measures might be uniquely different and needs
further exploration in future research.
Table 10. Summary of Results for Experiment 1
Framing Mitigation vs.
Insurance
Interaction between Framing and
Mitigation vs. Insurance
Hurricanes Yes1 Yes2 No
Floods Yes1 No No
Earthquakes No Yes2 Yes3
1: Gain frame was associated with stronger risk averse tendencies.
2: Risk mitigation was associated with stronger risk averse tendencies.
3: Mitigation versus insurance moderates the gain-loss framing effect: gain frame was associated
with more risk averse tendencies for insurance purchase, but the opposite effect was reported for
risk mitigation.

The opposite direction of gain-loss framing effects for risk mitigation could potentially be
related to the unique characteristics of earthquakes. Compared to floods and hurricanes which
have a longer, gradual developing process, earthquakes have comparatively shorter durations and
can cause tremendous damage to building structures through strong ground motion within
51

seconds (McCann Jr & Shah, 1979). Furthermore, the frequency of large magnitude earthquakes
that resulted in substantial lives lost or damages to properties is relatively low in the United
States (on average less than once per year), whereas flood season and hurricane season occur
annually (USGS, 2020). In addition, there is currently no reliable method of predicting
earthquakes (Martínez-Álvarez, Reyes, Morales-Esteban & Rubio-Escudero, 2013), yet for
floods and hurricanes, weather forecasts, amount of rainfall and projected paths of tropical
cyclones and other various indicators all effectively serve as forecasts of impending disasters
(Elsner, 2003; Price, Yair, Mugnai, Lagouvardos, Llasat, Michaelides, ... & Harats, 2011). In the
decision vignette, the recommended risk mitigation measure is retrofitting current property,
which is not only costly, but also causes interference to daily routines. For a hazard that has a
lo freqenc of occrrence and nknon probabilit of casing damage, its n derstandable
that respondents in the earthquake sample would be reluctant to invest in risk mitigation and
would rather invest in earthquake insurances, as the process of buying insurance plans virtually
causes no extra burden to everyday life. In addition, the lack of framing effect for earthquake
mitigation might be a result of overall indifferent attitudes toward risk mitigation under gain
versus loss frame. From Figure 3 in previous section, for earthquake risk mitigation, the
percentages of participants choosing the risk averse option for gain frame (63.3%) and loss frame
(70.6%) are about equal: only a 7% difference was observed with overlapping error bars, while a
12% difference with non-overlapping error bars was observed for insurance investment. For a
strong enough large-scale earthquake, currently there are no mitigation method that could
prevent the house from falling (Tsonos, 2008; Güneyisi & Altay, 2008; Gu, Wu, Wu & Wu,
2010), therefore for earthquakes, participants might perceive a lack of efficacy for the retrofitting
method described in the scenarios. Since participants are overall indifferent between gain and
52

loss frames for risk mitigation, the interaction effect between framing and risk mitigation versus
insurance might be a result of the expected gain-loss effect for insurance purchase.
Analyses conducted with demographic variables as predictors also revealed interesting
findings. For the earthquake sample, current home ownership was found to be a significant
predictor of the mitigation decision; renters were more likely to select the risk averse option
compared to homeowners. From the previous discussion, residents from earthquake prone areas
prefer to purchase insurance as precaution measures rather than retrofitting their homes.
Earthquake sample has the lowest percentages of participants who are currently homeowners
compared to the hurricane sample and flood sample. Prospective homeowners who are currently
renters might be less well off financially and cannot risk the large loss outcome described in the
earthquake scenario, which explains why the earthquake sample were more likely to mitigate or
buy insurance than actual homeowners.
For the flood scenario, residents with higher household income are more likely to be risk
averse compared to participants with lower household income. Intuitively, risk mitigation
precautions that involves spending significant amount of money will naturally discourage lower
income households from voluntarily participating in such programs. In previous flood mitigation
studies, household income has consistently been found to be correlated with willingness to
participate in risk mitigation programs that require outlay of money (Edwards, 1993; Russell,
Goltz & Bourque, 1995; Lindell & Hwang, 2008; Peacock, Brody& Highfield, 2005; Fothergill
& Peek, 2004). Peacock (2003) found that among homeowners who live in flood-prone regions
in Florida, households with higher income have better quality shutters and protections for their
homes compared to low income households, which is consistent with my finding. However,
previous research found that for risk mitigation programs that are not costly (such as free
53

education meetings or simply purchasing a flashlight), the impact of household income on
willingness to participate diminished significantly (Lindell & Perry, 2000). These findings point
to an important policy implication low cost or incentive-based risk mitigation programs should
be developed to promote protective actions for low income households.
Interestingly, previous experience with natural disasters did not predict risk preferences
for any of the three disaster scenarios. Previous risk mitigation studies have reported mixed
findings with regards to the relationship between experience with the disaster and willingness to
take protective actions (Peacock, 2003, Tierney, Lindell, & Perry, 2001; Lindell & Hwang, 2008;
Peacock, Brody, & Highfield, 2005). The non-significant associations between previous
exposure to the hazard and willingness to take protective actions could be a result of unique and
idiosyncratic exposure experiences by respondents. Multiple studies have found that exposure
experiences that did not cause direct loss to personal property or bodily injury have very limited
influence on future protective actions. Moreover, at-risk populations who have directly
experienced a hazard but suffered very minor negative consequences might be subject to
optimism bias in the future and falsely underestimate the probability of risk for future hazards
(Peacock, 2003; Ge, Peacock & Lindell, 2011), which in the near miss literature is referred to as
a tendency to overestimate the resilience of the hazard mitigation system (Tinsley, Dillon &
Cronin, 2012). A near miss eent is defined as an eent that had a nontriial probabilit of
ending badly, but by chance did not (Dillon & Tinsley, 2008; Dillon, Tinsley & Cronin, 2011).
Tinsley, Dillon and Cronin (2012) demonstrated that when near misses are interpreted as
disasters that did not occur (defined as resilient near miss), people tend to underestimate the
danger be more risk-seeking, such as choosing not to take protective actions for the potential
hazard. If near misses were interpreted as disasters that almost happened (defined as vulnerable
54

near miss), people are more risk averse and more likely to take protective actions. Near miss
events also impact natural disaster decision making as well. Dillion, Tinsley and Cronin (2011)
reported that respondents who received resilient near miss information (house surviving a
hurricane without damage) were less likely to invest in flood insurance. It might be intuitive to
assume that people with previous experiences with a particular natural disaster would be more
likely to take protective action to mitigate the risk of that natural disaster, but these results
provide an important warning to policy makers to not overlook the impact of previous near miss
experiences. Future research could include more details, such as the extent of damage to personal
health and property when inquiring previous experience with the disaster, and explore the joint
effects of near miss experiences and gain-loss framing on intention to take protective actions.
Experiment I has several limitations. First, the mitigation decisions are hypothetical, and
I did not evaluate the effect of incentives on risk averse tendencies. Although the decision
vignettes used in the current study are realistic and my results demonstrated the experimental
manipulations were effective, there ere no conseqences folloing the respondents decisions.
The monetary reward described in the scenarios was not paid out and the gambles involved were
not resolved. Future research could explore whether including consequences of decisions impact
future risk propensities using a multi-stage sequential design.
Second, the PADM posits that social stakeholder perception is one of the three core
elements in the decision-making process of protective actions (Lindell & Perry, 2004, 2012).
Social stakeholders involved in natural disaster preparation decisions are defined as authorities
(government), emergency management agencies (such as National weather service), watchdogs
(media, environmental groups), employers and households (Pijawka & Mushkatel, 1991; Lang &
Hallman, 2005). In the decision vignettes implemented in this experiment, social stakeholder
55

perception is incorporated through a description of expert opinions (for example, quote from
seismic expert or the National Weather Service; see Appendix A for a complete description).
However, social influences could also come from sources that are psychologically closer to the
resident, such as neighbors, friends and family members. Peacock (2003) found that households
located in counties that require installation of hurricane shutters have shutters of significantly
better quality. Mileti and Darlington (1997) also reported that respondents are much more likely
to adopt earthquake prevention adjustments if other people are participating in such programs.
Therefore, future research could explore whether the psychological distance of social
stakeholders plas a role in inflencing peoples illingness to mitigate risk. If at-risk
populations are more likely to be influenced by social stakeholders they feel close to and trust,
then policy makers could consider targeting specific neighborhoods and encourage residents to
relay relevant information to family and friends to improve compliance with recommended
protective actions.
To summarize, in this chapter I evaluated the effects of gain-loss framing and risk
mitigation versus insurance purchase on risk preferences. My results demonstrate that gain-loss
framing effects are effective in nudging people toward risk averse preferences for flood,
hurricanes and earthquake insurance investments, but not for earthquake risk mitigation. The
lack of framing effect for earthquake mitigation might be a result of an overall disbelief in the
efficacy of earthquake retrofitting measures. My data also suggest that respondents from
earthquake and hurricane prone states are more likely to adopt retrofitting as a precaution
measure, yet participants at risk for floods did not exhibit any preference. Overall, results from
the current experiment offer integral insights for policy makers, and results suggest that the
effectiveness of gain-loss framing and preference for risk mitigation versus insurance purchase
56

might be context dependent. In the following chapter, I present a second experiment that further
explored whether the effect of gain-loss framing is robust when two scenario variables, EV and
disaster probability, are manipulated.

57

Chapter 6: Effects of Framing and Decision Determinants on Individual Choices -
Experiment 2
Results from Experiment 1 provide preliminary evidence that gain-loss framing can
influence individual disaster mitigation decisions, and gain-frame messages effectively nudged
respondents to be risk averse in both mitigation decision contexts for hurricanes and floods, and
for earthquake insurance purchase decisions. As previously discussed in Chapter 4, to better
understand the effect of gain-loss framing at individual level, Experiment 2 was conducted to
explore whether perceived risk and disaster probability moderate gain-loss framing effects. In
this experiment, perceived risk is operationalized by expected value (EV), which can be
calculated by estimating the probability weighted average value over all possible outcomes of a
particular alternative. Probability of incurring damage from the disaster was kept at 10% for EV
manipulations, and in high EV condition, EV was set at -500 (spending or losing $500); in low
EV condition, EV was set at -1,000 (spending or losing $1,000). Disaster probability refers to the
probability of incurring damage from the upcoming disaster, and EV was set as -500 (spending
or losing $500) for disaster probability manipulations. In high probability scenario, the
probability of incurring damage was 50%, whereas in low disaster probability scenario, the
probability of incurring damage was 5%.
While Experiment 1 compared risk mitigation versus insurance purchase, Experiment 2
focused on risk mitigation scenarios only. To the athors knoledge, the effect of EV and
disaster probability on risk mitigation decisions in the context of natural disasters has yet to be
empirically examined. The current experiment aimed to bridge the gap in literature by exploring
the interplay among gain-loss framing, EV, and disaster probability for individual mitigation
decision makign. Similar to Experiment 1, three natural disasters (earthquakes, floods and
58

hurricanes) were included in the design. The same screening procedures from Experiment 1
(regarding geolocations and homeownership) were implemented for Experiment 2.
Method
Design Overview
Experiment 2 was composed of two different designs, exploring the joint effects of gain-
loss framing and EV (Part I) and the joint effects of gain-loss framing and disaster probability
(Part II). Both Part I and II implemented a 2 by 2 factorial design (gain vs. loss framing by high
EV vs. low EV; gain vs. loss framing by high probability vs. low probability), as shown in Table
11. Gain-loss framing is manipulated by shifting the reference point (similar to Experiment 1).
EV is manipulated by shifting the expected value of outcomes (in high EV condition, expected
value is fixed at -500; in low EV condition, expected value is fixed at -1000) while keeping
disaster probability constant at 10%. Disaster probability is manipulated by changing probability
of incurring damage from the disaster (in high probability condition, probability of incurring
damage from disaster is fixed at 50%; in low probability condition, probability of incurring
damage from disaster is fixed at 5%) while keeping EV constant at -500 (losing or spending
$500).
Participants were randomly assigned into one of the four conditions, starting with a
decision vignette related to natural disaster preparation, followed by one attention check question
asking which natural disaster was presented in the vignette, and concluded with assessments of
demographic variables. Participants who failed the attention check question were excluded from
analyses. Since objective numeracy was not predictive of mitigation decision in Experiment 1,
objective numeracy was not assessed in Experiment 2.

59

Table 11. Design Layout for Experiment 2
High EV*Gain Frame High EV*Loss Frame
Low EV*Gain Frame Low EV*Loss Frame

High Probability*Gain Frame High Probability*Loss Frame
Low Probability*Gain Frame Low Probability*Loss Frame

Procedure
Similar to Experiment 1, manipulation of gain-loss framing is achieved by shifting the
reference point. For Part I of the experiment where EV was manipulated, the loss frame (which is
more intuitive to consider) involves a choice between spending a certain amount on installing
storm shutters to prepare for the upcoming hurricane season, or to pick the gamble in which
theres a 90% probabilit of incrring no damage from hrricanes and a 10% probability of
incurring a certain amount of damages. The gain frame used a slightly different description
where the decision maker must put afront a certain amount of security deposit in escrow, and the
participant can either choose to use part of the deposit to purchase storm shutters and receive the
remaining back for sure, or to not invest in preparing for the upcoming hurricane season, in
hich theres a 90% probability of not incurring any damage and receive the entire deposit back,
and a 10% probability of incurring damage and no deposit is returned. In high EV condition, the
EV is fixed at -500 (spending or losing $500), and in low EV condition the EV is fixed at -1,000
(spending or losing $1,000); disaster probability is fixed at 10% for all four conditions.
For Part II of the experiment where disaster probability was manipulated, similar to Part
I, the loss frame involves a choice between spending a $500 on installing storm shutters to
prepare for the pcoming hrricane season, or to pick the gamble in hich theres a probabilit
60

of incurring no damage from hurricanes and a probability of incurring $500 worth of damages.
The gain frame described a scenario where the decision maker must put afront a certain amount
of security deposit in escrow, and the participant can either choose to use $500 of the deposit to
purchase storm shutters and receive the remaining back for sure, or to not invest in preparing for
the pcoming hrricane season, in hich theres a probabilit of not incrring an damage and
receive the entire deposit back, and a probability of incurring damage and no deposit is returned.
In high disaster probability condition, the probability of incurring damage from the disaster is
fixed at 50%; in low disaster probability condition, the probability of incurring damage from the
disaster is fixed at 5%. EV is fixed at -500 (spending or losing $500) for all four conditions.
For both Part I and II, in the loss frame condition, the reference point is the status quo of
maintaining everything as is, and any amount spent on preparing for the hurricanes would be
perceived as a loss, whereas in the gain frame the reference point is having already spent a
certain amount as the security deposit, and any amount received back from the deposit would be
perceived as a gain. As an illustration, Figure 4 shows the eight decision trees for each of the
condition for the hurricane sample. Since the manipulations used for each of the three disasters
are the same and only description of disaster context changed accordingly, the decision trees will
remain the same for flood and earthquake scenarios.

61

Figure 4. Decision trees for hurricane scenario, Experiment 2

62

63

Decision Vignettes
Similar to Experiment 1, the decision vignette implemented in Experiment 2 described a
hypothetical scenario of selling a home due to job re-location. In the case of hurricanes, since the
property is located in a natural disaster-prone area and recent forecasts predicted an upcoming
hazard on the way, the homeowner in this scenario faces the choice of whether or not to invest in
storm shutters. In loss frame conditions, participants faced two options, first, to invest a certain
amount (for retrofitting the property to be hazard-proof) on preparing for the upcoming hurricane
season (which is the risk averse sure thing option), or to not spend any money and take their
chances (hich is the risk seeking gamble option). If the choose the gamble option, theres a
probabilit that nothing happened to the propert, and theres a chance that the haard cased
damages to the property and they would incur a considerable amount of repair costs.
In gain frame conditions, participants must put afront a security deposit into escrow first
and faced two options. First, to invest a certain amount from the security deposit on retrofitting
the property to be hazard-proof (which is the risk averse sure thing option), or to not use any of
the security deposit on preparing for the upcoming hurricane season and take their chances
(which is the risk seeking gamble option). If they choose to gamble, theres a probabilit that no
damage happened to the property and the deposit will be returned in full, and theres a chance
that the hazard caused damages to the property and they will lose all of the security deposit. In
the loss frame, the status quo is maintaining the current state as is, therefore any amount the
participant decides to invest on preparing for hurricanes would be perceived as a loss. In the gain
frame, the status quo is having already spent a certain amount as the security deposit, therefore
any amount returned back from the deposit would be perceived as a gain. Full description of the
decision vignettes can be found in Appendix B.
64

Participants
Respondents were recruited from Prolific.ac, an online questionnaire platform designed
specifically for crowdsourcing academic research. Previous research indicated that Prolific is
suitable and handy for conducting social and economic experiments (Palan & Schitter, 2018).
Each worker was awarded $0.55 for participating in the survey, and the survey took participants
on average 4 minutes to complete. Participants were not allowed to sign up for more than one
study in this experiment.
Six different samples were collected for hurricane EV, hurricane probability, flood EV,
flood probability, earthquake EV and earthquake probability, respectively. A power analysis
(Cohen, 1988) with power equal to or larger than 0.80 and sample size set to be sufficient
enough to obtain a moderate effect size (d = 0.50) revealed that for a 2 by 2 factorial design, each
condition must have at least 50 participants (d = 0.527), therefore the target population for each
of the six samples was set as 200 participants.
Table 12 summarizes the full demographic information of the six samples for
Experiment 2. For hurricane EV sample, a total of 200 participants (50 participants in each
condition) were recruited, and for hurricane probability sample, a total of 204 participants (51
participants in each condition) were recruited. For flood EV sample, a total of 204 participants
(51 participants in each condition) were recruited, and for flood probability sample, a total of 204
participants (51 participants in each condition) were recruited. For earthquake EV sample, a total
of 204 participants (51 participants in each condition) were recruited, and for earthquake
probability sample, a total of 208 participants (52 participants in each condition) were recruited.
All participants met the screening criteria of geolocations and attention check questions, and in
all six samples, < 2% participants were dropped due to failure to answer attention check question
65

correctly.
Across all six samples, the majority of participants have previous experience with the
disaster (with the earthquake sample reporting the highest percentage), less than half currently
owned their home (the rest were previous home owners or soon to be homeowners) and about
2/3 received some college education or college degree. The four hurricane and flood samples are
overall more politically conservative compare to the two earthquake samples, and the two
earthquake samples overall have higher household annual income compared to the hurricane and
flood samples.
Table 12. Demographics Table for Experiment 2
Characteristic (%)
Hurricane
EV
(N=200)
Hurricane
Probability
(N=204)
Flood EV
(N=204)
Flood
Probability
(N=204)
Earthquake
EV
(N=204)
Earthquake
Probability
(N=208)
Response Time
Average (minutes) 3.79 4.07 4.05 3.90 3.95 3.73
IQR (minutes) 1.70 1.36 1.99 1.31 1.58 1.55
Previous Experience with the disaster
Yes 67.8 64.0 44.1 45.8 93.6 94.1
Gender
Female 54.0 57.1 49.5 57.2 47.0 46.3
Age (years)
Mean 35.6
(SD = 14)
33.1
(SD = 11.2)
32.9(SD
= 12.5)
33.6(SD =
11.8)
34.9(SD =
12.3)
34(SD =
12.8)
Median 31 31 29 31 31.5 30
Education
High school or less 13.9 7.9 12.4 6.5 8.9 8.8
Some college or college degree 68.8 76.3 69.8 75.6 77.7 75.1
Masters degree or aboe 17.3 15.8 17.8 17.4 13.4 16.1
Prefer not to answer 0 0 0 0.5 0.0 0
Own or rent current residence
Own 46.5 38.9 47.0 42.3 45.0 47.3
Household Annual Income
< $10,000 to $39,999 38.6 34.0 36.6 34.3 23.8 24.9
66

$40,000 to $79,999 35.2 39.4 39.6 40.3 32.6 33.1
$80,000 to $150,000 or more 24.7 24.6 21.8 22.9 42.1 39.1
Prefer not to answer 1.5 2.0 2.0 2.5 1.5 2.9
Political Orientation
Liberal or leaning toward
liberal
59.9 57.6 55.9 60.7 58.5 65.7
Neutral 20.8 22.7 23.3 18.4 25.0 19.1
Conservative or leaning toward
conservative
19.3 19.7 20.8 20.9 16.5 15.2

Data Analyses
First, a 2 by 2 (message frame by EV level) logistic regression model with demographic
variables included as predictors was used to evaluate the effects of gain-loss framing and high
versus low EV on risk preferences. Second, a 2 by 2 (message frame by probability level)
logistic regression model with demographic variables included as predictors was used to evaluate
the effects of gain-loss framing and high versus low disaster probability on risk preferences.
Gain-loss framing effects, high vs. low EV and high vs. low probabilities were coded as -0.5 and
0.5 using contrast coding, and the odds ratios obtained are therefore the average effects of both
groups. The same data analyses procedure and coding were used for all six samples.
Since Experiment 2 was exploratory in nature to determine whether EV serves as a
moderator of gain-loss framing effect and whether gain-loss framing effect is robust over EV
manipulation, I make no specific hypothesis in regards to the direction of effect. Consistent with
Experiment 1, I postulate that gain frame will be associated with more risk averse tendencies,
and loss frame will be associated with more risk seeking choices.
Results
Figure 5 summarizes the distribution of risk averse choices for all six samples. For
hurricane EV sample, when participants were assigned to high EV condition, 72% picked the
67

risk averse option when the vignette is in gain frame, whereas 64% chose the risk averse option
when the vignette is in loss frame. For participants assigned to low EV condition, 90% picked
the risk averse option when the vignette is in gain frame, whereas 74% chose the risk averse
option when the vignette is in loss frame.
For hurricane probability sample, for participants assigned to high disaster probability
condition, 92% picked the risk averse option when the vignette is in gain frame, whereas 72%
chose the risk averse option when the vignette is in loss frame. For participants assigned to low
disaster probability condition, 88% picked the risk averse option when the vignette is in gain
frame, whereas 48% chose the risk averse option when the vignette is in loss frame.
For flood EV sample, 90% of participants in high EV conditions picked the risk averse
option when the vignette is in gain frame, whereas 86% chose the risk averse option when the
vignette is in loss frame. For participants assigned to low EV condition, 88% picked the risk
averse option when the vignette is in gain frame, whereas 54% chose the risk averse option when
the vignette is in loss frame.
For flood probability sample, of all the participants assigned to high disaster probability
condition, 92% picked the risk averse option when the vignette is in gain frame, whereas 82%
chose the risk averse option when the vignette is in loss frame. For participants assigned to low
disaster probability conditions, 82% picked the risk averse option when the vignette is in gain
frame, whereas 56% chose the risk averse option when the vignette is in loss frame.
For earthquake EV sample, for participants assigned to high EV condition, 74% picked
the risk averse option when the vignette is in gain frame, whereas 76% chose the risk averse
option when the vignette is in loss frame. For participants assigned to low EV condition, 84%
picked the risk averse option when the vignette is in gain frame, whereas 80% chose the risk
68

averse option when the vignette is in loss frame.
For earthquake probability sample, of all the participants assigned to high disaster
probability condition, 72% picked the risk averse option when the vignette is in gain frame,
whereas 88% chose the risk averse option when the vignette is in loss frame. For participants
assigned to low disaster probability condition, 60% picked the risk averse option when the
vignette is in gain frame, whereas 64% chose the risk averse option when the vignette is in loss
frame. In the next section, I present and discuss results from logistic regression for all six
samples.
Figure 5. Risk Preference Distribution for Experiment 2

Hurricane EV Hurricane Probability

Flood EV Flood Probability
72
90
64
74
40
50
60
70
80
90
100
High EV Low EV
Percentage of Risk Averse Choice
Gain Frame Loss Frame
92
88
72
48
40
50
60
70
80
90
100
High Disaster Probability Low Disaster Probability
Percentage of Risk Averse Choice
Gain Frame Loss Frame
90
88
86
54
40
50
60
70
80
90
100
High EV Low EV
Percentage of Risk Averse Choice
Gain Frame Loss Frame
92
82 82
56
40
50
60
70
80
90
100
High Disaster Probability Low Disaster Probability
Percentage of Risk Averse Choice
Gain Frame Loss Frame
69

Earthquake EV Earthquake Probability
Hurricane EV Sample Results
Table 13 presents the logistic regression model for hurricane EV sample. First, focusing
on the effects of gain-loss framing and high versus low EV, logistic regression revealed a
significant main effect for gain-loss framing (p = 0.007, = 1.22, 95% CI: 1.40 -6.13), indicating
participants are more likely to choose risk averse option when they were assigned to gain-frame
vignettes. The odds ratio for gain-loss framing is 3.37, suggesting that the odds of participants in
the gain frame condition being risk averse over the odds of participants in the loss framing being
risk averse is 3.37. A significant main effect of high versus low EV was also observed (p =
0.002, = -1.39, 95% CI: 0.10-0.60), indicating participants are more likely to choose risk
averse option when they were assigned to low EV conditions. The odds ratio for high versus low
EV is 4, suggesting that the odds of participants in the low EV condition being risk averse were 4
times the odds of participants in the high EV condition being risk averse. A significant
interaction effect between high versus low EV and gain-loss framing was also detected (p = 0.04,
= -1.83, 95% CI: 0.03-0.83). The odds ratio for the interaction effect is 5, for low EV
condition, the odds of participants in the gain frame condition being risk averse over the odds of
participants in the loss framing being risk averse is 2.46; for high EV condition, the odds of
participants in the gain frame condition being risk averse over the odds of participants in the loss
74
84
76
80
40
50
60
70
80
90
100
High EV Low EV
Percentage of Risk Averse Choice
Gain Frame Loss Frame
72
60
88
64
40
50
60
70
80
90
100
High Disaster Probability Low Disaster Probability
Percentage of Risk Averse Choice
Gain Frame Loss Frame
70

framing being risk averse is 1.78, which suggest that the odds of participants who saw gain-
frame messages in the low EV condition choosing the risk averse option were higher compared
to the odds of participants assigned to gain-frame messages in the high EV condition being risk
averse. The significant interaction effect suggests that EV moderates the gain-loss framing effect,
and gain-loss framing works more effectively in nudging people to choose risk averse option
when perceived EV is low.
Demographic variables (including previous experience with earthquakes, gender, age,
education, political orientation, own or rent current home, and household income) were included
as predictors to examine potential impacts on risk preferences. None of the demographic
variables was found to be a significant predictor.
Table 13. Logistic Regression Model for Hurricane EV Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 1.22 0.47 0.007* 3.37 (1.40, 6.13)
High vs. Low EV -1.39 0.47 0.002* 4 (0.10, 0.60)
Framing*EV Interaction -1.83 0.92 0.04* 5 (0.03, 0.83)
Previous experience 0.47 0.43 0.28 1.60 (0.69, 3.72)
Gender 0.63 0.39 0.10 1.87 (0.88, 2.99)
Age 0.01 0.04 0.75 1.00 (0.98, 1.03)
Education 0.20 0.15 0.18 1.23 (0.91, 1.65)
Own or Rent 0.83 0.43 0.08 1.32 (0.88, 2.34)
Household Income -0.04 0.06 0.54 0.97 (0.86, 1.08)
Political Orientation 0.11 0.12 0.40 1.11 (0.87, 1.41)
*: p-value is significant at 0.05 level.
Hurricane Probability Sample Results
Table 14 presents the logistic regression model for hurricane probability sample. First,
focusing on the effects of gain-loss framing and high versus low EV, logistic regression revealed
a significant main effect for gain-loss framing (p = 0.001, = 1.76, 95% CI: 2.71 -12.52),
indicating participants are more likely to choose risk averse option when they were assigned to
71

gain-frame vignettes. The odds ratio for gain-loss framing is 5.82, suggesting that the odds of
participants in the gain frame condition being risk averse over the odds of participants in the loss
framing being risk averse is 5.82. A significant main effect of high versus low disaster
probability was also observed (p = 0.04, = 0.80, 95% CI: 1.03 -4.78), indicating participants are
more likely to choose risk averse option when they were assigned to high disaster probability
conditions. The odds ratio for high versus low EV is 2.22, suggesting that the odds of
participants in the high disaster probability condition being risk averse were 2.22 times the odds
of participants in the low disaster probability condition being risk averse. No significant
interaction effect was detected.
Demographic variables (including previous experience with earthquakes, gender, age,
education, political orientation, own or rent current home, and household income) were included
as predictors to examine potential impacts on risk preferences. None of the demographic
variables was found to be a significant predictor.
Table 14. Logistic Regression Model for Hurricane Probability Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 1.76 0.41 0.001* 5.82 (2.72, 12.52)
High vs. Low Disaster Probability 0.80 0.40 0.04* 2.22 (1.03, 4.78)
Framing*Probability Interaction 0.48 0.81 0.56 0.62 (0.13, 3.06)
Previous experience -0.11 0.40 0.79 0.89 (0.41, 1.95)
Gender 0.76 0.40 0.08 2.13 (0.97, 3.71)
Age 0.01 0.02 0.80 1.01 (0.97, 1.04)
Education -0.33 0.17 0.09 0.72 (0.51, 1.00)
Own or Rent -0.01 0.42 0.99 0.99 (0.43, 2.31)
Household Income 0.06 0.07 0.38 1.06 (0.93, 1.21)
Political Orientation -0.17 0.12 0.17 0.85 (0.67, 1.07)
*: p-value is significant at 0.05 level.
Flood EV Sample Results
Table 15 presents the logistic regression model for flood EV sample. First, focusing on
72

the effects of gain-loss framing and high versus low EV, logistic regression revealed a significant
main effect for gain-loss framing (p = 0.003, = 1.19, 95% CI: 1.51 -7.23), indicating
participants are more likely to choose risk averse option when they were presented with gain
frame messages. The odds ratio for gain-loss framing is 3.30, suggesting that the odds of
participants in the gain frame condition being risk averse over the odds of participants in the loss
framing being risk averse is 3.30. Data also showed a significant main effect of high versus low
EV (p = 0.03, = 0.88, 95% CI: 1.11 -5.23), indicating participants are more likely to choose risk
averse option when they were assigned to high EV conditions. The odds ratio for gain-loss
framing is 2.42, suggesting that the odds of participants in the high EV condition being risk
averse were 2.42 times the odds of participants in the low EV condition being risk averse. No
significant interaction effects were detected.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
were included as predictors to examine potential impacts on risk preferences. None of the
demographic variables was found to be significant predictors.
Table 15. Logistic Regression Model for Flood EV Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 1.19 0.42 0.003* 3.30 (1.51, 7.23)
High vs. Low EV 0.88 0.42 0.03* 2.42 (1.11, 5.23)
Framing*EV Interaction -1.34 0.84 0.11 0.26 (0.05, 1.35)
Previous experience -0.32 0.40 0.42 0.73 (0.33, 1.58)
Gender 0.23 0.39 0.55 1.26 (0.59, 2.70)
Age -0.04 0.02 0.83 0.99 (0.97, 1.03)
Education -0.45 0.17 0.19 0.64 (0.45, 0.89)
Own or Rent 0.55 0.42 0.19 1.73 (0.76, 3.93)
Household Income 0.02 0.07 0.74 1.02 (0.89, 1.17)
Political Orientation 0.01 0.13 0.96 1.01 (0.78, 1.30)
*: p-value is significant at 0.05 level.
73

Flood Probability Sample Results
Table 16 presents the logistic regression model for flood probability sample. First,
focusing on the effects of gain-loss framing and high versus low probabilities, logistic regression
revealed a significant main effect for gain-loss framing (p = 0.003, = 1.15, 95% CI: 1.47 -6.82),
indicating participants are more likely to choose risk averse option when they were presented
with gain frame messages. The odds ratio for gain-loss framing is 3.17, suggesting that the odds
of participants in the gain frame condition being risk averse over the odds of participants in the
loss framing being risk averse is 3.17. Data also showed a significant main effect of high versus
low disaster probabilities (p = 0.007, = 1.05, 95% CI: 1.32 -6.14), indicating participants are
more likely to choose risk averse option when they were assigned to high disaster probability
conditions. The odds ratio for high vs. low probability is 2.86, suggesting that the odds of
participants in the high disaster probability condition being risk averse were 2.86 times the odds
of participants in the low disaster probability condition being risk averse. No significant
interaction effects were detected.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
were included as predictors to examine potential impacts on risk preferences. None of the
demographic variables was found to be significant predictors.
Table 16. Logistic Regression Model for Flood Probability Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 1.15 0.39 0.003* 3.17 (1.47, 6.82)
High vs. Low Disaster Probability 1.05 0.40 0.007* 2.86 (1.32, 6.14)
Framing*Probability Interaction 0.41 0.81 0.61 0.66 (0.14, 3.23)
Previous experience 0.33 0.37 0.37 1.39 (0.68, 2.85)
Gender -0.03 0.38 0.93 0.97 (0.46, 2.04)
Age -0.01 0.02 0.60 0.99 (0.96, 1.02)
74

Education 0.07 0.16 0.67 1.07 (0.78, 1.48)
Own or Rent -0.16 0.35 0.65 0.85 (0.43, 1.71)
Household Income -0.02 0.06 0.73 0.98 (0.87, 1.11)
Political Orientation -0.10 0.12 0.38 0.90 (0.72, 1.13)
*: p-value is significant at 0.05 level.
Earthquake EV Sample Results
Table 17 presents the logistic regression model summary for earthquake EV sample.
First, focusing on the effects of gain-loss framing and high versus low EV, logistic regression
revealed no significant main effect for gain-loss framing or high versus low EV. No significant
interaction effect was detected either.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
were included as predictors to examine potential impacts on risk preferences. Gender was found
to be a significant predictor for risk preference (p = 0.02, = -0.90, 95% CI: 0.26-0.58) with an
odds ratio of 2.44, suggesting that the odds of female participants being risk averse are 2.44
times the odds of male participants being risk averse.
Table 17. Logistic Regression Model for Earthquake EV Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing -0.02 0.37 0.97 0.98 (0.47, 2.03)
High vs. Low EV -0.38 0.38 0.31 0.68 (0.33, 1.43)
Framing*EV Interaction -0.21 0.74 0.78 0.81 (0.19, 3.47)
Previous experience -0.34 0.77 0.66 0.71 (0.16, 3.23)
Gender 0.90 0.39 0.02* 2.44 (0.26, 0.58)
Age -0.02 0.02 0.27 0.98 (0.95, 1.01)
Education -0.25 0.17 0.84 0.78 (0.68, 1.05)
Own or Rent 0.13 0.39 0.75 1.14 (0.53, 2.46)
Household Income -0.10 1.09 0.18 0.90 (0.80, 1.02)
Political Orientation -0.19 0.13 0.14 0.83 (0.65, 1.07)
*: p-value is significant at 0.05 level.

75

Earthquake Probability Sample Results
Table 18 presents the logistic regression model summary for earthquake probability
sample. First, focusing on the effects of gain-loss framing and high versus low disaster
probabilities, logistic regression revealed a significant main effect for high versus low disaster
probability (p = 0.005, = 0.96, 95% CI: 1.36 -3.93), indicating participants are more likely to
choose risk averse option when they were presented with high disaster probability scenarios. The
odds ratio for high vs. low probability is 2.62, suggesting that the odds of participants being risk
averse in the high disaster probability condition were 2.62 times the odds of participants being
risk averse in the low disaster probability condition. No significant effect of gain-loss framing
nor a significant interaction effect between gain-loss framing and high versus low disaster
probabilities was detected.
Secondly, demographic variables (including previous experience with earthquakes,
gender, age, education, political orientation, own or rent current home, and household income)
were included as predictors to examine potential impacts on risk preferences. None of the
demographic variables was found to be significant predictors.
Table 18. Logistic Regression Model for Earthquake Probability Sample, Experiment 2
Predictor Standard
Error (SE)
p Exponential
Beta
95%
Confidence
Interval
Gain vs. Loss Framing 0.49 0.34 0.15 0.61 (0.32, 1.18)
High vs. Low Disaster Probability 0.96 0.34 0.005* 2.62 (1.36, 3.93)
Framing*Probability Interaction 0.80 0.67 0.29 0.50 (0.13, 1.84)
Previous experience -0.56 0.64 0.38 0.57 (0.16, 1.99)
Gender 0.90 0.31 0.02* 2.44 (0.26, 0.58)
Age 0.01 0.01 0.98 0.99 (0.97, 1.03)
Education -0.05 0.15 0.77 0.96 (0.71, 1.29)
Own or Rent 0.17 0.29 0.56 1.19 (0.67, 2.11)
Household Income 0.01 0.05 0.98 1.01 (0.91, 1.11)
Political Orientation -0.13 0.11 0.25 0.88 (0.71, 1.09)
*: p-value is significant at 0.05 level.
76

Discussion
Building on findings from Experiment 1, Experiment 2 explores the effects of message
frame (gain vs. loss), EV (high vs. low) and disaster probability (high vs. low) on risk
preferences. This research bridges the gap in current research by providing both a direct
comparison of varying levels of EV and disaster probability (manipulated independently) and an
investigation of whether EV or disaster probability moderate gain-loss framing for individual
mitigation decisions. Through the implementation of realistic decision vignettes, the current
experiment empirically and quantitatively tested how two scenario variables related to perceived
risk affect impact mitigation decision making and how they independently influence gain-loss
framing effects. In this section I discuss how findings from the current experiment shed light on
multiple inspirations for natural disaster related policies, and also highlight several research
directions for future researchers.
Table 19 provides a summary of main effect results for Experiment 2 (no interaction
effect was discovered for Experiment 2). First, regarding gain-loss framing effects, gain-loss
framing yielded predicted significant main effects for both floods and hurricanes, but not for
earthquakes. Respondents who are at risk for floods and hurricanes tend to be risk averse when
assigned to gain-frame vignettes, yet for both of the earthquake samples, gain-loss framing did
not yield any significant effect in risk preference. The impact of gain-loss framing in promoting
risk averse choices for flood and hurricane scenarios is consistent with previous research
findings (Rothman, Bartels, Wlaschin & Salovey, 2006; Marti, Stauffacher, Matthes & Wiemer,
2018). Lack of any influence for gain-loss framing on both the earthquake EV sample and
earthquake disaster probability sample are consistent with the finding from Experiment 1.
Considering Experiment 1 was conducted on MTurk and Experiment 2 was conducted on
77

Prolific, the replication of results across two different survey platforms further strengthens the
robustness of these findings. Furthermore, despite manipulations of EV and disaster probability,
gain-loss framing results across all three natural disasters exhibited a remarkably consistent
pattern with findings from Experiment 1, providing further empirical support for the robustness
of gain-loss framing effects across different levels of disaster risk (EV and disaster probability).
Results from Experiment 1 and 2 point to the conclusion that in the context of natural disaster
preparation, the effectiveness of gain-loss framing is context dependent. In Chapter 5 I provided
a discussion of how earthquakes uniquely differ from floods and hurricanes in terms of
frequency, prediction method and duration; the consistent pattern of findings across two
experiments opens promising avenues for further research on how context moderates gain-loss
framing. Future researchers should incorporate multiple natural hazards with diverse
characteristics and investigate the gain-loss framing on mitigation decisions across different
hazard contexts.
Second, in regards to the manipulation of EV, Table 19 reveals that EV did not yield any
significant effect for earthquakes. Although significant main effects of EV were discovered for
the other two natural disasters, the direction of effects is opposite for hurricanes and floods.
Respondents who are subject to flood risks are more likely to be risk averse when they perceive a
high degree of EV, however, respondents from the hurricane sample who viewed high EV
vignettes are more risk seeking compared to participants presented with low EV vignettes. The
opposite direction of effects from EV on risk preferences for hurricanes and floods and absence
of effect for earthquakes suggest that the effect of EV is moderated by disaster context, similar to
findings from gain-loss framing where framing effects were detected for hurricanes and floods
but not for earthquakes in both experiments.
78

Third, Table 19 indicates that the effect of disaster probability was identified
consistently in the same direction across all three types of natural disasters. Respondents in all
three disaster scenarios (earthquake, flood and hurricane) exhibited the same pattern of greater
risk aversion in high disaster probability conditions compared to low disaster probability
conditions. In both high and low disaster probability conditions, participants favored the risk
averse sure thing option more than the risk seeking gamble option (in nearly all conditions, the
percent of participants picking the gamble is less than 50%, with the only exception of hurricane
low disaster probability condition where the percent of participants picking gamble is 52%).
Respondents are more likely to pick risk averse option for high disaster probability than for low
disaster probability, and even for low disaster probability conditions, participants are still more
likely to pick the risk averse option than gamble. Furthermore, the robust association between
high probability of disaster and more risk averse tendencies point to an insightful account for at-
risk poplations reluctance to mitigate natural disaster related risks in real life. In the current
experiment, the low end of disaster probability is 5%, which is still higher than the true
probability of any particular individual experiencing damages to property from a disaster at a
given year. Results from Experiment 2 suggest that people at risk for natural disasters are more
willing to gamble when the probability of disaster is low (5%) compared to high (50%),
realistically, the true probability of someone incurring damages from the disaster may be well
below the low end probability in the current experiment (5%), thus overall tendency to invest in
risk mitigation measures would be further reduced.

79

Table 19. Summary of Results for Experiment 2
Risk Mitigation Context Only
Framing EV Disaster Probability
Hurricanes Yes1 Yes2 Yes4
Floods Yes1 Yes3 Yes4
Earthquakes No No Yes4
1: Gain frame was associated with more risk averse tendencies.
2: Low EV was associated with more risk averse tendencies.
3: High EV was associated with more risk averse tendencies.
4: High probability of disaster was associated with more risk averse tendencies.

Probability of incurring damage from the disaster is also related to how much control the
decision maker can exert on the situation at hand. In emotion psychology research, control has
long been theorized as a crucial part of the coping appraisal process (Rotter, 1966; Scherer,
1982). Previous research also introduced two different types of control: internal control, which
refers to situations in which the outcome of the decision is dependent on the decision-maker
herself, and external control, which refers to circumstances in which the outcome of the decision
is contingent on factors external to the decision maker, such as the actions of others or chance
(Rotter, 1966; Langer, 1975; Lichtenstein & Fischhoff, 1977). Considering the consistent
relationship between high disaster probability and increased likelihood to mitigate disaster risk, a
more thorough investigation of how internal and external control impacts risk preferences in the
context of natural disaster preparations is called for.
Analyses conducted with demographic variables as predictors yielded an interesting
finding. For earthquake EV sample, gender was found to be a significant predictor for risk
preference; specifically, female participants were more likely to select the risk averse option
compared to male participants in the context of earthquake risk mitigations. Previous research
80

from natural hazard preventions have noted mixed findings regarding the role of gender on risk
attitudes. Vinnell, McClure and Milfont (2017) conducted a study of residents from earthquake
prone areas, and reported male participants were more likely to vote for the legislation when the
message is framed positively. On the other hand, other studies have reported females to be more
risk averse compared to males (Brody, 1984; Stern, Dietz & Kalof, 1993; Fothergill, 1996).
Previous research has also found interesting interactions between gender and race, specifically,
white males have significantly lower levels of risk perceptions compared to white females, black
males and black females (Flynn, Slovic & Mertz, 1994). Therefore, another potential research
direction is to further investigate the joint role of gender and other demographic variables on
decisions to mitigate risk from natural disasters.
One potential direction for future research is to expand the investigation to include other
parameters related to perceived risk, such as skewness of distribution, range of the gamble and
assessing risk preferences over multiple (repeated) gambles. Coombs and colleagues offered a
rich body of research in exploring variables that could impact perceived risk (Coombs & Pruitt,
1960; Coombs & Huang, 1970; Coombs & Bowen, 1971; Coombs, 1975). Coombs model
hypothesizes that each decision maker has a most preferred risk level and the decision maker
tends to choose the option thats closest to her ideal risk leel. Coombs and colleag es also
reported that the perceived risk of gambles with equal EVs increases monotonically with the
amount lost, and perceived risk varies with skewness of a gamble (Coombs & Pruitt 1960,
Coombs & Huang 1970; Coombs & Bowen, 1971). Investigating whether other decision
determinants moderate gain-loss framing effects could bring novel insights for designing more
effective risk communication programs, which should be explored in future studies.
One limitation of the current study is that only EV and disaster probability were
81

manipulated, and future research could consider including more parameters to build a more
thorough model for estimating risk propensities. More recently, Ito and Marsella (2013)
proposed a computational utility modeling approach for incorporating various modeling
parameters (including pleasantness sensitivity, goal congruence sensitivity, control sensitivity
and degree of risk-seeking/aversion), and reported improved accuracy in predicting decisions
when accounting for multiple factors related to the decision at hand. Future research should
explore the possibility of developing a quantitative framework that integrates the effects from
framing, EV, disaster probabilities and other relevant constructs, which could aid in determining
the most suitable option for the decision maker.
A second limitation of the study is that only vignettes related to risk mitigation were
included in this experiment. As discussed in Experiment 1, residents from hurricane and
earthquake prone states favor risk mitigation over insurance purchase, and are more likely to be
risk averse when presented with risk mitigation plans. It would be intriguing to explore whether
findings from the current experiment still hold for insurance purchase contexts, and results would
provide important insights for both retrofitting and insurance promotion programs.
In summary, this chapter explored whether EV and disaster probability influence
mitigation decision making directly, and whether either moderates gain-loss framing effects in
the context of hurricane, flood and earthquake preparation. Results indicate that consistent with
Experiment 1, gain framing is effective in evoking risk averse mitigation decisions for floods and
hurricanes, however gain-loss framing had no impact on earthquake preparation decisions. The
consistent pattern of results across two experiments indicate that gain-loss framing effects are
context dependent, and in order to improve compliance with protective actions, policy makers
should bear in mind the unique characteristics of various natural hazards. Second, my data also
82

indicate that EV affects flood and hurricane samples in opposite direction, but earthquake
samples did not report any significant effect. Participants from flood prone states are more likely
to be risk averse when EV is perceived as high, but respondent at risk for hurricanes displayed
the opposite pattern. Future research should further examine the impact of EV on risk preference
for various types of natural hazards. Lastly, the effect of disaster probability is highly consistent
across all three types of natural disasters, specifically, when people perceive the probability of
incurring damage from the disaster is high, they are more likely to be risk averse. This finding
provides a critical implication for policy makers future natural disaster education programs
could concentrate more on presenting risk mitigation methods using a gain frame, and de-
emphasize the probability of incurring damage from the upcoming disaster in order to elicit
higher compliance with risk mitigation measures. Overall, results from the current experiment
largely replicated with findings from Experiment 1, and results suggested that gain-loss framing
is effective and robust in influencing individual level choices in the contexts of floods and
hurricanes, across various manipulations of EV and disaster probability. In the next chapter, I
provide a general discussion of this project.
83

Chapter 7: General Discussion
This dissertation was motivated by the rich literature embedded in both Prospect Theory
and the PADM framework, and presented two empirical studies involving effects of gain-loss
framing for individual mitigation decision making. In the first experiment (Chapter 5), I
evaluated the effects of both gain-loss framing and risk mitigation versus insurance, and reported
gain frame messages effectively encouraged risk averse choices. I also noted that residents at risk
for earthquakes and hurricanes favor retrofitting their homes considerably over insurance
purchases, yet residents from flood-prone states did not show any significant preference.
Experiment 1 is one of the first to empirically examine gain-loss framing effects for individual
household disaster mitigation decisions, and one of the first to provide a direct comparison
between risk mitigation and insurance purchases.
The second study (Chapter 6) built upon findings from Experiment 1 and further
explored whether gain-loss framing effects are robust when EV and disaster probability were
manipulated independently of each other. Consistent with Experiment 1, gain-framed messages
successfully promoted risk averse tendencies for floods and hurricanes, but not for earthquakes.
Despite manipulations of both EV and disaster probability, gain-loss framing showed similar
effect patterns for the three natural disasters across two experiments, which provided powerful
support for the robustness of gain-loss framing effects for individual level decisions. The effect
of EV differed drastically for the three natural disasters; for earthquakes EV did not possess any
significant impact on risk preferences, for floods perceived higher levels of EV is associated with
risk averse tendencies, and for hurricanes respondents are more risk seeking under high EV
conditions. Contrary to the diverse pattern observed for EV, the effect of disaster probability is
consistent across earthquakes, floods and hurricanes. For all three natural disasters, higher
84

probability of incurring damage was associated with increased likelihood of selecting the risk
averse mitigation option. The robustness of findings indicate that disaster probability should be
de-emphasized in designing risk communications.
Contributions to current literature
Table 20 summarizes findings from both Experiment 1 and 2, which indicate consistent
main effects for gain-loss framing, across two experiments for floods and hurricanes.
Specifically, respondents assigned to gain frames are more likely to be risk averse compared to
respondents assigned to loss frames, consistent with my hypothesis. Regarding the main effects
of risk mitigation versus insurance purchase, respondents at risk for earthquakes and hurricanes
both reported significant preference to mitigation over purchasing insurance, but the flood
sample did not yield any significant results. Interestingly, results from the earthquake disaster in
Experiment 1 included a significant interaction effect between gain-loss framing and type of
mitigation (structural mitigation vs. insurance purchase). More specifically, respondents assigned
to the insurance purchase decision were more risk averse under the gain frame than the loss
frame (consistent with the other two disaster contexts), but there was no such effect (somewhat
in the reverse direction) for structural mitigation decisions. As discussed in Chapter 5, with a
strong enough large scale earthquake, no retrofitting method can prevent a house from falling
(Tsonos, 2008; Güneyisi & Altay, 2008). The lack of efficacy for earthquake retrofitting
measures might explain the results from earthquake risk mitigation condition: for gain frames
and loss frames, the percentages of participants picking the risk averse option are approximately
equal. For earthquake insurance, participants are more likely to be risk averse under the gain
frame compared to loss frame, consistent with my hypothesis. As a result, a significant
interaction effect between framing and risk mitigation versus insurance investment was detected.
85

For the manipulation of perceived risk, results were varied across the three disasters. For the
earthquake EV sample, manipulation of EV failed to elicit significant effects. Respondents at
risk for floods and hurricanes reported significant main effect of perceived risk in opposite
directions: respondents from flood-prone states are more likely to be risk averse when they
perceive a high degree of EV, yet respondents from hurricane-prone states are more likely to be
risk seeking when presented with high EV vignettes. Lastly, the manipulation of disaster
probability elicited robust main effects across all three types of natural disasters. For
earthquakes, hurricanes and floods, high probability of incurring damage from the upcoming
natural disaster was consistently associated with stronger risk averse tendencies, highlighting the
need to de-emphasize or omit description of the low probability of encountering disaster for any
particular individual in risk communications.
Table 20. Summary of Results for Experiment 1 and 2
Experiment 1 Experiment 2
(Mitigation Context Only)
Framing Mitigation versus
Insurance
Interaction Framing EV Disaster
Probability
Hurricane Yes1 Yes2 No Yes1 Yes4

Yes6
Flood Yes1 No No Yes1 Yes5 Yes6
Earthquake No Yes2 Yes3 No No Yes6
1: Gain frame was associated with more risk averse tendencies.
2: Risk mitigation was associated with more risk averse tendencies.
3: Mitigation versus insurance moderates the gain-loss framing effect: gain frame was associated
with more risk averse tendencies for insurance purchase, but the opposite effect was reported for
risk mitigation.
4: Low EV was associated with more risk averse tendencies.
5: High EV was associated with more risk averse tendencies.
6: High probability of disaster was associated with more risk averse tendencies.

86

Findings from Experiment 1 and 2 offer several theoretical and practical contributions.
First, despite the abundance of literature on gain-loss framing effects (Krishnamurthy, Carter &
Blair, 2001; Kühberger, 1998; Levin, Gaeth, Schreiber& Lauriola, 2002), previous research has
focused only on policy-level decisions, which are not comparable to common decisions
individuals and households face in real life. While gain-loss framing effects might be similar for
policy-level and individual hazard mitigation decisions, there was no empirical evidence to
support this conjecture. Results from the current experiments provide the first empirical evidence
demonstrating the applicability of individual level gain-loss framing effects in the context of
natural disaster preparation. In addition, gain-loss framing effects were highly consistent across
the two experiments, despite variation in the crowdsourcing platform used for recruitment, and
the manipulations of EV and disaster probability, providing strong empirical evidence for the
effect of gain-loss framing in particular contexts (flood and hurricane mitigation).
Another important finding is that people are overall more risk averse for individual level
decisions: across two experiments and a total of nine different samples with over 3,000
participants, in almost all conditions the percentages of participants choosing the risk averse
option are larger than 50% (with the exception of 2 cells out of 36 different conditions over the 2
experiments, both approximately 50%). The overall percentage of risk averse choices across all
36 conditions averaged to 71.71%. In previous literature that evaluated gain-loss framing effects
using policy-level scenarios, the difference between risk seeking and risk averse choices can be
quite drastic. For example, in the original Asian Disease Problem, 72% participants chose the
risk averse option under gain frame whereas only 22% opted for the risk averse option under loss
frame (Tversky & Kahneman, 1981). These results indicate that people perhaps possess a higher
risk averse baseline for individual level decisions compared to policy level decisions. This
87

finding can be intuitively understood by considering the difference between policy and
individual level decisions. For policy level decision problems, the outcome of the decision
makers choice impacts the well being of other people, whereas for individual level decision
problems, the decision outcome impacts only the decision maker (and family) directly.
Furthermore, my results demonstrated that even when participants are overall more risk averse,
gain frames were still effective in nudging people to choose the risk averse option for both
hurricane and flood samples across two experiments, which further demonstrated the
applicability and strength of gain-loss framing effects in the context of natural disaster
preparation.
Second, most research in natural disaster preparation literature only focused on one type
of disaster (Palm & Hodgson, 1992; Peacock, 2003; Kerjan, 2010; Ge, Peacock & Lindell,
2011), and this dissertation included multiple types of natural disasters to evaluate whether gain-
loss framing effects might be moderated by disaster context, and the extent to which framing
effects can be generalized across natural disasters. Results did provide support for the
effectiveness of gain-loss framing across two different hazard contexts in both experiments
framing effect was prominent for floods and hurricanes, but not for earthquakes. Furthermore,
both experiments screened participants based on both geolocation and home ownership, assuring
the sampled respondents are representative of the target decision maker population -
homeowners who are at risk for upcoming natural disasters. The representativeness of the
samples not only contributed to the consistency of findings across two experiments, but also
added strength to the generalizability of the findings.
Lastly, results from Experiment 1 and 2 highlights the importance of natural disaster
context. Gain-loss framing effects were detected for floods and hurricanes which share many
88

characteristics, but not for earthquakes which possess unique attributes in terms of prediction
techniques, frequency and duration. Furthermore, significant differences were observed for
earthquakes and hurricanes with a preference for retrofitting over insurance purchases, but not
for flood insurance which differs from earthquake and hurricane insurance in terms of
availability and affordability (Browne & Hoyt, 2000; Michel-Kerjan, 2010; Michel-Kerjan &
Kunreuther, 2011). Taken together, these findings indicate that risk attitudes toward natural
disaster preparation might be context dependent. Furthermore, these findings offer practical
implications to promote risk averse choices. Policy makers should take into consideration the
unique attributes of various hazards, and utilize gain-loss framing and perceptions of disaster risk
(EV and disaster probability) in accordance with these attributes. For example, a perceived high
probability of incurring damage has been a reliable factor in eliciting risk averse tendencies, in
designing effective risk communication messages, policy makers could consider making the
probability of someone encountering a disaster salient, and de-emphasize or omit discussing the
probability of disaster for any particular individual such as the decision maker itself.
Does framing violate free will?
Although findings have demonstrated gain-loss framing and perceived risk (especially
disaster probability) could serve as useful tools in shaping individual disaster mitigation
decisions, potential criticism could arise as to whether policy makers have the authority to
interfere with ciilians choices. Nobel laurate Richard Thaler and law scholar Cass Sunstein
proposed the concept of nudge as a method to move people to select options that will increase
their overall happiness or satisfaction in life (Thaler & Sunstein, 2009). In the nudge framework,
all decisions are deemed as having contexts associated with them, and any context incorporated
with the decision could potentially nudge people toward different options. To elicit desired
89

responses from decision makers, the context in which the decision is made could be modified by
choice architects (people with the responsibility or authority to define or construct the decision
contexts) to make the (risk averse) option selected for the decision maker appear more attractive.
Findings from Experiment 1 and 2 relate to this framework in that manipulation of gain-loss
framing is a method for nudging respondents to choose the risk averse option. Skeptics of the
nudge framework often criticize the approach by asserting ndge iolates peoples freedom of
choice (Hausman& Welch, 2010), which could also be relevant in applying gain-loss framing for
mitigation decision making.
I old like to proide a jstification of sing framing effects in shaping peoples
decisions by introducing the libertarian paternalism account from Thaler and Sunstein (2009).
This account might seem contradictory as libertarian and paternalism have opposite meanings,
but it can be understood as consisting of two parts. First, nudge is libertarian because people are
still free to choose; no interferences will be in place to force decision makers to pick the option
that policy makers approve of. Second, nudge is paternalistic in the sense that parties responsible
for setting up the decision context (policy makers or even governments) can alter the context to
make risk averse options more appealing. Therefore, the combination of both accounts produce a
justification for steering peoples choices in the direction that old make them better off
because decision makers are still making the decision based on their free will, and any
manipulation of gain-loss framing functions only to encourage people to choose the risk averse
option.
Concluding remarks
The current dissertation provides empirical evidence regarding the effects of individual
level gain-loss framing, EV and disaster probability on mitigation decisions for natural disasters.
90

Findings from the two experiments are meaningful because they: 1. Provide empirical evidence
supporting the hypothesis that gain-loss framing effects at the policy or organizational level
would generalize to gain-loss framing effects for individual and household decisions; 2.
Demonstrate the generalizability and robustness of gain-loss framing effects across different
natural disaster contexts and different levels of perceived disaster risk; and 3. Highlight the
integral role of context dependency in designing risk communication messages, indicating that
policy makers should consider unique attributes of each natural disaster and tailor the decision
context accordingly to nudge decision makers to more prudent, risk-averse options. In future
research, I plan to expand the scope of this dissertation and incorporate psychological factors and
develop more effective ways to steer people into risk averse choices.

91

References
Agnew, J. R., Anderson, L. R., Gerlach, J. R., & Szykman, L. R. (2008). Who chooses annuities?
An experimental investigation of the role of gender, framing, and defaults. American
Economic Review, 98(2), 418-22.
Ashley, W. S. (2007). Spatial and temporal analysis of tornado fatalities in the United States:
1880 2005. Weather and Forecasting, 22(6), 1214-1228.
Baddeley, A. (2000). The episodic buffer: a new component of working memory?. Trends in
Cognitive Sciences, 4(11), 417-423.
Ballantyne, M., Paton, D., Johnston, D., Kozuch, M. and Daly, M. (2000), Information on
Volcanic and Earthquake Hazards: The Impact on Awareness and Preparation. Institute
of Geological and Nuclear Sciences Report, No. 2, Institute of Geological and Nuclear
Sciences, Lower Hutt.
Bodine, D. J., Kumjian, M. R., Palmer, R. D., Heinselman, P. L., & Ryzhkov, A. V. (2013).
Tornado damage estimation using polarimetric radar. Weather and Forecasting, 28(1),
139-158.
Boruff, B. J., Easoz, J. A., Jones, S. D., Landry, H. R., Mitchem, J. D., & Cutter, S. L. (2003).
Tornado hazards in the United States. Climate Research, 24(2), 103-117.
Brody, C. J. (1984). Differences by sex in support for nuclear power. Social Forces, 63(1), 209-
228.
Browne, M. J., & Hoyt, R. E. (2000). The demand for flood insurance: empirical
evidence. Journal of Risk and Uncertainty, 20(3), 291-306.
Brown, J. R. (2007). Rational and behavioral perspectives on the role of annuities in retirement
planning (No. w13537). National Bureau of Economic Research.
92

Bron, J. R., Kling, J. R., Mllainathan, S., & Wrobel, M. V. (2008). Wh dont people insre
late-life consumption? A framing explanation of the under-annuitization
puzzle. American Economic Review, 98(2), 304-09.
Buda, R., & Zhang, Y. (2000). Consumer product evaluation: the interactive effect of message
framing, presentation order, and source credibility. Journal of Product & Brand
Management, 9(4), 229-242.
Burton, S. (1990). The framing of purchase for services. Journal of Services Marketing, 4(4), 55-
67.
Camerer, C. F. (2004). Prospect theory in the wild: Evidence from the field. In C. F. Camerer, G.
Loewenstein, & M. Rabin (Eds.), Advances in behavioral economics (pp. 148-161).
Princeton, NJ: Princeton University Press.
Cherubini, P., Rumiati, R., Rossi, D., Nigro, F., & Calabrò, A. (2005). Improving Attitudes
Toward Prostate Examinations by Loss ‐Framed Appeals. Journal of Applied Social
Psychology, 35(4), 732-742.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Lawrence Erlbaum.
Coombs, C. H. (1975). Portfolio theory and the measurement of risk. In M. F. Kaplan & S.
Schwartz (Eds.), Human judgement and decision (pp. 63-68). New York: Academic
Press.
Coombs, C. H., & Bowen, J. N. (1971). A test of VE-theories of risk and the effect of the central
limit theorem. Acta Psychologica, 35(1), 15-28.
Coombs, C. H., & Huang, L. C. (1970). Polynomial psychophysics of risk. Journal of
Mathematical Psychology, 7(2), 317-338.
93

Coombs, C. H., & Pruitt, D. G. (1960). Components of risk in decision making: Probability and
variance preferences. Journal of Experimental Psychology, 60(5), 265.
Dahl, K. A., Spanger-Siegfried, E., Caldas, A., & Udvardy, S. (2017). Effective inundation of
continental United States communities with 21st century sea level rise. Elementa Science
of the Anthropocene, 5, 37.
Dash, N., & Gladwin, H. (2007). Evacuation decision making and behavioral responses:
Individual and household. Natural Hazards Review, 8(3), 69-77.
Davis, M. S. (1989). Living along the fault line: An update on earthquake awareness and
preparedness in Southern California. Urban Resources, 5, 8-14.
Detweiler, J. B., Bedell, B. T., Salovey, P., Pronin, E., & Rothman, A. J. (1999). Message
framing and sunscreen use: gain-framed messages motivate beach-goers. Health
Psychology, 18(2), 189.
Dillon, R. L., & Tinsley, C. H. (2008). How near-misses influence decision making under risk: A
missed opportunity for learning. Management Science, 54(8), 1425-1440.
Dillon, R. L., Tinsley, C. H., & Cronin, M. (2011). Why near ‐miss events can decrease an
individual's protective response to hurricanes. Risk Analysis, 31(3), 440-449.
Doyle, J. K., McClelland, G. H., Schulze, W. D., Elliott, S. R., & Russell, G. W. (1991).
Protective responses to household risk: A case study of radon mitigation. Risk
Analysis, 11(1), 121-134.
Edwards, M. L. (1993). Social location and self-protective behavior: Implications for earthquake
preparedness. International Journal of Mass Emergencies and Disasters, 11(3), 293-303.
94

Elder, K., Xirasagar, S., Miller, N., Bowen, S. A., Glover, S., & Piper, C. (2007). African
Americans decisions not to eacate Ne Orleans before Hrricane Katrina: A
qualitative study. American Journal of Public Health, 97(Supplement_1), S124-S129.
Elsner, J. B. (2003). Tracking hurricanes. Bulletin of the American Meteorological
Society, 84(3), 353-356.
Evans, J. S. B. (2003). In two minds: dual-process accounts of reasoning. Trends in Cognitive
Sciences, 7(10), 454-459.
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.
Flynn, J., Slovic, P., & Mertz, C. K. (1994). Gender, race, and perception of environmental
health risks. Risk analysis, 14(6), 1101-1108.
Fothergill, A. (1996). Gender, risk, and disaster. International Journal of Mass Emergencies and
Disasters, 14(1), 33-56.
Fothergill, A., & Peek, L. A. (2004). Poverty and disasters in the United States: A review of
recent sociological findings. Natural Hazards, 32(1), 89-110.
Gallagher, K. M., & Updegraff, J. A. (2011). Health message framing effects on attitudes,
intentions, and behavior: a meta-analytic review. Annals of Behavioral Medicine, 43(1),
101-116.
Garfield, L. (2018, July 23). Flood-prone US states have billions of dollars of property at stake in
the next 15 years - here's where it will be worst. Retrieved from
https://www.businessinsider.com/us-states-climate-change-flooding-risk-2018-7
Ge, Y., Peacock, W. G., & Lindell, M. K. (2011). Florida hoseholds epected responses to
hurricane hazard mitigation incentives. Risk Analysis, 31(10), 1676-1691.
95

Gigerenzer, G., & Gaissmaier, W. (2011). Heuristic decision making. Annual Review of
Psychology, 62, 451-482.
Gintis, H. (2000). Game theory evolving: A problem-centered introduction to modeling strategic
behavior. Princeton University Press.
Golden, J. H., & Adams, C. R. (2000). The tornado problem: Forecast, warning, and
response. Natural Hazards Review, 1(2), 107-118.
Grace, M. F., Klein, R. W., & Liu, Z. J. (2005). Increased hurricane risk and insurance market
responses. Journal of Insurance Regulation, 24(2), 3-32.
Grinblatt, M., & Han, B. (2005). Prospect theory, mental accounting, and momentum. Journal of
Financial Eeconomics, 78(2), 311-339.
Grothmann, T., & Reusswig, F. (2006). People at risk of flooding: why some residents take
precautionary action while others do not. Natural Hazards, 38(1-2), 101-120.
Gu, D. S., Wu, G., Wu, Z. S., & Wu, Y. F. (2010). Confinement effectiveness of FRP in
retrofitting circular concrete columns under simulated seismic load. Journal of
Composites for Construction, 14(5), 531-540.
Güneyisi, E. M., & Altay, G. (2008). Seismic fragility assessment of effectiveness of viscous
dampers in R/C buildings under scenario earthquakes. Structural Safety, 30(5), 461-480.
Handel, B. R. (2013). Adverse selection and inertia in health insurance markets: When nudging
hurts. American Economic Review, 103(7), 2643-82.
Hasseldine, J., & Hite, P. A. (2003). Framing, gender and tax compliance. Journal of Economic
Psychology, 24(4), 517-533.
Hausman, D. M., & Welch, B. (2010). Debate: To nudge or not to nudge. Journal of Political
Philosophy, 18(1), 123-136.
96

Henrich, L., McClure, J., & Crozier, M. (2015). Effects of risk framing on earthquake risk
perception: life-time frequencies enhance recognition of the risk. International Journal of
Disaster Risk Reduction, 13, 145-150.
Hoekstra, S., Klockow, K., Riley, R., Brotzge, J., Brooks, H., & Erickson, S. (2011). A
preliminary look at the social perspective of warn-on-forecast: Preferred tornado warning
lead time and the general pblics perceptions of eather risks. Weather, Climate, and
Society, 3(2), 128-140.
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.
Hurnen, F., & McClure, J. (1997). The effect of increased earthquake knowledge on perceived
preventability of earthquake damage. Australas. J. Disaster Trauma Stud., (3).
Ito, J., & Marsella, S. (2013, September). Modeling framing effects: Comparing an appraisal-
based model with existing models. In 2013 Humaine Association Conference on Affective
Computing and Intelligent Interaction (pp. 381-386). IEEE.
Janis, I. L., & Feshbach, S. (1954). Personality differences associated with responsiveness to
fear ‐arousing communications 1. Journal of Personality, 23(2), 154-166.
Jarrell, J., Mayfield, M., Rappaport, E., & Landsea, C. (2001). The deadliest, costliest, and most
intense United States hurricanes from 1900 to 2000 (and other frequently requested
hurricane facts) (pp. 1-29, Tech.). Miami, FL: NOAA Technical Memorandum NWS
TPC-3.
Jenner, E. A., Jones, F., Fletcher, B. C., Miller, L., & Scott, G. M. (2005). Hand hygiene posters:
motivators or mixed messages? Journal of Hospital Infection, 60(3), 218-225.
97

Jia, J., & Dyer, J. S. (1997) Risk-Value Theory. (Risk-Value Study Series Working Paper 94/95-
3-4). Graduate School of Business, University of Texas at Austin.
Jia, J., Dyer, J. S., & Butler, J. C. (1999). Measures of perceived risk. Management
Science, 45(4), 519-532.
Johnson, E. J., Hershey, J., Meszaros, J., & Kunreuther, H. (1993). Framing, probability
distortions, and insurance decisions. Journal of Risk and Uncertainty, 7(1), 35-51.
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.
Kahneman, D., & Egan, P. (2011). Thinking, fast and slow (Vol. 1). New York: Farrar, Straus
and Giroux.
Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute substitution in
intuitive judgment. In T. Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics and
biases (pp. 49 81). New York: Cambridge University Press.
Kahneman, D., & Tversky, A. (1979). Prospect Theory. An Analysis of Decision Making Under
Risk. Econometrica, 47, 263-291. doi:10.21236/ada045771
Keller, L. R., Sarin, R. K., & Weber, M. (1986). Empirical investigation of some properties of
the perceived riskiness of gambles. Organizational Behavior and Human Decision
Processes, 38(1), 114-130.
Koehler, J. J. (1996). The base rate fallacy reconsidered: Descriptive, normative, and
methodological challenges. Behavioral and Brain Sciences, 19(1), 1-17.
98

Krishnamurthy, P., Carter, P., & Blair, E. (2001). Attribute framing and goal framing effects in
health decisions. Organizational Behavior and Human Decision Processes, 85(2), 382-
399.
Kühberger, A. (1998). The influence of framing on risky decisions: A meta-
analysis. Organizational Behavior and Human Decision Processes, 75(1), 23-55.
Kunreuther, H. (1996). Mitigating disaster losses through insurance. Journal of Risk and
Uncertainty, 12(2-3), 171-187.
Kunreuther, H., & Kleffner, A. E. (1992). Should earthquake mitigation measures be voluntary
or required?. Journal of Regulatory Economics, 4(4), 321-333.
Kunreuther, H., Meyer, R., & Michel-Kerjan, E. (2013). Overcoming decision biases to reduce
losses from natural catastrophes (E. Shafir, Ed.). In The behavioral foundations of public
policy (pp. 398-415). Princeton, NJ: Princeton University Press.
Landsea, C. (2018, August 1). TCFAQ E19) How many direct hits by hurricanes of various
categories. Retrieved April 23, 2019, from
https://www.aoml.noaa.gov/hrd/tcfaq/E19.html
Lang, J. T., & Hallman, W. K. (2005). Who does the public trust? The case of genetically
modified food in the United States. Risk Analysis, 25(5), 1241-1252.
Levin, I. P., Gaeth, G. J., Schreiber, J., & Lauriola, M. (2002). A new look at framing effects:
Distribution of effect sizes, individual differences, and independence of types of
effects. Organizational Behavior and Human Decision Processes, 88(1), 411-429.
Lichtenstein, S., & Fischhoff, B. (1977). Do those who know more also know more about how
much they know? Organizational Behavior and Human Performance, 20(2), 159-183.
99

Lindell, M. K., & Hwang, S. N. (2008). Households' perceived personal risk and responses in a
multihazard environment. Risk Analysis, 28(2), 539-556.
Lindell, M. K., Lu, J. C., & Prater, C. S. (2005). Household decision making and evacuation in
response to Hurricane Lili. Natural Hazards Review, 6(4), 171-179.
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.
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. (2002). Risk Area Residents Perceptions and Adoption of
Seismic Hazard Adjustments 1. Journal of Applied Social Psychology, 32(11), 2377-
2392.
Lindell, M. K., & Whitney, D. J. (2000). Correlates of household seismic hazard adjustment
adoption. Risk Analysis, 20(1), 13-26.
List, J. A. (2004). Neoclassical theory versus prospect theory: Evidence from the
marketplace. Econometrica, 72(2), 615-625.
Luce, R. D. (1980). Several possible measures of risk. Theory and Decision, 12(3), 217-228.
Marteau, T. M., Ogilvie, D., Roland, M., Suhrcke, M., & Kelly, M. P. (2011). Judging nudging:
can nudging improve population health? BMJ: British Medical Journal (Online), 342.
100

Marti, M., Stauffacher, M., Matthes, J., & Wiemer, S. (2018). Communicating earthquake
preparedness: The influence of induced mood, perceived risk, and gain or loss frames on
homeoners attitdes toard general precationar measres for earthqakes. Risk
Analysis, 38(4), 710-723.
Martínez-Álvarez, F., Reyes, J., Morales-Esteban, A., & Rubio-Escudero, C. (2013).
Determining the best set of seismicity indicators to predict earthquakes. Two case
studies: Chile and the Iberian Peninsula. Knowledge-Based Systems, 50, 198-210.
Merz, B., Kreibich, H., Schwarze, R., & Thieken, A. (2010). Review article" Assessment of
economic flood damage". Natural Hazards and Earth System Sciences, 10(8), 1697-
1724.
McCann Jr, M. W., & Shah, H. C. (1979). Determining strong-motion duration of
earthquakes. Bulletin of the Seismological Society of America, 69(4), 1253-1265.
McClure, J., Walkey, F., & Allen, M. (1999). When earthquake damage is seen as preventable:
Attributions, locus of control and attitudes to risk. Applied Psychology, 48(2), 239-256.
McClure, J., White, J., & Sibley, C. G. (2009). Framing effects on preparation intentions:
distinguishing actions and outcomes. Disaster Prevention and Management: An
International Journal, 18(2), 187-199.
McIvor, D., & Paton, D. (2007). Preparing for natural hazards: normative and attitudinal
influences. Disaster Prevention and Management: An International Journal, 16(1), 79-
88.
Meyer, R. J. (2006). Why We Under-Prepare for Hazards. In R. J. Daniels, D. F. Kettl, & H.
Kunreuther (Eds.), On Risk and Disaster: Lessons from Hurricane Katrina (pp. 153-173).
Philadelphia, PA, Pennsylvania: University of Pennsylvania Press.
101

Meyer R., & Kunreuther, H. C. (2017). The ostrich paradox: Why we underprepare for disasters.
Philadelphia, Pennsylvania: Wharton digital press.
Michel-Kerjan, E. O. (2010). Catastrophe economics: the national flood insurance
program. Journal of Economic Perspectives, 24(4), 165-86.
Michel-Kerjan, E., & Kunreuther, H. (2011). Redesigning flood insurance. Science, 333(6041),
408-409.
Mileti, D. S., & Darlington, J. D. (1997). The role of searching in shaping reactions to
earthquake risk information. Social Problems, 44(1), 89-103.
Mileti, D. S., & Fitzpatrick, C. (1992). The causal sequence of risk communication in the
Parkfield earthquake prediction experiment. Risk Analysis, 12(3), 393-400.
Mileti, D. S., & Peek, L. (2000). The social psychology of public response to warnings of a
nuclear power plant accident. Journal of Hazardous Materials, 75(2-3), 181-194.
Mileti, D. S., & Sorensen, J. H. (1987). Why people take precautions against natural
disasters. Taking care: Why people take precautions, 296-320.
Miller, R. L., & Mulligan, R. D. (2002). Terror management: The effects of mortality salience
and locus of control on risk-taking behaviors. Personality and Individual
Differences, 33(7), 1203-1214.
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.
Mulilis, J. P., Duval, T. S., & Lippa, R. (1990). The effects of a large destructive local
earthquake on earthquake preparedness as assessed by an earthquake preparedness
scale. Natural Hazards, 3(4), 357-371.
102

Mulilis, J. P., & Duval, T. S. (1995). Negative threat appeals and earthquake preparedness: a
person ‐relative ‐to ‐event (PrE) model of coping with threat. Journal of Applied Social
Psychology, 25(15), 1319-1339.
Mulilis, J. P., & Duval, T. S. (1997). The PrE model of coping and tornado preparedness:
Moderating effects of responsibility. Journal of Applied Social Psychology, 27(19), 1750-
1766.
Palan, S., & Schitter, C. (2018). Prolific. ac A subject pool for online experiments. Journal of
Behavioral and Experimental Finance, 17, 22-27.
Palm, R., & Hodgson, M. (1992). Earthquake insurance: mandated disclosure and homeowner
response in California. Annals of the Association of American Geographers, 82(2), 207-
222.
Paton, D., Smith, L., & Johnston, D. (2000). Volcanic hazards: risk perception and
preparedness. New Zealand Journal of Psychology, 29(2), 86-91.
Paton, D. (2003). Disaster preparedness: a social-cognitive perspective. Disaster Prevention and
Management: An International Journal, 12(3), 210-216.
Paul, B. K., Emerson, L., Brock, V. T., & Csiki, S. (2013). Public response to tornado warnings:
A comparative study of the May 4, 2003, Tornadoes in Kansas, Missouri, and Tennessee
(Rep. No. 165). Boulder, CO: Natural Hazards Research and Applications Information
Center, University of Colorado Boulder.
Peacock, W. G. (2003). Hurricane mitigation status and factors influencing mitigation status
among Floridas 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.
103

Petersen, M.D., Moschetti, M.P., Powers, P.M., Mueller, C.S., Haller, K.M., Frankel, A.D.,
Zeng, Yuehua, Rezaeian, Sanaz, Harmsen, S.C., Boyd, O.S., Field, Ned, Chen, Rui,
Rukstales, K.S., Luco, Nico, Wheeler, R.L., Williams, R.A., and Olsen, A.H., 2014,
Documentation for the 2014 update of the United States national seismic hazard maps:
U.S. Geological Survey Open-File Report 2014 1091, 243
p., https://dx.doi.org/10.3133/ofr20141091.
Picard, P. (2008). Natural disaster insurance and the equity ‐efficiency trade ‐off. Journal of Risk
and Insurance, 75(1), 17-38.
Pijawka, K. D., & Mushkatel, A. H. (1991). Public opposition to the siting of the high ‐level
nuclear waste repository: The importance of trust. Review of Policy Research, 10(4), 180-
194.
Pollatsek, A., & Tversky, A. (1970). A theory of risk. Journal of Mathematical Psychology, 7(3),
540 553. https://doi.org/10.1016/0022-2496(70)90039-8
Price, C., Yair, Y., Mugnai, A., Lagouvardos, K., Llasat, M. C., Michaelides, S., ... & Harats, N.
(2011). The FLASH Project: using lightning data to better understand and predict flash
floods. Environmental Science & Policy, 14(7), 898-911.
Raiffa, H., & Schlaifer, R. (1970). Additive Utility, Opportunity Loss, and the Value of
Information. In Applied statistical decision theory (5th ed., pp. 79-92). Cambridge, MA:
Division of Research, Graduate School of Business Administration, Harvard University.
Raschky, P. A., & Weck-Hannemann, H. (2007). Charity hazard A real hazard to natural
disaster insurance? Environmental Hazards, 7(4), 321-329.
Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change1. The
Journal of Psychology, 91(1), 93-114.
104

Rothman, A. J., Martino, S. C., Bedell, B. T., Detweiler, J. B., & Salovey, P. (1999). The
systematic influence of gain-and loss-framed messages on interest in and use of different
types of health behavior. Personality and Social Psychology Bulletin, 25(11), 1355-1369.
Rothman, A. J., Bartels, R. D., Wlaschin, J., & Salovey, P. (2006). The strategic use of gain-and
loss-framed messages to promote healthy behavior: How theory can inform
practice. Journal of Communication, 56(s1), S202-S220.
Rotter, J. B. (1966). Generalized expectancies for internal versus external control of
reinforcement. Psychological Monographs: General and Applied, 80(1), 1-28.
Rouse, S. V. (2015). A reliability analysis of Mechanical Turk data. Computers in Human
Behavior, 43, 304-307.
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.
Rüstemli, A., & Karanci, A. N. (1999). Correlates of earthquake cognitions and preparedness
behavior in a victimized population. The Journal of Social Psychology, 139(1), 91-101.
Schapira, M. M., Davids, S. L., McAuliffe, T. L., & Nattinger, A. B. (2004). Agreement between
scales in the measurement of breast cancer risk perceptions. Risk Analysis, 24(3), 665-
673.
Scherer, K. R. (1982). Emotion as a process: Function, origin and regulation. Social Science
Information, 21(4 5), 555 570. https://doi.org/10.1177/053901882021004004
Schwartz, L. M., Woloshin, S., Black, W. C., & Welch, H. G. (1997). The role of numeracy in
understanding the benefit of screening mammography. Annals of Internal
Medicine, 127(11), 966-972.
105

Shaw, R., Kurita, T., Nakamura, A., Kodama, M., & Colombage, S. R. (2006). Tsunami public
awareness and the disaster management system of Sri Lanka. Disaster Prevention And
Management: An International Journal, 15(1), 92-110.
Shead, N. W., & Hodgins, D. C. (2009). Probability discounting of gains and losses: Implications
for risk attitudes and impulsivity. Journal of the Experimental Analysis of
Behavior, 92(1), 1-16.
Sherman, D. K., Mann, T., & Updegraff, J. A. (2006). Approach/avoidance motivation, message
framing, and health behavior: Understanding the congruency effect. Motivation and
Emotion, 30(2), 164-168.
Simmons, J. P., Nelson, L. D., Galak, J., & Frederick, S. (2010). Intuitive biases in choice versus
estimation: Implications for the wisdom of crowds. Journal of Consumer
Research, 38(1), 1-15.
Slovic, P. (1967). The relative influence of probabilities and payoffs upon perceived risk of a
gamble. Psychonomic Science, 9(4), 223-224.
Smith, K. (2013). Environmental hazards: Assessing risk and reducing disaster. Abingdon,
Oxon, UK: Routledge.
Smith, A. (2017). 2016: A historic year for billion-dollar weather and climate disasters in U.S.
Retrieved September 04, 2017, from https://www.climate.gov/news-
features/blogs/beyond-data/2016-historic-year-billion-dollar-weather-and-climate-
disasters-us
Sorensen, J. H., Vogt, B. M., & Mileti, D. S. (1987). Evacuation: an assessment of planning and
research (No. ORNL-6376). Oak Ridge National Lab., TN (USA).
106

Spence, A., & Pidgeon, N. (2010). Framing and communicating climate change: The effects of
distance and outcome frame manipulations. Global Environmental Change, 20(4), 656-
667.
Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the
rationality debate? Behavioral and Brain Sciences, 23(5), 645-665.
Stanovich, K. E., & West, R. F. (2008). On the relative independence of thinking biases and
cognitive ability. Journal of Personality and Social Psychology, 94(4), 672.
Stern, P. C., Dietz, T., & Kalof, L. (1993). Value orientations, gender, and environmental
concern. Environment and Behavior, 25(5), 322-348.
Tanaka, K. (2005). The impact of disaster education on public preparation and mitigation for
earthquakes: a cross-country comparison between Fukui, Japan and the San Francisco
Bay Area, California, USA. Applied Geography, 25(3), 201-225.
Terpstra, T., & Lindell, M. K. (2013). Citiens perceptions of flood haard adjst ments: an
application of the protective action decision model. Environment and Behavior, 45(8),
993-1018.
Thaler, R. H. (2008). Nudge: Improving decisions about health, wealth, and happiness.
Thaler, R. H., & Sunstein, C. R. (2009). Nudge: Improving decisions about health, wealth, and
happiness. Penguin.
Tierney, K. J., Lindell, M. K., & Perry, R. W. (2001). Facing the unexpected: Disaster
preparedness and response in the United States. Washington, D.C.: Joseph Henry Press.
Tinsley, C. H., Dillon, R. L., & Cronin, M. A. (2012). How near-miss events amplify or attenuate
risky decision making. Management Science, 58(9), 1596-1613.
107

Toplak, M. E., West, R. F., & Stanovich, K. E. (2014). Assessing miserly information
processing: An expansion of the Cognitive Reflection Test. Thinking & Reasoning, 20(2),
147-168.
Tsonos, A. G. (2008). Effectiveness of CFRP-jackets and RC-jackets in post-earthquake and pre-
earthquake retrofitting of beam column sub assemblages. Engineering Structures, 30(3),
777-793.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of
choice. Science, 211(4481), 453-458.
Tversky, A., & Kahneman, D. (1986). Rational choice and the framing of decisions. Journal of
Business, S251-S278.
Tversky, A., & Koehler, D. J. (1994). Support theory: A nonextensional representation of
subjective probability. Psychological Review, 101(4), 547.
USGS (2020) Earthquake Statistics. Retrived from https://www.usgs.gov/natural-
hazards/earthquake-hazards/lists-maps-and-statistics
Vinnell, L. J., McClure, J., & Milfont, T. L. (2017). Do framing messages increase support for
earthquake legislation? Disaster Prevention and Management: An International
Journal, 26(1), 28-40.
Weber, E. U. (1984). Combine and conquer: A joint application of conjoint and functional
approaches to the problem of risk measurement. Journal of Experimental Psychology:
Human Perception and Performance, 10(2), 179.
Weber, E. U., Anderson, C. J., & Birnbaum, M. H. (1992). A theory of perceived risk and
attractiveness. Organizational Behavior and Human Decision Processes, 52(3), 492-523.
108

Weber, E. U., & Bottom, W. P. (1989). Axiomatic measures of perceived risk: Some tests and
extensions. Journal of Behavioral Decision Making, 2(2), 113-131.
Weber, E. U., & Bottom, W. P. (1990). An empirical evaluation of the transitivity, monotonicity,
accounting, and conjoint axioms for perceived risk. Organizational Behavior and Human
Decision Processes, 45(2), 253-275.
Wiener, J. L., Gentry, J. W., & Miller, R. K. (1986). The framing of the insurance purchase
decision. Advances in Consumer Research, 13(1), 251-256.
Woloshin, S., Schwartz, L. M., Byram, S., Fischhoff, B., & Welch, H. G. (2000). A new scale for
assessing perceptions of chance: a validation study. Medical Decision Making, 20(3),
298-307.
Zhao, M., Rosoff, H., & John, R. S. (2019). Media disaster reporting effects on public risk
perception and response to escalating tornado warnings: A natural experiment. Risk
analysis, 39(3), 535-552.

109

Appendix A: Decision Problems for Experiments 1
1. Hurricanes

❖ Risk Mitigation: invest in storm shutters to secure your home for the upcoming hurricane
season

Gain frame:
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Cent er has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your city in the upcoming hurricane season (June 1st to Nov
30th). State real estate regulations require that the seller deposit $10,000 in escrow to pay for any
damages incurred from the upcoming hurricane season. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after hurricane season officially ends on Nov 30th, 2020. Since your home is
located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install the storm shutters using part of the $10,000 deposit; you will receive the
remaining $9,000 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 9 0% chance that no hurricane causes wind damage to your property during the 2020
hurricane season, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that a hrricane does case ind damage to or propert dring the
2020 hurricane season, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $9000 of the deposit back after
November 30, regardless any hurricane activity.
● I choose to not install the storm shutters, and receive the entire $10,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

Loss frame:
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Center has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
110

the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your city in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install storm shutters for a total cost of $1,000, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 90% chance that no hrricane dring the 2020 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 10% chance t hat a hurricane during the 2020 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1000 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

❖ Insurance purchase: whether to buy hurricane insurance to secure your property for the next
hurricane season

Gain frame:
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Center has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your city in the upcoming hurricane season. State real estate
regulations require that the seller deposit $10,000 in escrow to pay for any damages incurred
from the upcoming hurricane season. The deposit will be returned to you in full after your
propert has been transferred to the ne oners name and or propert remains in good
condition after hurricane season officially ends on Nov 30th, 2020. Since your home is located on
high ground, your home will be safe from flooding during hurricane season. Home owners who
face the risks of hurricane winds typically purchase hurricane insurance, which guarantee your
home will be covered if any wind damage occurred from a hurricane. (Note that investing in
hurricane insurance will not affect the sale price of the home.) Unfortunately, your home does
not have hurricane insurance as of now. You have two options:

111

You may invest in hurricane insurance for the upcoming hurricane season using part of the
$10,000 deposit; you will receive the remaining $9,000 of your deposit back for sure at the end
of the 2020 hurricane season.

You may choose not to invest in hurricane insurance for the upcoming hurricane season, in
which you will accept the following possibility:

1. Theres a 90% chance that no hrricane cases ind damage to or propert dring the 2020
hurricane season, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that a hurricane does cause wind damage to your property during the
2020 hurricane season, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to invest in hurricane insurance and receive $9000 of the deposit back after
November 30, regardless any hurricane activity.
● I choose to not invest in hurricane insurance, and receive the entire $10,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

Loss frame:
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Center has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your city in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Home
owners who face the risks of hurricane winds typically purchase hurricane insurance, which
guarantee your home will be covered if any wind damage occurred from a hurricane. (Note that
the shutters will not affect the sale price of the home.) Unfortunately, your home does not have
storm shutters. You have two options:

You may buy hurricane insurance for the upcoming hurricane season for a total cost of $1,000,
and you will not incur any cost related to hurricane damage during the 2020 hurricane season for
sure.

You may choose not to buy hurricane insurance, in which you will accept the following
possibility:
1. Theres a 90% chance that no hrricane dring the 2020 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 10% chance that a hrricane dring the 2020 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to spend $1000 and buy hurricane insurance.
112

● I choose to not buy hurricane insurance and either pay nothing extra if there is no damage
to my home or pay $10,000 to repair my home before it can be sold.

2. Earthquakes
❖ Risk Mitigation: retrofit your home to prepare for earthquakes

Gain frame:
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $10,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or prope rty remains
in good condition. Homes located in earthquake prone areas typically are retrofitted to prevent
damages from earthquakes. (Note that retrofitting will not affect the sale price of the home.)
Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home using part of the $10,000 deposit; you will receive the remaining
$9,000 of your deposit back for sure after 12 months.

You may choose not to retrofit, in which you will accept the following possibility:

1. Theres a 90% chance that no earthquake causes damage to your property during the 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that an earthquake does cause damage to your property during the 12
months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to retrofit and receive $9,000 of the deposit back after 12 months, regardless any
seismic activity.
● I choose to not retrofit, and receive the entire $10,000 deposit after 12 months if there is
no earthquake damage to my home, or receive nothing if there is earthquake damage to
my home within 12 months.

Loss frame:
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkeley. The Southern California Earthqake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
113

within the next 12 months. Homes located in earthquake prone areas typically are retrofitted to
prevent damages from earthquakes. (Note that retrofitting will not affect the sale price of the
home.) Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home for a total cost of $1,000, and you will not incur any damage to your
property for sure.

You may choose not to retrofit, in which you will accept the following possibility:
1. Theres a 90% chance that no earthquakes during the next 12 months causes damage to your
home and you will lose nothing.
2. Theres a 10% chance that an earthquake occurs within the next 12 months and damages your
home; the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and retrofit my home.
● I choose to not retrofit and either pay nothing extra if there is no damage to my home or
pay $10,000 to repair my home before it can be sold.

❖ Insurance purchase: whether to buy earthquake insurance to secure your property for the next
12 months

Gain frame:
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkeley. The Southern California Earthquake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $10,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after 12 months. Home owners in earthquake prone areas typically invest in
earthquake insurance to prevent incurring damages from earthquakes. (Note that investing in
earthquake insurance will not affect the sale price of the home.) Unfortunately, your home does
not have earthquake insurance as of now. You have two options:

You may invest in earthquake insurance using part of the $10,000 deposit; you will receive the
remaining $9,000 of your deposit back for sure after 12 months.

You may choose not to invest in earthquake insurance, in which you will accept the following
possibility:

1. Theres a 90% chance that no earthquake causes damage to your property during the next 12
months, and your $10,000 deposit will be returned to you in full.
114

2. Theres a 10% chance that an earthquake does cause wind damage to your property during the
next 12 months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to invest in earthquake insurance and receive $9,000 of the deposit back after 12
months, regardless any seismic activity.
● I choose to not invest in earthquake insurance, and receive the entire $10,000 deposit
after 12 months if there is no earthquake damage to my home, or receive nothing if there
is earthquake damage to my home within the next 12 months.

Loss frame:
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkeley. The Southern California Earthqake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. Home owners in earthquake prone areas typically invest in
earthquake insurance to prevent damages from earthquakes. (Note that earthquake insurance will
not affect the sale price of the home.) Unfortunately, your home does not have earthquake
insurance as of now.

You have two options:

You may pay for earthquake insurance for a total cost of $1,000, and you will not incur any cost
related to earthquake damages during the next 12 months for sure.

You may choose not to pay for earthquake insurance, in which you will accept the following
possibility:
1. Theres a 90% chance that no earthquake during the next 12 months causes damage to your
home and you will lose nothing.
2. Theres a 10% chance that an earthquake during the next 12 months damages your home; the
cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and get earthquake insurance.
● I choose to not pay for earthquake insurance and either pay nothing extra if there is no
damage to my home or pay $10,000 to repair my home before it can be sold.

3. Flood

❖ Risk Mitigation: invest in sump pump to prevent your home from water damage

Gain frame:
115

As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell your home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $10,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the ne oners name and or propert remains in good condition after
12 months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for your home using part of the $10,000 deposit; you will receive
the remaining $9,000 of your deposit back for sure after 12 months.

You may choose not to install sump bumps, in which you will accept the following possibility:

1. Theres an 90% chance that no flood causes damage to your property during the next 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that an earthquake does cause damage to your property during the next
12 months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the sump bump and receive $9,000 of the deposit back after 12 months,
regardless any flooding activity.
● I choose to not install the sump bump, and receive the entire $10,000 deposit after 12
months if there is no flooding damage to my home, or receive nothing if there is flooding
damage to my home within the next 12 months.

Loss frame:
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for a total cost of $1,000, and you will not incur any damage to
your property for sure.

You may choose not to install sump pumps, in which you will accept the following possibility:
1. Theres an 90% chance that no flood during the next 12 months causes damage to your home
and you will lose nothing.
2. Theres a 10% chance that a flood occurs within the next 12 months and damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and install sump pumps for my home.
116

● I choose to not install sump pumps and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

❖ Insurance purchase: whether to buy earthquake insurance to secure your property for the next
12 months

Gain frame:
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $10,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the ne oners name and or propert remains in good condition after
12 months. Home owners in flood prone areas typically purchase flood insurance to protect
properties. (Note that investing in flood insurance will not affect the sale price of the home.)
Unfortunately, your home is not insured for flood as of now. You have two options:

You may invest in flood insurance using part of the $10,000 deposit; you will receive the
remaining $9,000 of your deposit back for sure after 12 months.

You may choose not to invest in flood insurance, in which you will accept the following
possibility:

1. Theres an 90% chance that no flood causes damage to your property during the next 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that a flood does cause water damage to your property during the next
12 months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to invest in flood insurance and receive $9,000 of the deposit back after 12
months, regardless any flooding activity.
● I choose to not invest in flood insurance, and receive the entire $10,000 deposit after 12
months if there is no flood damage to my home, or receive nothing if there is flood
damage to my home within the next 12 months.

Loss frame:
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell your home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Home owners in flood prone areas typically purchase flood insurance to protect
properties. (Note that investing in flood insurance will not affect the sale price of the home.)
Unfortunately, your home is not insured for flood as of now. You have two options:

You may pay for flood insurance for a total cost of $1,000, and you will not incur any cost
related to flood damages during the next 12 months for sure.

117

You may choose not to pay for flood insurance, in which you will accept the following
possibility:
1. Theres an 90% chance that no flood during the next 12 months causes damage to your home
and you will lose nothing.
2. Theres a 10% chance that a flood during the next 12 months damages your home; the cost to
repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and get flood insurance.
● I choose to not pay for flood insurance and either pay nothing extra if there is no damage
to my home or pay $10,000 to repair my home before it can be sold.

118

Appendix B: Decision Vignettes for Experiment 2
Participants will be randomized into one of the eight conditions, and make decisions for
each of the three disasters. For hurricane, the decision is to invest in storm shutters to secure your
home for the next upcoming hurricane season or not. For earthquake, the decision is to invest in
retrofitting to secure your home for the next 12 months or not. For flood, the decision is to install
sump bump to protect your home from water damage for the next 12 months or not. Effect of EV
is tested by manipulating the expected value of outcomes, and effect of disaster probabilities is
tested by manipulating the variance of the gamble probabilities. Experiment 2 will focus solely
on risk mitigation.

Decision Problems
1. High EV*gain frame
A. Hurricane
As a homeowner living in a high risk area for hurricane, due to a recent job re-location
o mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your area in the upcoming hurricane season (June 1st to Nov
30th). State real estate regulations require that the seller deposit $5,000 in escrow to pay for any
damages incurred from the upcoming hurricane season. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after hurricane season officially ends on Nov 30th, 2020. Since your home is
located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install the storm shutters using part of the $5,000 deposit; you will receive the
remaining $4,500 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 90% chance that no hrricane cases ind damage to or propert dring the 2020
hurricane season, and your $5,000 deposit will be returned to you in full.
2. Theres a 10% chance that a hrricane does case ind damage to or propert dring the
2020 hurricane season, and none of the $5,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $4,500 of the deposit back after
November 30, regardless any hurricane activity.
119

● I choose to not install the storm shutters, and receive the entire $5,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $5,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
full after or propert has been transferred to the ne oners name and or propert remains
in good condition. Homes located in earthquake prone areas typically are retrofitted to prevent
damages from earthquakes. (Note that retrofitting will not affect the sale price of the home.)
Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home using part of the $5,000 deposit; you will receive the remaining
$4,500 of your deposit back for sure after 12 months.

You may choose not to retrofit, in which you will accept the following possibility:

1. Theres a 90% chance that no earthqake cases damage to or propert dring the 12
months, and your $5,000 deposit will be returned to you in full.
2. Theres a 10% chance tha t an earthquake does cause damage to your property during the 12
months, and none of the $5,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to retrofit and receive $4,500 of the deposit back after 12 months, regardless any
seismic activity.
● I choose to not retrofit, and receive the entire $5,000 deposit after 12 months if there is no
earthquake damage to my home, or receive nothing if there is earthquake damage to my
home within 12 months.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $5,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the new owners name and or propert remains in good condition after
12 months. Homes located in flood prone areas typically have sump pumps to prevent water
120

damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for your home using part of the $5,000 deposit; you will receive
the remaining $4,500 of your deposit back for sure after 12 months.

You may choose not to install sump bumps, in which you will accept the following possibility:

1. Theres an 90% chance that no flood cases damage to or propert dring the net 12
months, and your $5,000 deposit will be returned to you in full.
2. Theres a 10% chance that an earthqake does case damage to or propert dring the net
12 months, and none of the $5,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the sump bump and receive $4,500 of the deposit back after 12 months,
regardless any flooding activity.
● I choose to not install the sump bump, and receive the entire $5,000 deposit after 12
months if there is no flooding damage to my home, or receive nothing if there is flooding
damage to my home within the next 12 months.

2. High EV*loss frame
A. Hurricanes
As a homeowner living in a high risk area for hurricanes, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Climate Prediction Cen ter has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your area in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install storm shutters for a total cost of $500, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 90% chance that no hrricane dring the 20 20 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 10% chance that a hr ricane during the 2020 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $5,000.

Please make a decision based on the information provided above:
121

● I choose to pay $500 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage to
my home or pay $5,000 to repair my home before it can be sold.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 10% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. Homes located in earthquake prone areas typically are retrofitted to
prevent damages from earthquakes. (Note that retrofitting will not affect the sale price of the
home.) Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to retrofit, in which you will accept the following possibility:
1. Theres a 90% chance that no earthqakes dring the net 12 months cases damage to or
home and you will lose nothing.
2. Theres a 10% chance that an earthqake occ rs within the next 12 months and damages your
home; the cost to repair this damage to your home before sale is estimated to be $5,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and retrofit my home.
● I choose to not retrofit and either pay nothing extra if there is no damage to my home or
pay $5,000 to repair my home before it can be sold.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell your home. The National Weather Serice predicted theres a 10% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to install sump pumps, in which you will accept the following possibility:
1. Theres an 90% chance that no flood dring the net 12 months cases damage to or home
and you will lose nothing.
122

2. Theres a 10% chance that a flood occrs ithin the net 12 months and damages or home;
the cost to repair this damage to your home before sale is estimated to be $5,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and install sump pumps for my home.
● I choose to not install sump pumps and either pay nothing extra if there is no damage to
my home or pay $5,000 to repair my home before it can be sold.

3. Low EV* gain frame
A. Hurricane
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Center has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 60% chance a hurricane will hit your area in the upcoming hurricane season (June 1st to Nov
30th). State real estate regulations require that the seller deposit $10,000 in escrow to pay for any
damages incurred from the upcoming hurricane season. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after hurricane season officially ends on Nov 30th, 2020. Since your home is
located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install the storm shutters using part of the $10,000 deposit; you will receive the
remaining $9,000 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 90% chance that no hurricane causes wind damage to your property during the 2020
hurricane season, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that a hurricane does cause wind damage to your property during the
2020 hurricane season, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $9,000 of the deposit back after
November 30, regardless any hurricane activity.
● I choose to not install the storm shutters, and receive the entire $10,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

B. Earthquake
123

As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 60% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $10,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
full after your property has been transferred to the ne oners name and or propert remains
in good condition. Homes located in earthquake prone areas typically are retrofitted to prevent
damages from earthquakes. (Note that retrofitting will not affect the sale price of the home.)
Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home using part of the $10,000 deposit; you will receive the remaining
$9,000 of your deposit back for sure after 12 months.

You may choose not to retrofit, in which you will accept the following possibility:

1. Theres a 90% chance that no earthquake causes damage to your property during the 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that an earthquake does cause damage to your property during the 12
months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to retrofit and receive $9,000 of the deposit back after 12 months, regardless any
seismic activity.
● I choose to not retrofit, and receive the entire $10,000 deposit after 12 months if there is
no earthquake damage to my home, or receive nothing if there is earthquake damage to
my home within 12 months.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 60% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $10,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the ne oners name and your property remains in good condition after
12 months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for your home using part of the $10,000 deposit; you will receive
the remaining $9,000 of your deposit back for sure after 12 months.
124

You may choose not to install sump bumps, in which you will accept the following possibility:

1. Theres an 90% chance that no flood causes damage to your property during the next 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 10% chance that an earthquake does cause damage to your property during the next
12 months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the sump bump and receive $9,000 of the deposit back after 12 months,
regardless any flooding activity.
● I choose to not install the sump bump, and receive the entire $10,000 deposit after 12
months if there is no flooding damage to my home, or receive nothing if there is flooding
damage to my home within the next 12 months.

4. Low EV* loss frame
A. Hurricane
As a homeowner living in a high risk area for hurricane, due to a recent job re-location
o mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your area in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install storm shutters for a total cost of $1,000, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 90% chance that no hurricane during the 2020 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 10% chance that a hurricane during the 2020 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

125

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 60% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. Homes located in earthquake prone areas typically are retrofitted to
prevent damages from earthquakes. (Note that retrofitting will not affect the sale price of the
home.) Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home for a total cost of $1,000, and you will not incur any damage to your
property for sure.

You may choose not to retrofit, in which you will accept the following possibility:
1. Theres a 90% chance that no earthquakes during the next 12 months causes damage to your
home and you will lose nothing.
2. Theres a 10% chance that an earthquake occurs within the next 12 months and damages your
home; the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $1,000 and retrofit my home.
● I choose to not retrofit and either pay nothing extra if there is no damage to my home or
pay $10,000 to repair my home before it can be sold.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 60% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for a total cost of $1,000, and you will not incur any damage to
your property for sure.

You may choose not to install sump pumps, in which you will accept the following possibility:
1. Theres an 90% chance that no flood during the next 12 months causes damage to your home
and you will lose nothing.
2. Theres a 10% chance that a flood occurs within the next 12 months and damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
126

● I choose to pay $1,000 and install sump pumps for my home.
● I choose to not install sump pumps and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

5. Low disaster probability*gain frame
A. Hurricane
As a homeowner living in a high risk area for hurricane, due to a recent job re-location you must
sell or home. The National Weather Serice (NWS)s Climate Prediction Center has reported
that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 5% chance a hurricane will hit your area in the upcoming hurricane season (June 1st to Nov
30th). State real estate regulations require that the seller deposit $10,000 in escrow to pay for any
damages incurred from the upcoming hurricane season. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after hurricane season officially ends on Nov 30th, 2020. Since your home is
located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install the storm shutters using part of the $10,000 deposit; you will receive the
remaining $9,500 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 95% chance that no hrricane cases ind damage to or propert dring the 20 20
hurricane season, and your $10,000 deposit will be returned to you in full.
2. Theres a 5% chance that a hrricane does case ind damage to or propert dring the
2020 hurricane season, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $9500 of the deposit back after
November 30, regardless any hurricane activity.
● I choose to not install the storm shutters, and receive the entire $10,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
127

that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 5% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $10,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
full after your property has been transferred to the ne oners name and or propert remains
in good condition. Homes located in earthquake prone areas typically are retrofitted to prevent
damages from earthquakes. (Note that retrofitting will not affect the sale price of the home.)
Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home using part of the $10,000 deposit; you will receive the remaining
$9,500 of your deposit back for sure after 12 months.

You may choose not to retrofit, in which you will accept the following possibility:

1. Theres a 95% chance that no earthqake cases damage to or propert dring the 12
months, and your $10,000 deposit will be returned to you in full.
2. Theres a 5% chance that an earthqake does ca use damage to your property during the 12
months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to retrofit and receive $9,500 of the deposit back after 12 months, regardless any
seismic activity.
● I choose to not retrofit, and receive the entire $10,000 deposit after 12 months if there is
no earthquake damage to my home, or receive nothing if there is earthquake damage to
my home within 12 months.
C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 5% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $10,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the ne oners name and or propert y remains in good condition after
12 months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for your home using part of the $10,000 deposit; you will receive
the remaining $9,500 of your deposit back for sure after 12 months.

You may choose not to install sump bumps, in which you will accept the following possibility:

1. Theres an 95% chance that no flood cases damage to or propert dring the net 12
months, and your $10,000 deposit will be returned to you in full.
128

2. Theres a 5% chance that an earthqake does case damage to or propert dring the net
12 months, and none of the $10,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the sump bump and receive $9,500 of the deposit back after 12 months,
regardless any flooding activity.
● I choose to not install the sump bump, and receive the entire $10,000 deposit after 12
months if there is no flooding damage to my home, or receive nothing if there is flooding
damage to my home within the next 12 months.

6. Low disaster probability*loss frame
A. Hurricane
As a homeowner living in a high risk area for hurricanes, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Clim ate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 10% chance a hurricane will hit your area in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install storm shutters for a total cost of $500, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 95% chance that no hrricane dring the 20 20 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 5% chance that a hurricane during the 2020 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
129

California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 5% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. Homes located in earthquake prone areas typically are retrofitted to
prevent damages from earthquakes. (Note that retrofitting will not affect the sale price of the
home.) Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to retrofit, in which you will accept the following possibility:
1. Theres a 95% chance that no earthqakes dring the net 12 months cases damage to or
home and you will lose nothing.
2. Theres a 5% chance that an earthqake occrs ithin the net 12 months and damages or
home; the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and retrofit my home.
● I choose to not retrofit and either pay nothing extra if there is no damage to my home or
pay $10,000 to repair my home before it can be sold.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 5% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to install sump pumps, in which you will accept the following possibility:
1. Theres an 95% chance that no flood dring the net 12 months cases damage to or home
and you will lose nothing.
2. Theres a 5% chance that a flood occrs within the next 12 months and damages your home;
the cost to repair this damage to your home before sale is estimated to be $10,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and install sump pumps for my home.
● I choose to not install sump pumps and either pay nothing extra if there is no damage to
my home or pay $10,000 to repair my home before it can be sold.

7. High disaster probability*gain frame
130

A. Hurricane
As a homeowner living in a high risk area for hurricanes, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Mexico and Western Atlantic and an increase of hurricanes in the Pacific. NWS predicted theres
a 50% chance a hurricane will hit your area in the upcoming hurricane season (June 1st to Nov
30th). State real estate regulations require that the seller deposit $1,000 in escrow to pay for any
damages incurred from the upcoming hurricane season. The deposit will be returned to you in
fll after or propert has been transferred to the ne oners name and or propert remains
in good condition after hurricane season officially ends on Nov 30th, 2020. Since your home is
located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install the storm shutters using part of the $1,000 deposit; you will receive the
remaining $500 of your deposit back for sure at the end of the 2020 hurricane season.

You may choose not to install the shutters, in which you will accept the following possibility:

1. Theres a 50% chance that no hrricane cases ind damage to or propert dring the 20 20
hurricane season, and your $1,000 deposit will be returned to you in full.
2. Theres a 50% chance that a hrricane does case ind damage to or propert dring the
2020 hurricane season, and none of the $1,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the storm shutters and receive $500 of the deposit back after November
30, regardless any hurricane activity.
● I choose to not install the storm shutters, and receive the entire $1,000 deposit after
November 30 if there is no wind damage to my home, or receive nothing if there is wind
damage to my home before November 30.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 50% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. State real estate regulations require that the seller deposit $1,000 in
escrow to pay for any damages incurred from earthquakes. The deposit will be returned to you in
full after your property has been transferred to the ne oners name and or propert remains
131

in good condition. Homes located in earthquake prone areas typically are retrofitted to prevent
damages from earthquakes. (Note that retrofitting will not affect the sale price of the home.)
Unfortunately, your home has not been retrofitted. You have two options:

You may retrofit your home using part of the $1,000 deposit; you will receive the remaining
$500 of your deposit back for sure after 12 months.

You may choose not to retrofit, in which you will accept the following possibility:

1. Theres a 50% chance that no earthqake cases damage to or propert dring the 12
months, and your $1,000 deposit will be returned to you in full.
2. Theres a 50% chance that an earthqake does cause damage to your property during the 12
months, and none of the $1,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to retrofit and receive $9,500 of the deposit back after 12 months, regardless any
seismic activity.
● I choose to not retrofit, and receive the entire $19,000 deposit after 12 months if there is
no earthquake damage to my home, or receive nothing if there is earthquake damage to
my home within 12 months.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 50% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. State real estate regulations require that the seller deposit $1,000 in escrow to pay for
any damages incurred from floods. The deposit will be returned to you in full after your property
has been transferred to the ne oners name and o ur property remains in good condition after
12 months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for your home using part of the $1,000 deposit; you will receive
the remaining $500 of your deposit back for sure after 12 months.

You may choose not to install sump bumps, in which you will accept the following possibility:

1. Theres an 50% chance that no flood cases damage to or propert dring the net 12
months, and your $1,000 deposit will be returned to you in full.
2. Theres a 50% chance that an earthqake does case damage t o your property during the next
12 months, and none of the $1,000 deposit is returned to you.

Please make a decision based on the information provided above:
● I choose to install the sump bump and receive $500 of the deposit back after 12 months,
regardless any flooding activity.
132

● I choose to not install the sump bump, and receive the entire $1,000 deposit after 12
months if there is no flooding damage to my home, or receive nothing if there is flooding
damage to my home within the next 12 months.

8. High disaster probability*loss frame
A. Hurricane
As a homeowner living in a high risk area for hurricanes, due to a recent job re-location you
mst sell or home. The National Weather Serice (NWS)s Climate Prediction Center has
reported that El Nino conditions are present and are expected to continue through the Northern
Hemisphere. El Niño events can cause far-reaching global disruption in the general circulation of
the Pacific Ocean and atmosphere. Additionally, there is a decrease of hurricanes in the Gulf of
Meico and Western Atlantic and an increase of hrricanes in the Pacific. NWS predicted theres
a 50% chance a hurricane will hit your area in the upcoming hurricane season. Since your home
is located on high ground, your home will be safe from flooding during hurricane season. Homes
vulnerable to hurricane winds typically have storm shutters, which guarantee your home will not
incur wind damage from a hurricane. (Note that the shutters will not affect the sale price of the
home.) Unfortunately, your home does not have storm shutters. You have two options:

You may install storm shutters for a total cost of $500, and you will not incur any damage to
your property during the 2020 hurricane season for sure.

You may choose not to install the shutters, in which you will accept the following possibility:
1. Theres a 50% chance that no hrricane dring the 20 20 hurricane season causes wind damage
to your home and you will lose nothing.
2. Theres a 50% chance that a hrricane dring the 20 20 hurricane season damages your home;
the cost to repair this damage to your home before sale is estimated to be $1,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and install storm shutters.
● I choose to not install storm shutters and either pay nothing extra if there is no damage to
my home or pay $1,000 to repair my home before it can be sold.

B. Earthquake
As a homeowner living in San Bernardino Southern California, due to a recent job re-location
you must sell your home. Seismologists have said California is due for the "big one," another
massive earthquake that would cause significant damage. "There is an extremely high chance
that there will be a damaging quake (magnitude greater than or equal to 6.7) somewhere in
California in the next 30 years," said Peggy Hellweg, a seismologist at the University of
California, Berkele. The Sothern California Earthqake Center predicted theres a 50% chance
an earthquake near the San Andreas Fault (which is in proximity to San Bernardino) will occur
within the next 12 months. Homes located in earthquake prone areas typically are retrofitted to
prevent damages from earthquakes. (Note that retrofitting will not affect the sale price of the
home.) Unfortunately, your home has not been retrofitted. You have two options:

133

You may retrofit your home for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to retrofit, in which you will accept the following possibility:
1. Theres a 50% chance that no earthqakes dring the next 12 months causes damage to your
home and you will lose nothing.
2. Theres a 50% chance that an earthqake occrs ithin the net 12 months and damages or
home; the cost to repair this damage to your home before sale is estimated to be $1,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and retrofit my home.
● I choose to not retrofit and either pay nothing extra if there is no damage to my home or
pay $1,000 to repair my home before it can be sold.

C. Flood
As a homeowner living in a high risk area for flooding, due to a recent job re-location you must
sell or home. The National Weather Serice predicted theres a 50% chance a flooding of at
least moderate magnitude (76 gauges and above) will occur in your area within the next 12
months. Homes located in flood prone areas typically have sump pumps to prevent water
damage. (Note that installing sump pump will not affect the sale price of the home.)
Unfortunately, your home has not installed sump pumps. You have two options:

You may install sump pumps for a total cost of $500, and you will not incur any damage to your
property for sure.

You may choose not to install sump pumps, in which you will accept the following possibility:
1. Theres an 50% chance that no flood dring the net 12 months cases damage to or home
and you will lose nothing.
2. Theres a 50% chance that a flood occrs ithin the net 12 months and damages or home;
the cost to repair this damage to your home before sale is estimated to be $1,000.

Please make a decision based on the information provided above:
● I choose to pay $500 and install sump pumps for my home.
● I choose to not install sump pumps and either pay nothing extra if there is no damage to
my home or pay $1,000 to repair my home before it can be sold.

134

Appendix C: Cognitive Reflection Test (CRT)
(1) A bat and a ball cost $1.10 in total. The bat costs a dollar more than the ball. How much does
the ball cost? ____ cents [Correct answer ¼ 5 cents; intuitive answer ¼ 10 cents]
(2) If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to
make 100 widgets? ____ minutes [Correct answer ¼ 5 minutes; intuitive answer ¼ 100 minutes]
(3) In a lake, there is a patch of lily pads. Every day, the patch doubles in size. If it takes 48 days
for the patch to cover the entire lake, how long would it take for the patch to cover half of the
lake? ____ days [Correct answer ¼ 47 days; intuitive answer ¼ 24 days]
(4) If John can drink one barrel of water in 6 days, and Mary can drink one barrel of water in 12
days, how long would it take them to drink one barrel of water together? _____ days [correct
answer ¼ 4 days; intuitive answer ¼ 9]
(5) Jerry received both the 15th highest and the 15th lowest mark in the class. How many
students are in the class? ______ students [correct answer ¼ 29 students; intuitive answer ¼ 30]
(6) A man buys a pig for $60, sells it for $70, buys it back for $80, and sells it finally for $90.
How much has he made? _____ dollars [correct answer ¼ $20; intuitive answer ¼ $10]
(7) Simon decided to invest $8,000 in the stock market one day early in 2008. Six months after
he invested, on July 17, the stocks he had purchased were down 50%. Fortunately for Simon,
from July 17 to October 17, the stocks he had purchased went up 75%. At this point, Simon has:
a. broken even in the stock market
b. is ahead of where he began
c. has lost money
[correct answer ¼ c, because the value at this point is $7,000; intuitive response ¼ b].

Abstract (if available)

Abstract Promoting natural disaster preparation plays a critical role in natural disaster research. Adoption of protective actions could effectively reduce the risk of injury and damage to property, however, despite considerable expenditures on public education programs, numerous studies reported that civilians still are under-prepared for natural disasters (Ballantyne, Paton, Johnston, Kozuch & Daly, 2000

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Disaster near-miss appraisal: effects of attribution, individual differences, psychological distance, and cumulative sequences

PDF

The moral foundations of needle exchange attitudes

PDF

Choice biases in making decisions for oneself vs. others

PDF

Psychological distance in the public’s response to terrorism: An investigation of the 2016 Orlando nightclub shooting

PDF

Measuring truth detection ability in social media following extreme events

PDF

The cost of missing objectives in multiattribute decision modeling

PDF

Evaluating aleatory uncertainty assessment

PDF

Comparing skipped correlations: the overlapping case

PDF

Bound in hatred: a multi-methodological investigation of morally motivated acts of hate

PDF

The role of primary care in the use of specialty mental health services: an investigation of utilization patterns among African Americans and non-Hispanic Whites

PDF

Validation of a neuroimaging task to investigate decisions involving visceral immediate rewards

PDF

Making terrorism risk assessments more useful for decision-making

PDF

A neuropsychological exploration of low-SES adolescents’ life goals and their motives

PDF

Making appropriate decisions for nonnormal data: when skewness and kurtosis matter for the nominal response model

PDF

Depression and anxiety symptom outcomes in adolescent users of smoked, vaporized, edible and blunt cannabis

PDF

Homeostatic imbalance and monetary delay discounting: effects of feeding on RT, choice, and brain response

PDF

Heart, brain, and breath: studies on the neuromodulation of interoceptive systems

PDF

Effects of childhood adversity on physiology and health in emerging adulthood

PDF

Predictors and outcomes across the transition to fatherhood

PDF

Affective neuropsychiatric symptoms and neural connectivity in the early stages of Alzheimer’s disease

Asset Metadata

Creator Zhao, Mengtian (author)

Core Title Preparing for natural disasters: investigating the effects of gain-loss framing on individual choices

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Publication Date 07/28/2020

Defense Date 05/11/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag natural disaster mitigation,OAI-PMH Harvest,risk mitigation

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor John, Richard Sheffield (committee chair), Bechara, Antoine (committee member), Lai, Hok Chio (committee member), Monterosso, John (committee member), von Winterfeldt, Detlof (committee member)

Creator Email jessicazhaomt@gmail.com,zhaomeng@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-348277

Unique identifier UC11663318

Identifier etd-ZhaoMengti-8805.pdf (filename),usctheses-c89-348277 (legacy record id)

Legacy Identifier etd-ZhaoMengti-8805.pdf

Dmrecord 348277

Document Type Dissertation

Rights Zhao, Mengtian

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA