University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Disaster Near-miss Appraisal: Effects of Attribution, Individual Differences, Psychological Distance, and Cumulative Sequences By Jinshu Cui 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) December 2017 i Acknowledgments I would like to extend my deepest gratitude to my advisor, Dr. Richard S. John, without whom I would not be who I am today. He helped me through every step of my academic development. I am always amazed by his brilliance, meticulousness, and kindness. Words cannot express how grateful I am for his tremendous support during my research career. I am also very grateful to my committee members, Drs. Detlof von Winterfeldt, John Monterosso, and Mortesa Dehghani, who provided insightful guidance and comments on my dissertation. I would also like to thank Dr. John J. McArdle. I learned a lot from his classes and he gave me invaluable advice on my second-year project and dissertation proposal. My gratitude also goes to Dr. Milind Tambe, who gave me guidance and inspiration; Dr. Heather Rosoff, who advised me on both academics and life; Dr. James Polk, who helped me in academic writing; and Dr. William Breland, who inspired my interests in teaching. Thanks also to Drs. Norbert Schwarz, JoAnn Farver, Margaret Gatz, Laura Baker, and many other professors, who have supported me in many ways. I would also like to thank my labmates: Zhiqin Chen, Kenneth Nguyen, Sarah Kusumastuti, Matthew Baucum, and Mengtian Zhao; my friends: Jinxia Ma, Jia Li, Pan Wang, Zhisen Jiang, Blake Cignarella, and Hao Wang; and staff members in the Psychology Department: Irene Takaragawa, Twyla Ponton, and Vivian Hsu-Tran. My special thanks go to my husband, Ruoren Yu, and my parents, Yu Cui and Xue Li. Their love, support, and encouragement have made me strong and happy. ii Table of Contents Acknowledgements .................................................................................................................................... i List of Tables ........................................................................................................................................... iii List of Figures .......................................................................................................................................... iv Chapter 1 : Introduction .................................................................................................................. 1 Chapter 2 : Public Response to a Near-miss Nuclear Accident Scenario Varying in Causal Attributions and Outcome Uncertainty ........................................................................................... 5 Method .................................................................................................................................................... 11 Results ..................................................................................................................................................... 18 Discussion ............................................................................................................................................... 27 Chapter 3 : Development of the Near-miss Appraisal Scale (NMAS) Using a Polytomous Item Response Theory Model ............................................................................................................... 32 Method .................................................................................................................................................... 38 Results ..................................................................................................................................................... 43 Discussion ............................................................................................................................................... 49 Chapter 4 : Effects of Psychological Distance on Near-miss Appraisals ..................................... 53 Method .................................................................................................................................................... 57 Results ..................................................................................................................................................... 62 Discussion ............................................................................................................................................... 65 Chapter 5 : Near-miss Appraisals in Sequential Near-miss Events .............................................. 68 Experiment I ........................................................................................................................................... 73 Method ................................................................................................................................................ 73 Results ................................................................................................................................................ 76 Experiment II .......................................................................................................................................... 80 Method ................................................................................................................................................ 80 Results ................................................................................................................................................ 81 Experiment III ......................................................................................................................................... 84 Method ................................................................................................................................................ 84 Results ................................................................................................................................................ 85 Discussion ............................................................................................................................................... 88 Chapter 6 : Conclusions ................................................................................................................ 91 References ..................................................................................................................................... 95 iii List of Tables Table 2.1. LOCA Scenario and Attribution and Outcome Uncertainty Manipulations ................ 13 Table 2.2. Sample Frequencies and Percentages by Sex, Income, Age, Children Under 18, and Distance from Nearest Nuclear Power Plant (N = 773) .............................................................. 16 Table 2.3. Items for Measures of Attitudes toward Nuclear Power, Risk Perception, and Behavioral Intention ..................................................................................................................... 18 Table 2.4. Means and Standard Deviations of Response Measures over Three Phases .............. 19 Table 2.5. Loadings of the Three Latent Constructs at Three Phases .......................................... 25 Table 3.1. Near-miss Appraisal Scale Items ................................................................................. 39 Table 3.2. Demographic Information of the Sample (N = 298) ................................................... 42 Table 3.3. Mean and Standard Deviation of the Measures .......................................................... 48 Table 4.1. Descriptions of two scenarios and manipulations ....................................................... 58 Table 4.2. Questions about predicted target’s behavior and risk perception .............................. 60 Table 4.3. Statements about predicted target’s perceived risks and benefits for the concert scenario ......................................................................................................................................... 61 Table 4.4. Demographic information of the sample (N = 348) .................................................... 62 Table 5.1. Design of the 20 outcomes in Experiment I ................................................................. 75 Table 5.2. Within-subjects Design of the six near misses in Experiment I ................................... 75 Table 5.3. Means, standard deviations, and correlations with sex for respondents’ NMAS and DOSPERT scores in Experiment I ................................................................................................ 76 Table 5.4. Demographic Information of the Sample (N = 159) ................................................... 80 Table 5.5. Means, standard deviations, and correlations with demographic variables for respondents’ NMAS, DOSPERT, and PRT scores in Experiment II ............................................. 81 Table 5.6. Design of the 20 outcomes in Experiment III .............................................................. 84 Table 5.7. Means, standard deviations, and correlations with sex for respondents’ NMAS, DOSPERT, and PRT scores in Experiment III ............................................................................. 85 Table 5.8. Kolmogorov-Smirnov test results and estimated parameters of the posterior Beta distributions .................................................................................................................................. 87 iv List of Figures Figure 2.1. Experimental design summary. .................................................................................. 12 Figure 2.2. Plots of mean negative affect, risk perception and behavioral intention by (a) initiating attribution (b) halting attribution and (c) outcome uncertainty. .................................... 23 Figure 2.3. SEM model with standardized parameter estimates. ................................................. 27 Figure 3.1. A Framework of the determinants of response following a near-miss ...................... 35 Figure 3.2. ICCs of two items (item 5 and item 15) from the 21-item NMAS ............................ 45 Figure 4.1. Mean predicted risk perception and behavioral intention to avoid the target location by spatial PD and social PD .......................................................................................................... 63 Figure 4.2. Mean scores of perceived benefits and risks by spatial PD. ...................................... 64 Figure 4.3. Mean scores of perceived benefits by spatial PD and concreteness. ......................... 65 Figure 4.4. An illustration of how psychological distance could change near-miss appraisal ..... 66 Figure 5.1. A picture of the scenario. ........................................................................................... 74 Figure 5.2. Mean of probability judgment and leaving intention by spatial and hypothetical psychological distance for the first (a) and second (b) near-miss sequence. ................................ 78 Figure 5.3. Mean of (a) probability judgment of damage in the next round and (b) leaving intention by spatial and hypothetical psychological distance over the 12 rounds of near-misses in Experiment I. ................................................................................................................................. 79 Figure 5.4. Mean of probability judgment of damage in the next round and leaving intention by spatial and hypothetical psychological distance for the first 10-trial near-miss sequence. .......... 82 Figure 5.5. Mean of leaving intention by hypothetical psychological distance over the 12 rounds of near-misses in Experiment II. ................................................................................................... 83 NEAR-MISS APPRAISALS IN DISASTER EVENTS 1 Chapter 1 : Introduction A near-miss is defined as an event with some “nontrivial expectation of ending in disaster” (Tinsley, Dillon, & Cronin, 2012). In contrast to a disaster that is accompanied by salient damage (e.g., death, injury, property damage, etc.), a near-miss occurs when no bad outcome is observed and is far more prevalent than a disaster. Near-misses are typically associated with events with potentially severe consequences, such as space vehicle accidents, commercial aviation accidents, and oil rig disasters. For instance, foam strike, which was the primary cause of the space shuttle Columbia disaster, had been seen at least 30 times in previous missions. A near-miss can be viewed as a signal of vulnerability and an opportunity for decision makers to learn how to avoid future failure (Carroll, 2004; Khakzad, Khan, & Paltrinieri, 2014; Yi & Bier, 1998). In fact, many industries and organizations have established near-miss reporting systems (e.g., Aviation Safety Reporting System, National Firefighter Near-miss Reporting System of the United States). However, previous research has shown that decision makers do not always engage in the same safety improvements following near-misses as they do following observable failures (Dillon & Tinsley, 2008; Dillon, Tinsley, & Burns, 2014; Rosoff, Cui, & John, 2013). For instance, Madsen, Dillon, and Tinsley (2016) reviewed data from the U.S. commercial airline industry from 1990 to 2007 and concluded that airlines learned from accidents and near-misses with salient possibilities of danger (i.e., the same category of event had resulted in accidents before), while they failed to learn from near-misses with no clear sign of danger (i.e., the same category of event had not caused accidents before). Literature on near-miss has suggested that people do not learn from resilient near-misses. Dillon and Tinsley (2008) suggested that decision makers are more likely to take protective NEAR-MISS APPRAISALS IN DISASTER EVENTS 2 actions when a near-miss event is perceived as vulnerable 1 rather than resilient. They defined vulnerable near-misses as those that highlight how a disaster almost happened. Individuals tend to interpret a vulnerable near-miss as dangerous and associated with a need to take precautionary measures (Tinsley et al., 2012; Dillon et al., 2014; Rosoff et al., 2013). For example, although no causal link to an increase in the rate of cancer was established from the 1979 Three Mile Island (TMI) nuclear power plant accident, the public viewed TMI as a serious accident due to a focus on the severity of the possible consequences that could have resulted. On the other hand, resilient near-misses were defined as those that highlight how a disaster did not happen. Individuals tend to interpret a resilient near-miss as not dangerous and for this reason do not take precautionary measures (Dillon & Tinsley, 2005; Dillon & Tinsley, 2008; Dillon, Tinsley, & Cronin, 2011; Meyer, 2012). For example, Dillon and her colleagues (2011) showed that respondents with resilient near-miss information (those whose house survived a hurricane) were less likely to purchase flood insurance and evacuate from a hurricane. However, characteristics that determine whether a near-miss is resilient or vulnerable have not been delineated in the literature. The aim of this dissertation is to explore various factors that could influence whether people would assess a near-miss as vulnerable or resilient. Specifically, I hypothesized that the interpretation of a near-miss is related to the causal attribution of the near-miss event (Chapter 2), the individual trait that determines the general tendency an individual would assess a near-miss event (Chapter 3), psychological distance between the decision maker and the event (Chapters 4 and 5), as well as the near-miss sequence (Chapter 5). 1 Note that use of the terms vulnerable, resilient, and threat in this paper generally refer to perceptions related to extreme events, and not to calculations based on definitions from the security or engineering communities. NEAR-MISS APPRAISALS IN DISASTER EVENTS 3 Chapter 2 presents a study that examined the role of causal attributions and outcome uncertainty in public responses to a near-miss nuclear accident. I first motivate why it is important to study public responses in near-miss events in a nuclear power facility. The chapter then reviews the literature covering the association between attribution and risk perception and proposes how causal attributions in different phases of a near-miss nuclear accident could affect people’s reactions. I then present an experiment that examined respondents’ self-reported negative affect, risk perception, and behavioral intention responses in a three-phase near-miss Locus-of-Control Accident scenario where initiating attribution, halting attribution, and outcome uncertainty were manipulated sequentially in the three phases. Chapter 3 proposes that the interpretation of a near-miss is not only determined by the event itself, but is also a process of subjective assessment, which I call near-miss appraisal. The chapter introduces a framework that illustrates the process of near-miss appraisal and proposes that an individual has a stable tendency (i.e., a trait) to appraise near-misses as vulnerable or resilient. The chapter first explains the trait from a theoretical standpoint. I then present the development and validation of a Near-miss Appraisal Scale to measure this tendency and demonstrate that the scale could determine an individual’s interpretation of a near-miss message. Since near-miss appraisal is a subjective process, I further propose that individual- specific characteristics could also influence the assessment of a near-miss. Chapter 4 focuses on the effect of the psychological distance between a decision maker and the near-miss event on near-miss appraisals. The chapter first introduces psychological distance (as illustrated in Construal Level Theory). I then review the literature on psychological distance, particularly on the relationship between psychological distance and risk perception, and lay out the effect of psychological distance on risk perception in near-miss events. The chapter then presents an NEAR-MISS APPRAISALS IN DISASTER EVENTS 4 online behavioral experiment that examined the role of psychological distance in a single near- miss terrorist attack scenario. Furthermore, Chapter 5 focuses on responses over a sequence of near-miss events. The chapter first reviews the relevant literature on individuals’ cognitive and behavioral responses in a sequence of risk events. The chapter then presented an online behavioral game that involved a sequence of near-miss natural disaster scenarios. I examined how respondents’ sequential responses could be influenced by psychological distance and could change as near-misses accumulated. Lastly, Chapter 6 concludes the findings from the four studies and points out the contributions the studies could make to the near-miss literature and practice. The chapter also identifies future directions of research on near-misses. NEAR-MISS APPRAISALS IN DISASTER EVENTS 5 Chapter 2 : Public Response to a Near-miss Nuclear Accident Scenario Varying in Causal Attributions and Outcome Uncertainty Abstract Many studies have investigated public reactions to nuclear accidents. However, few studies focused on more common events when a serious accident could have happened but did not (i.e., near-miss events). This study evaluated public response (emotional, cognitive, and behavioral) over three phases of a near-miss nuclear accident. Simulating a Loss-of-Coolant Accident (LOCA) scenario, I manipulated (1) attribution for the initial cause of the incident (software failure vs. cyber terrorist attack vs. earthquake), (2) attribution for halting the incident (fail-safe system design vs. an intervention by an individual expert vs. a chance coincidence), and (3) level of uncertainty (certain vs. uncertain) about risk of a future radiation leak after the LOCA is halted. A total of 773 respondents were sampled using a 3 x 3 x 2 between-subjects design. Results from both MANCOVA and Structural Equation Modeling (SEM) indicated that respondents experienced more negative affect, perceived more risk, and avoidance behavioral intention when the near-miss event was initiated by an external attributed source (e.g., earthquake) compared to an internally attributed source (e.g., software failure). Likewise, respondents also indicated greater negative affect, perceived risk and avoidance behavioral intentions when the future impact of the near-miss incident on people and the environment remained uncertain. Results from SEM analyses also suggested that negative affect predicted risk perception, and both predicted avoidance behavior. Affect, risk perception, and avoidance behavior demonstrated high stability (i.e., reliability) from one phase to the next. Keywords: nuclear power, risk perception, near-miss, causal attribution, Structural Equation Modeling NEAR-MISS APPRAISALS IN DISASTER EVENTS 6 Public Response to a Near-miss Nuclear Accident Scenario Varying in Causal Attributions and Outcome Uncertainty Nuclear power has been an essential source of electricity generation for several decades. As of 2016, over 450 power reactors in 31 countries with over 390,000 MWe of total capacity are operating commercially, generating 2,476 TWh annually, which is about 11% of the total world electricity output. The United States accounts for the most reactors (100) and the largest capacity (approximately 100,000 MWe), producing approximately 800 TWh annually, which is just under 20% of generated electricity in the US. Nuclear power accounts for an even greater percentage of generated electricity in 14 European countries, led by France with 58 reactors with a total capacity of over 63,000 MWe, producing over 72% of total generated electricity. Currently, there are approximately 60 new reactors under construction, with another 160 reactors “firmly planned.” (International Atomic Energy Agency, 2017; World Nuclear Association, 2017) Although disposal and storage of nuclear waste remains the subject of heated debate (Kunreuther, Easterling, Desvousges, & Slovic, 1990; Slovic, Layman, & Flynn, 1991), more public fear and concern about nuclear power would likely result if a nuclear accident occurred (van Der Pligt, 1985; de Boer & Catsburg, 1988). In particular, studies have found past major nuclear accidents (e.g., Chernobyl, Fukushima) negatively impacted perceived risks (McDaniels, 1988; Lindell & Perry, 1990; Katsuya, 2001), behavioral intentions (Prati & Zani, 2013), public attitudes (Verplanken, 1989; Wittneben, 2012; Kasperson, et al., 1988) and policies (Wittneben, 2012) related to nuclear power. Even in the case of no direct casualties or long-term disease or deaths from radiation exposure (e.g., the Three Mile Island accident), the occurrence of a nuclear accident would likely impact society and the economy substantially due to social amplification of risk (Kasperson, et al., 1988; Denning & Mubayi, 2016). For instance, greater public NEAR-MISS APPRAISALS IN DISASTER EVENTS 7 opposition to nuclear power after an accident could cause stricter regulations and reduce ongoing and planned operation of nuclear reactors. While many studies have investigated the influence of major nuclear power plant accidents on the public, few studies have focused on incidents that have much less severe outcomes but are more common than accidents. In this study, I am particularly interested in near- miss nuclear accidents. I consider a near-miss nuclear accident where direct consequences for people and the environment (e.g., direct casualty, significant release of radioactive substances) could have happened but did not. According to the International Nuclear and Radiological Event Scale (INES) (International Atomic Energy Agency, 2013), the safety significance of a nuclear event is rated from level 1 to level 7, with levels 1 to 3 as incident-levels and levels 4 to 7 as accident-levels. The near-miss nuclear accidents considered in the current study are incident- level events. A nuclear accident is breaking news that would almost certainly draw enormous attention. However, public reaction might change once the possible disaster is deemed a near- miss event. Public perceptions and concerns might also change when the story fades and life returns to normal. Because public reactions to disaster (and near-disaster) events are dynamic, it is important to look at public response at different phases of an unfolding disaster, such as a near-miss nuclear accident. In the current study, I created a three-phase near-miss nuclear accident scenario and manipulated various features of the scenario to systematically investigate how these scenario characteristics influence public reaction. Specifically, I focused on (1) attributions for the initiating event that caused the incident that could have led to a devastating disaster, (2) attributions for the controlling mechanism that halted the incident sequence and NEAR-MISS APPRAISALS IN DISASTER EVENTS 8 averted disaster, leading to a near-miss, and (3) level of uncertainty regarding long-term damage to the environment and negative impacts on society from the incident. The current study explores the relationship between near-miss nuclear accidents, appraised as either vulnerable or resilient, and causal attributions associated with the incident sequence. Attribution theory (Kelley & Michela, 1980; Weiner, 1985, 1995) suggested that observers generate causal attributions to explain behaviors and events in everyday life. Causal attributions could affect emotional experiences and expectancies of outcomes and could further guide future behaviors when individuals encounter similar situations. Causal attributions can be characterized as either internal (intrinsic characteristics, related to an individual’s ability or effort) or external (situational characteristics such as luck or other uncontrollable factors). Previous research on the association between attribution and risk perception has focused on how people make attributions and indicates that people are more likely to make external attributions for bad outcomes (e.g., experienced illness during an epidemic because of a chance encounter with another infected individual). For instance, Kahlor et al. (2002) showed that individuals tended to ascribe their infections from a drinking water outbreak to external causes. Rickard (2014) looked at public response to accidents in U.S. parks and found that individuals who perceived the park as more dangerous were more likely to attribute responsibility to an external (e.g., park employees) rather than an internal factor (e.g., victims themselves). Among the few that examined how attribution could influence public response to risk events, Griffin and colleagues (2008) found that respondents who ascribed flood damage to an external factor (poor government management) were angrier with the agencies than those who ascribed flood damage to an internal reason (people living near a river). NEAR-MISS APPRAISALS IN DISASTER EVENTS 9 In the scenario of a near-miss nuclear accident, it is reasonable to think that when the incident first happened, people would perceive the incident that caused by a more external attribution less controllable and more risky. Indeed, Rickard (2014) demonstrated that external attribution was associated with greater perceived controllability of the park-related risks. When the incident is controlled, people might also consider that an incident that averted by chance is more alarming. Contrarily, they might think that as the facility is prepared to deal with such an incident, the facility expects the failure and unstableness of the system, which makes the risk of nuclear power more concerning. Therefore, I hypothesized that when a nuclear incident is initiated, external causal attributions would be associated with greater negative affect, risk perception, and avoidance behavioral intention. I made no specific predictions regarding what halted the incident sequence but anticipated that negative affect, risk perception, and avoidance behavioral intention after learning the disaster is averted would be affected by the halting attribution. Specifically, I investigated three initiating attributions: (1) an internal cause, software failure, which is similar to the cause of the Three Mile Island accident, (2) an external cause, unforeseen earthquake, which is similar to the direct cause of the Fukushima Daiichi accident, and (3) a cause both external and internal (ambiguous), a cyber terrorist attack. A cyber attack is partly internal in that it indicates a failure of defense to protect the plant, and is external in that it is directed by an intelligent adversary. I anticipated that a near-miss accident caused by an unforeseen earthquake would have a more negative impact on public responses than one caused by software failure. I suspected that an external cause is associated with unpredictable circumstances, highlighting the vulnerability of the system, whereas an internal cause is associated with more well-understood circumstances associated with system resilience. NEAR-MISS APPRAISALS IN DISASTER EVENTS 10 As for the halting attribution that averted disaster, I also investigated three attributions: (1) an internal cause, fail-safe system design, (2) an external cause, a coincidence, and (3) a cause both internal and external (ambiguous), an individual expert. An individual expert is partly internal in that the person is employed by the facility, and is partly external in that the expert happened to be at the right place at the right time. I made no specific predictions regarding the halting attributions. I anticipated that a disaster prevented by an externally or internally attributed mechanism would have an impact on individuals’ responses. In addition, I focus on a third phase of a near-miss accident – after the incident sequence is halted, and I explore the effect of outcome uncertainty on public response. I anticipated that greater uncertainty regarding radioactive release or harm to the environment would result reports of greater negative affect, greater perceived risk, and greater intention to engage in avoidance behavior. The public would likely have larger concerns about nuclear power if the long-term impact on society if the eventual consequences of the near-miss event were uncertain. Moreover, I hypothesized that public response (negative affect, risk perception, behavioral intention) from an earlier phase of a near-miss nuclear accident could influence response in later phases of the unfolding event. Previous research has found that public perceptions and beliefs before a nuclear accident could influence public response after the accident. For instance, in a study evaluating public acceptance of nuclear power in Switzerland before and after the Fukushima Daiichi disaster (Visschers & Siegrist, 2013), researchers found that public trust related to nuclear power plants before the accident strongly affected public trust after the accident. I also hypothesized that at a given phase of the incident, negative affective response could influence perceived risk and both negative affect and risk perception could influence NEAR-MISS APPRAISALS IN DISASTER EVENTS 11 intended behavior. Previous research has demonstrated causal relationships among emotional, cognitive and behavioral responses (Rubin, Amlôt, Page, & Wessely, 2009; Rundmo, 2002; Slovic, Peters, Finucane, & MacGregor, 2005). For instance, negative emotion has been shown to influence risk perception (Peters & Slovic, 1996) and decision making (Smith, Kay, Hoyt, & Bernard, 2009; Terpstra, 2011). In one study about public response to terrorist attacks, the effect of negative affect on behavioral intention was found to be mediated by perceived risk (Cui, Rosoff, & John, 2016). I also anticipated that individual difference variables such as sex, age, and income could influence responses throughout a nuclear incident. Previous studies have shown that females generally report higher perceived risk than males. For example, Lerner et al. (2003) found that females perceived more risk than males in the aftermath of 9/11. I hypothesized that females would report greater negative affect, risk perception, and avoidance behavioral intention than males. I also hypothesized that older respondents would exhibit attenuated responses to a nuclear incident, since they have more experience with historical nuclear plant events. In addition, I anticipated respondents with higher incomes would perceive less risk about a nuclear incident, as those with higher incomes may pay more attention to the commercial benefits of nuclear power plants. Method Scenario and Design A scenario of a Loss-of-Coolant Accident (LOCA) that is quickly controlled and has no confirmed direct impact on people and the environment (a near-miss) was constructed. Figure 2.1 depicts the three phases of the incident and the manipulated variables at each phase. Negative affect, risk perception, and behavioral intention were repeatedly measured at all three phases. NEAR-MISS APPRAISALS IN DISASTER EVENTS 12 The incident starts when the primary cooling pumps at a nuclear facility fail (phase 1). The failure disables the plant’s cooling system and introduces the risk of a radiation leak. A radiation leak could bring serious consequences to people and the environment (cost of cleanup, evacuation, etc.). Attributions for the pump failure were manipulated as caused alternatively by a system software failure (internal), a cyber terrorist attack (ambiguous), or an earthquake (external). By the end of the day, the primary cooling pumps are restored (phase 2). Attributions for halting the LOCA were manipulated as the robustness of the original design of the system (internal), intervention by an individual expert (ambiguous), or a coincidental scheduled plant shutdown (external). Four days after the original pump failure, as the incident is contained and the pumps are running normally (phase 3), the uncertainty level regarding damage to the nuclear plant core and a possible radiation leak was manipulated as either known (certain) or unknown (uncertain). A complete description of all three scenes, including the manipulations at each phase, is provided in Table 2.1. Figure 2.1. Experimental design summary. Phase&1& Phase&2 Phase&3 (3&groups) (9&groups) (18&groups) Nega4ve& Affect& & Risk& Percep4on& & Behavioral& Inten4on& Nega4ve& Affect& & Risk& Percep4on& & Behavioral& Inten4on& Nega4ve& Affect& & Risk& Percep4on& & Behavioral& Inten4on& Ini4a4ng& AAribu4on& Hal4ng& AAribu4on& Outcome& Uncertainty& soHware&failure& (internal)& earthquake& (external)& cyber&aAack& (ambiguous)& system&design& (internal)& coincidence& (external)& individual&expert& (ambiguous)& uncertain& certain& NEAR-MISS APPRAISALS IN DISASTER EVENTS 13 The three phases of the incident scenario were developed as three videos 2 , which simulated news broadcasts that take place on Day 1 (phase 1), the end of Day 1 (phase 2), and Day 5 (phase 3), respectively, of a pump failure incident at the Byron Nuclear Power Facility near Chicago. The lengths of the videos ranged from 1:16 to 2:13 minutes. A unique independent groups experimental design was utilized, resulting in a 3 (Day 1: initiating attribution) by 3 (end of Day 1: halting attribution) by 2 (Day 5: outcome uncertainty) factorial design. While the design resembles a between-subjects design, since respondents were randomly assigned to one of the eighteen conditions, it is important to realize that responses depend only on manipulations prior to the responses; specifically, phase 1 responses are independent of the manipulation at phases 2 and 3, and phase 2 responses are independent of the manipulation at phase 3. Table 2.1. LOCA Scenario and Attribution and Outcome Uncertainty Manipulations 2 Access to the videos is available upon request. Scenario Day 1 Initiating Attribution A software failure has shut off the primary pumps at the Byron Nuclear Power Facility near Chicago, Illinois. Without pumps, the cooling system is disabled and introduces the risk of nuclear core temperatures risking to unsafe levels. (software failure) Cyber terrorists have hacked into the Byron Nuclear Power Facility near Chicago, Illinois and shut off the primary pumps. Without pumps, the plant’s cooling system is disabled and introduces the risk of nuclear core temperatures rising to unsafe levels. (cyber terrorist attack) A magnitude 7.8 earthquake just struck Byron County, Illinois. The earthquake brought down the computer system at the Byron Nuclear Power Facility, shutting off the primary pumps. Without pumps, the cooling system is disabled and introduces the risk of nuclear core temperatures risking to unsafe levels. (earthquake) According to the NRC, the U.S. Nuclear Regulatory Commission, unsafe temperatures could lead to a partial meltdown within the reactor or a high pressure buildup of radioactive steam within the nuclear containment vessel. Consequences of either event could result in the need to vent radioactive steam, a radioactive leak or a large explosion, all of which releases radioactive material into the atmosphere. Technicians are on the scene working to determine how to get the pumps back up and running. While it really is too early to say how this happened, Initiating Attribution computer experts believe that the problem originated in the program code responsible for sending plant managers and computer experts believe the system was hacked remotely computer experts believe the facility’s computer server, which is responsible for regulating the flow of water NEAR-MISS APPRAISALS IN DISASTER EVENTS 14 signals to the cooling system. (software failure) using very advanced technology. (cyber terrorist attack) through the coolant pumps, was damaged during the earthquake. (earthquake) Day 1 – end of the day We just received word that the primary cooling pumps are back up and running at the Byron Nuclear Power Facility. According to the plant manager, the facility’s back up power system activated and power was restored to the cooling pumps just a short while ago. Jake Howard, a nuclear physicist and operator at the Byron Nuclear Power Facility, was present during the ordeal. He joins us live via satellite: Anchor: Hello Jake – thanks for taking the time to speak with me. How are you and the rest of your team? Halting Attribution Jake: .. Anchor: So the computer systems that run the facility were designed to respond to these kinds of events? Jake: .. Anchor: “Well, thanks Jake for your time. As we just heard, the plant’s system design prevented what could have been a devastating disaster. (system design) Jake: .. Anchor: So I am guessing you are really proud to have Andrew on your team? Jake: .. Anchor: “Well, thanks Jake for your time. As we just heard, the heroism of Andrew Cash prevented what could have been a devastating disaster. (individual expert) Jake: .. Anchor: So, how often does the plant conduct scheduled shut-downs? Jake: .. Anchor: “Well, thanks Jake for your time. As we just heard, the coincidental scheduling of the plant failure prevented what could have been a devastating disaster. (coincidence) However, NRC officials are still uncertain about damage to the nuclear core. Damage to the core could cause a pressurized buildup of chemical hydrogen in the form of radioactive steam. To avoid such damage, operators would have to vent some of this steam into the atmosphere, meaning hazardous materials would have to be released into the environment. We will continue to provide updates as the story develops. Day 5 Good evening everyone. What started five days ago as the threat of a nuclear power plant meltdown at the Byron Nuclear Power Facility is now contained – the cooling pumps are back up and running. However, the effects of the temporary failure are still very uncertain. What we do know is that when the cooling pumps failed, water stopped circulating within the reactor, increasing the core’s temperature. This increase in temperature was due to heat produced by radioactive decay. As the core temperature increased, so did the pressure within the core containment system, increasing the risk of explosions from the release of chemical hydrogen. As of now, the risk of this occurring has been minimized because the cooling systems are back online. Outcome Uncertainty Just coming in… the assessment of the core has been completed and officials are reporting that the level of damage is minimal and there is no risk, I repeat no risk of radioactive release into the air or water supply. The pressure within the core containment system has been stabilized. According to a spokesman at the NRC, “The plant is structurally sound and completely safe.” (certain no damage) What remains uncertain is whether there has been any damage to the facility’s nuclear core. If there is significant damage to the core, the potential effects vary in both type and severity. In many cases, core damage leads to a radioactive release into the surrounding air or water sources because of cracks in the containment vessel or various pipelines that lead to or away from it. Technicians and operators are currently working to assess the damage to the nuclear core. According to a spokesman at the NRC, “Until we understand the extent of the damage to the reactor, decisions with respect to public safety and facility clean up are in a holding pattern.” (uncertain damage) NEAR-MISS APPRAISALS IN DISASTER EVENTS 15 Before watching the video incident scenario, respondents were first asked to evaluate their general attitudes toward nuclear power. Following each video, respondents were asked to evaluate their present feelings, risk perception, and intended avoidance behavior. After watching all three videos, respondents answered two attention check questions about the three videos 3 and were asked to assess their general attitudes toward nuclear power again and to report their personal experiences related to nuclear power plants and nuclear accident events. They also reported demographic information including sex (male or female), age (year born), and income (gross annual house income in one of seven ranges). Respondents Nine hundred and one respondents recruited from Amazon Mechanical Turk (AMT) completed the survey. After excluding 106 respondents who failed to answer one or more of the two attention check questions correctly and 22 respondents who completed the survey in less than 11 minutes 4 , 773 respondents were included in the analysis. Each respondent was paid $1 for completing the survey. There were 409 (53%) female respondents. The median age of the sample was 34 years (IQR = 27 – 47). The median completion time was 17 minutes (IQR = 15 – 21). Only 21 (2.7%) had educational or work experience in the field of nuclear power, and 174 (22.5%) reported that the nearest nuclear power plant was less than fifty miles away from their residence 5 . A more detailed description of the sample is included in Table 2.2. 3 The two attention check questions are: “What was the cause of the nuclear power facility’s computer failure?” and “What prevented a meltdown after the incident?” 4 I expected respondents to need 6 minutes to watch all three videos and five minutes to answer all questions. 5 According to the 2010 U.S. census, one in three people live within the 50-mile radius around nuclear power plants. NEAR-MISS APPRAISALS IN DISASTER EVENTS 16 Table 2.2. Sample Frequencies and Percentages by Sex, Income, Age, Children Under 18, and Distance from Nearest Nuclear Power Plant (N = 773) Variables Values Frequencies and Percentages (N = 773) Sex Male 358 (46.3%) Female 406 (52.5%) Income < $20,000 144 (18.6%) $20,000-$40,000 206 (26.6%) $40,000-$60,000 171 (22.1%) $60,000-$80,000 110 (14.2%) $80,000-$100,000 63 (8.2%) $100,000-$150,000 56 (7.2%) > $150,000 21 (2.7%) Age 18-27 200 (26.0%) 28-34 189 (24.5%) 35-47 193 (25.0%) 48-82 188 (24.3%) Children under 18 Yes 248 (32.1%) No 523 (67.7%) Distance from nearest nuclear power plant < 10 miles 9 (1.2%) 10-50 miles 163 (21.2%) > 50 miles 291(37.6%) I don’t know 308 (39.8%) Measures Attitudes toward nuclear power. Before and after watching all three videos, respondents were asked to evaluate their general attitudes toward nuclear power. Respondents indicated level of agreement with eight statements about nuclear power using a Likert scale from 1 (strongly disagree) to 6 (strongly agree). The eight statements are included in Table 2.3. The eight-item scale was internally consistent (Cronbach’s 𝛼 = 0.92 and 0.93, respectively, before and after the incident scenario videos were viewed). Composite scores using unit weighting were NEAR-MISS APPRAISALS IN DISASTER EVENTS 17 calculated for the eight items at both time points. A higher score indicated greater support for nuclear power. Negative affect. After watching each video, respondents rated their current affect described by emotion words from the Positive and Negative Affect Scale (PANAS) (Watson, Clark, & Tellegen, 1988) using a 5-point scale from 1 (not at all) to 5 (extremely) 6 . The negative affect items included: (1) distressed, (2) upset, (3) scared, (4) nervous, (5) afraid, (6) hostile, (7) irritable, (8) ashamed, (9) jittery, and (10) guilty. Two items, “ashamed” and “guilty,” were excluded because Cronbach’s 𝛼 increased after removal. The remaining eight items were internally consistent (Cronbach’s 𝛼 = 0.94, 0.95, 0.95, respectively, for the three phases). Composite scores using equal weighting were calculated for all three negative affect scores. A higher score indicates greater negative affect. Risk perception. Following completion of the affect scale, respondents answered questions about the risk they perceived related to the incident. Respondents indicated the level of agreement with three statements using a scale from 1 (strongly disagree) to 6 (strongly agree). The three statements are also included in Table 2.3. The three questions were internally consistent (Cronbach’s 𝛼 = 0.93, 0.92, and 0.90). A composite score using unit weighting was calculated for the three scores. A higher score indicated more perceived risk. Behavioral intention. After answering questions about perceived risk, respondents reported their intended behavior regarding the scenario described in the video. Respondents indicated the likelihood that they would engage in six behaviors using a scale from 1 (very unlikely) to 6 (very likely). The six statements are again included in Table 2.3. Item 2 was excluded because Cronbach’s 𝛼 increased after removal. The five-item scale was internally 6 I administered the full 20-item PANAS but analyzed only negative affect since I had no hypotheses related to positive affect. NEAR-MISS APPRAISALS IN DISASTER EVENTS 18 consistent (Cronbach’s 𝛼 = 0.88, 0.91 and 0.92). Composite scale scores using unit weighting were calculated for the five items at each time point. A higher score indicated greater avoidance behavior intention. Table 2.3. Items for Measures of Attitudes toward Nuclear Power, Risk Perception, and Behavioral Intention Attitudes toward nuclear power 1 The United States should continue to invest in nuclear power. 2* There are serious risks associated with nuclear power. 3* The risks associated with nuclear power are greater than the benefits. 4 I have confidence in nuclear power. 5 I have confidence in nuclear plant operators’ ability to prevent severe nuclear incidents. 6 I have confidence in nuclear plant operators’ ability to contain severe nuclear incidents. 7* Nuclear security policy needs to be reformed. 8* All nuclear power plants in the United States should be closed. Risk perception 1 The nuclear power plant incident has created a great deal of anxiety for me. 2 The nuclear power plant incident caused me to worry about my own safety. 3 I believe the nuclear incident poses a serious risk to me. Behavioral intention 1 I would evacuate the area. 2 I would stay in my home. 3* I would continue with my normal daily lifestyle. 4* I would continue to take my children to school if open. 5 I would change my daily routine to avoid being near the plant or being exposed to radiation. 6 I would take necessary precautions to avoid potential radiation exposure when outside. Note: * Reverse-coded items Results I conducted the analysis in two steps. First, I performed Multivariate Analysis of Covariance (MANCOVA) on responses at each of the three phases in the near-miss accident progression to evaluate the effect of the three manipulated variables of the incident on reactions of negative affect, risk perception and behavioral intention. Second, I conducted an analysis with Structural Equation Modeling (SEM) to test the relationships among emotional, cognitive, and behavioral reactions at the same phase and across different phases. NEAR-MISS APPRAISALS IN DISASTER EVENTS 19 MANCOVA A between-subjects MANCOVA was performed in SPSS on the three dependent variables (negative affect, risk perception, and behavioral intention) for each of the three phases, corresponding to the three phases of the incident progression scenario. The means and standard deviations of the three dependent variables for the three phases are included in Table 2.4. At phase 1, initiating attribution of the pump failure is the independent variable; at phase 2, both initiating attribution and halting attribution are independent variables; at stage 3, outcome uncertainty, as well as initiating attribution and halting attribution, are independent variables. Adjustments were made for three covariates: (1) sex, (2) age, and (3) income. I also used contrast coding to examine specific hypotheses for responses to different levels of manipulations (initiating attribution: software failure vs. terrorist attack vs. earthquake = 0, -1/2, 1/2 and 2/3, - 1/3, -1/3; halting attribution: system design vs. expert intervention vs. coincidence = 0, -1/2, 1/2 and 2/3. -1/3, -1/3). For both initiating and halting attributions, the first contrast compares ambiguous with external conditions, and the second contrast compares internal with the combined ambiguous and external conditions. For the covariates, sex was a dichotomous variable; age was a continuous variable; income was an ordinal variable and was included as a linear contrast for the seven-ordered categories. Table 2.4. Means and Standard Deviations of Response Measures over Three Phases Phase 1 Phase 2 Phase 3 Mean S.D. Mean S.D. Mean S.D. Negative affect 2.66 1.08 2.12 0.96 1.99 0.92 Risk perception 4.39 1.32 4.13 1.30 4.02 1.35 Avoidance behavior 4.59 1.15 3.87 1.29 3.23 1.37 NEAR-MISS APPRAISALS IN DISASTER EVENTS 20 Figure 2.2 includes plots of mean negative affect, risk perception and avoidance behavior by initiating attribution (Figure 2.2 (a)), halting attribution (Figure 2.2 (b)), and outcome uncertainty (Figure 2.2 (c)) at phases 1, 2, and 3 where manipulations are significant. Results for phase 1 indicated that initiating attribution significantly predicted negative affect (𝐹(2,742) = 6.99, 𝑝 = 0.001, 𝜂 ! = 0.02), risk perception (𝐹(2,742) = 3.63, 𝑝 = 0.03, 𝜂 ! = 0.01), and behavioral intention (𝐹(2,742) = 6.56, 𝑝 = 0.001, 𝜂 ! = 0.02). Specifically, results from the contrast coding indicated that a pump failure caused by a software failure (internal) predicted less negative affect (𝑝 < 0.001), less perceived risk (𝑝 = 0.049), and less avoidance behavioral intention (𝑝 = 0.001) than that caused by a cyber terrorist attack (ambiguous) or an earthquake (external). In addition, females reported more negative affect (𝐹(1,742) = 16.48, 𝑝 < 0.001, 𝜂 ! = 0.02), more perceived risk (𝐹(1,742) = 23.94, 𝑝 < 0.001, 𝜂 ! = 0.03), and greater avoidance behavior (𝐹(1,742) = 25.09, 𝑝 < 0.001, 𝜂 ! = 0.03) than males. Older respondents reported less negative affect (𝐹(1,742) = 4.67, 𝑝 = 0.03, 𝜂 ! = 0.006), less perceived risk (𝐹(1,742) = 8.69, 𝑝 = 0.003, 𝜂 ! = 0.01), and less avoidance behavior (𝐹(1,742) = 5.23, 𝑝 = 0.02, 𝜂 ! = 0.007) than younger respondents. Respondents with higher incomes reported less perceived risk (𝐹(1,742) = 14.34, 𝑝 < 0.001, 𝜂 ! = 0.02) and less avoidance behavior (𝐹(1,742) = 7.32, 𝑝 = 0.007, 𝜂 ! = 0.01) than those with lower incomes. For phase 2, initiating attribution again significantly predicted negative affect (𝐹(2,740) = 4.34, 𝑝 = 0.01, 𝜂 ! = 0.01), risk perception (𝐹(2,740) = 4.36, 𝑝 = 0.01, 𝜂 ! = 0.01), and avoidance behavior (𝐹(2,740) = 7.89, 𝑝 < 0.001, 𝜂 ! = 0.02). Similar to phase 1 results, main effects based on contrast coding indicated that a pump failure caused by an earthquake (external) or a cyber attack (ambiguous) predicted responses of more negative affect (𝑝 = 0.003), perceived risk (𝑝 = 0.004) and avoidance behavior (𝑝 < 0.001) than did that caused by a software failure (internal). NEAR-MISS APPRAISALS IN DISASTER EVENTS 21 Attribution for halting the incident sequence did not significantly predict negative affect (𝐹 < 1) or risk perception (𝐹 < 1) for phase 2, but significantly predicted behavioral intention (𝐹(2,740) = 10.05, 𝑝 < 0.001, 𝜂 ! = 0.03). Respondents who found that an individual expert (ambiguous) had halted the failure reported more avoidance behavior (𝑝 < 0.001) than did those who found that the pump failure happened to be in a coincidental scheduled plant shutdown (external). In addition, females reported more negative affect (𝐹(1,740) = 14.51, 𝑝 < 0.001, 𝜂 ! = 0.02), perceived risk (𝐹(1,740) = 27.33, 𝑝 < 0.001, 𝜂 ! = 0.04), and avoidance behavior (𝐹(1,740) = 28.89, 𝑝 < 0.001, 𝜂 ! = 0.04) than did males. Older respondents reported less negative affect (𝐹(1,740) = 27.88, 𝑝 < 0.001, 𝜂 ! = 0.04), perceived risk (𝐹(1,740) = 28.74, 𝑝 < 0.001, 𝜂 ! = 0.04) and avoidance behavior (𝐹(1,740) = 30.21, 𝑝 < 0.001, 𝜂 ! = 0.04) than younger respondents. Respondents with higher incomes reported less perceived risk (𝐹(1,740) = 12.41, 𝑝 < 0.001, 𝜂 ! = 0.02) and less avoidance behavior (𝐹(1,740) = 4.02, 𝑝 = 0.045, 𝜂 ! = 0.005) than those with lower incomes. For phase 3, initiating attribution again significantly predicted negative affect 𝐹(2, 733) = 3.03, 𝑝 = 0.049, 𝜂 ! = 0.008), risk perception (𝐹(2, 733) = 4.18, 𝑝 = 0.02, 𝜂 ! = 0.01), and avoidance behavior (𝐹(2,733) = 4.34, 𝑝 = 0.01, 𝜂 ! = 0.01). As found at phases 1 and 2, pump failure caused by an earthquake (external) or a cyber terrorist attack (ambiguous) predicted responses of more negative affect (𝑝 = 0.01), more perceived risk (𝑝 = 0.01) and greater avoidance behavior (𝑝 = 0.005) than did that caused by a software failure (internal). Halting attribution did not significantly predict negative affect (𝐹 < 1), perceived risk (𝐹 < 1) or behavioral intention (𝐹(2,733) = 1.28, 𝑝 = 0.28, 𝜂 ! = 0.003). An uncertain damage to the nuclear core (uncertain) predicted more negative affect (𝐹(1,733) = 70.96, 𝑝 < 0.001, 𝜂 ! = 0.09), more perceived risk (𝐹(1,733) = 32.66, 𝑝 < 0.001, 𝜂 ! = 0.04), and more avoidance behavior (𝐹(1,733) NEAR-MISS APPRAISALS IN DISASTER EVENTS 22 = 112.14, 𝑝 < 0.001, 𝜂 ! = 0.13) than did a known absence of damage (certain). In addition, females reported more negative affect (𝐹(1,733) = 9.58, 𝑝 = 0.002, 𝜂 ! = 0.01), perceived risk (𝐹(1,733) = 25.33, 𝑝 < 0.001, 𝜂 ! = 0.03), and avoidance behavior (𝐹(1,733) = 18.16, 𝑝 < 0.001, 𝜂 ! = 0.02) than did males. Older respondents reported less negative affect (𝐹(1,733) = 26.04, 𝑝 < 0.001, 𝜂 ! = 0.03), perceived risk (𝐹(1,733) = 24.66, 𝑝 < 0.001, 𝜂 ! = 0.03), and avoidance behavior (𝐹(1,733) = 8.27, 𝑝 = 0.004, 𝜂 ! = 0.01) than younger respondents. Respondents with higher incomes reported less perceived risk (𝐹(1,733) = 12.74, 𝑝 < 0.001, 𝜂 ! = 0.02) than those with lower incomes 7 . (a) (b) 7 Results from MANOVAs when the three covariates were excluded showed that the pattern of significant effects of the manipulations remain the same. NEAR-MISS APPRAISALS IN DISASTER EVENTS 23 (c) Figure 2.2. Plots of mean negative affect, risk perception and behavioral intention by (a) initiating attribution (b) halting attribution and (c) outcome uncertainty. I also conducted repeated measures ANOVAs on negative affect, risk perception, and behavioral intention with time (phases 1, 2, and 3) as a linear predictor. Results indicated that reported negative affect (𝐹(1,748) = 518.36, 𝑝 < 0.001, 𝜂 ! = 0.41), perceived risk (𝐹(1,766) = 85.00, 𝑝 < 0.001, 𝜂 ! = 0.10) and avoidance behavior (𝐹(1,764) = 872.14, 𝑝 < 0.001, 𝜂 ! = 0.53) significantly decreased from phase 1 to phase 3. In addition, a repeated measures ANOVA on attitudes toward nuclear power with time (before and after watching the video scenario simulation) as a predictor suggested that respondents became less supportive of nuclear power after watching the three videos (𝐹(1,757) = 49.75, 𝑝 < 0.001, 𝜂 ! = 0.06). SEM Analysis I performed a SEM analysis to investigate the relationships among the various public response measures over the progression of the near-miss nuclear accident. A SEM combines confirmatory factor analysis that validates the measurement of a latent construct with regression analysis that examines the relationship between latent constructs. The dependent variable in one regression equation can serve as a predictor variable in another equation and multiple causal relationships can be evaluated simultaneously. One response variable can influence another directly or through other variables as mediators. A SEM model also allows examination of the NEAR-MISS APPRAISALS IN DISASTER EVENTS 24 causal relationship from a latent construct at an earlier phase to another latent construct at a later phase (Mayer & Carroll, 1987; McArdle, 2009). The proposed model. I applied SEM models for the responses to the incident progression scenarios. There are three latent constructs in the current analysis: negative affect (8 items), risk perception (3 items), and avoidance behavioral intention (5 items). I started from a parsimonious model that accounts for all three previously identified hypotheses concerning relationships among these three constructs. First, attribution for the initiation of the incident sequence could influence the three latent constructs at phase 1, attribution for halting the incident progression could influence the three constructs at phase 2, and outcome uncertainty could influence the three constructs at phase 3. Second, for each latent construct, responses from an earlier phase of the incident could influence the same latent construct in the next phase of the event (i.e., from phase 1 to phase 2 and from phase 2 to phase 3). Third, at a given phase of the incident, the effect of negative affect on behavioral intention could be mediated by risk perception. In addition, I tested three alternative models. First, attribution for initiation of the incident sequence could influence the latent constructs at all three phases and attribution for halting the incident progression could influence the latent constructs at phases 2 and 3. Second, each of the latent constructs (negative affect, risk perception, and behavioral intention) at phase 1 could also predict the same latent construct at phase 3. Third, negative affect could influence both risk perception and avoidance behavioral intention, and risk perception could also influence intended avoidance behavior at each of the three phases. I coded initiating attribution into the same two contrast variables as used in the MANCOVA analyses. NEAR-MISS APPRAISALS IN DISASTER EVENTS 25 Model estimation. I estimated the SEM models using the ‘lavaan’ package (Rosseel, 2012) in R. There were 38 (5%) cases that contained missing values in the sample. Full information maximum likelihood (fiml) in SEM (Enders & Bandalos, 2001) was used to deal with missing data. I used likelihood ratio tests to compare each of the alternative models with the proposed model (each pair of the models was nested) (Steiger, Shapiro, & Browne, 1985). The first two alternative models did not fit the data better than the proposed model, however, the third alternative model fit the data better than the proposed model (∆χ ! (3) = 92.92, 𝑝 < 0.001). Therefore, the third alternative model was chosen. The model fit the data moderately well (Hooper, Coughlan, & Mullen, 2008): χ ! (1296, N = 773) = 4520, 𝑝 < 0.001, CFI = 0.92, RMSEA = 0.057 (90% confidence interval from 0.055 to 0.059). The estimated model is depicted in Figure 2.3. Loadings for the three latent constructs at three phases are reported in Table 2.5. At phase 1, reported negative affect was higher for a pump failure caused by a cyber attack and earthquake than that caused by a software failure (𝛽 = 0.11, 𝑝 = 0.003). Perceived risk was not differentiated by the attribution for the initiating event. Reported avoidance behavior was also higher for a pump failure caused by a cyber attack and earthquake than that caused by a software failure (𝛽 = 0.07, 𝑝 = 0.02). Negative affect significantly predicted risk perception (𝛽 = 0.73, 𝑝 < 0.001), and risk perception significantly predicted avoidance behavioral intention (𝛽 = 0.52, 𝑝 < 0.001). Table 2.5. Loadings of the Three Latent Constructs at Three Phases Phase 1 Phase 2 Phase 3 Negative Affect Jittery 0.73 0.48 0.39 Irritable 0.59 0.43 0.34 Hostile 0.55 0.36 0.30 Distressed 0.88 0.61 0.48 Upset 0.86 0.58 0.47 NEAR-MISS APPRAISALS IN DISASTER EVENTS 26 Scared 0.93 0.64 0.51 Nervous 0.86 0.61 0.49 Afraid 0.94 0.64 0.50 Risk Perception Create anxiety 0.61 0.47 0.39 Worry own safety 0.64 0.47 0.39 Pose a serious risk 0.63 0.47 0.34 Behavioral Intention Evacuate 0.67 0.55 0.50 Change daily routine 0.55 0.47 0.43 Take children to school -0.66 -0.55 -0.53 Take precautions when outside 0.50 0.42 0.38 Continue daily routine -0.67 -0.58 -0.54 Note: All loadings are significant, 𝑝 < 0.05 At phase 2, reported negative affect and risk perception were not affected by how the incident was controlled (attribution for halting the incident sequence). More avoidance behavioral intention was reported when the pump failure was controlled by an individual expert (ambiguous) than if there was a coincidental system shutdown (external) (𝛽 = 0.14, 𝑝 < 0.001). Reported negative affect (𝛽 = 0.73, 𝑝 < 0.001), risk perception (𝛽 = 0.60, 𝑝 < 0.001), and behavioral intention (𝛽 = 0.52, 𝑝 < 0.001) were positively influenced by the same construct at phase 1. In addition, negative affect positively predicted risk perception (𝛽 = 0.36, 𝑝 < 0.001). Both negative affect (𝛽 = 0.20, 𝑝 < 0.001) and risk perception (𝛽 = 0.17, 𝑝 < 0.001) positively predicted avoidance behavioral intention. At phase 3, greater negative affect (𝛽 = 0.24, 𝑝 < 0.001), risk perception (𝛽 = 0.08, 𝑝 < 0.001) and behavioral intention (𝛽 = 0.23, 𝑝 < 0.001) were reported if damage to the nuclear core was uncertain compared to an undamaged core. Reported negative affect (𝛽 = 0.80, 𝑝 < 0.001), risk perception (𝛽 = 0.79, 𝑝 < 0.001), and behavioral intention (𝛽 = 0.60, 𝑝 < 0.001) were positively influenced by the same construct at phase 2. In addition, negative affect predicted risk perception (𝛽 = 0.17, 𝑝 < 0.001) and behavioral intention (𝛽 = 0.27, 𝑝 < 0.001) significantly. NEAR-MISS APPRAISALS IN DISASTER EVENTS 27 Figure 2.3. SEM model with standardized parameter estimates. Notes: ** 𝑝 < 0.05; * 𝑝 < 0.1. Discussion This study explored public response to a near-miss nuclear power plant accident progression using a LOCA scenario experienced via a series of news reports unfolding over time. In the first phase, a cooling pump failure occurred; in the second phase, the cooling pump regains normal functioning; in the third phase, the nuclear facility is operating normally. The event experienced involves the possibility of loss of life, environmental damage and source term release resulting from the LOCA sequence. The event is considered a near-miss because severe consequences were possible once the LOCA was initiated, but resulted in minimal damage to the nuclear power plant reactor, no injuries or damage to the environment, and no source term release. I manipulated two attributions related to the near-miss LOCA event: (1) the attribution for initiation of the incident sequence in the first phase, and (2) the attribution for halting the incident progression in the second phase. In addition, I manipulated outcome uncertainty 0.73** 0.80** 0.11** 0.73** 0.09* 0.36** 0.20** 0.18** 0.27** 0.24** 0.60** 0.79** 0.08** 0.52** 0.17** -0.01 0.07** 0.23** 0.52** 0.60** -0.14** so0ware5failure5vs.5 earthquake5and5 aAack5 NegaEve5 Affect5 Time515 Risk5 PercepEon5 Time515 Behavioral5 IntenEon5 Time515 NegaEve5 Affect5 Time525 Risk5 PercepEon5 Time525 Behavioral5 IntenEon5 Time525 NegaEve5 Affect5 Time535 Risk5 PercepEon5 Time535 Behavioral5 IntenEon5 Time535 outcome5 uncertainty5 expert5vs.5 coincidence5 NEAR-MISS APPRAISALS IN DISASTER EVENTS 28 resulting from the near-miss event in the third phase. I evaluated public responses related to negative affect, risk perception, and avoidance behavioral intention in each phase. Results from both MANCOVA and SEM indicated that respondents’ reactions to a near- miss nuclear accident were influenced by attributions of the incident. Specifically, results from MANCOVA indicated that respondents experienced more negative affect, perceived more risk and indicated more avoidance behavioral intention when the near-miss nuclear accident was initiated by a more external attribution (i.e., cyber attack or earthquake) rather than a more internal attribution (i.e., software failure). When the incident was controlled, respondents reported less avoidance behavioral intention when a more external attribution (i.e., coincidence) averted the disaster rather than a more internal attribution (i.e., individual expert). Further, results from SEM indicated that respondents’ reactions were either directly or indirectly affected by the attribution manipulations, in the same directions suggested by MANCOVAs. Attribution theory suggests that people tend to associate negative outcomes with external attributions (Weiner, 1985, 1995). My study consistently found that when an event that could lead to serious consequences occurs, public concern is sensitive to the cause of the event and the event is viewed as more vulnerable when ascribed to a more external attribution (e.g., earthquake). People may think nuclear facilities are unable to prevent an incident from happening when chance factors cause the incident, whereas it is possible for facilities to upgrade themselves in the future if the event is caused internally. When the event is halted and serious consequences are prevented, public concerns about the event decrease. While negative affect and risk perception were independent of the attribution halting the event, greater avoidance behavior was endorsed when the attribution was more internal than external. My interpretation is that incidents that NEAR-MISS APPRAISALS IN DISASTER EVENTS 29 could be anticipated by nuclear facilities result in a greater alarm level and call to action against nuclear power compared to incidents halted by chance causes. The current study focuses on the role of attributions in terms of the internality or externality of a nuclear facility. Further research could examine the role of other aspects or details of causal attributions in predicting public response; for instance, whether the incident is caused by personnel negligence or not. People may think negligence is easy to avoid and control therefore is less risky; people may also have more negative affect toward lack of responsibility, which could further exacerbate their negative view about the risk (Siegrist & Sütterlin, 2014). Results also suggest that respondents elicited more negative emotional, cognitive, and behavioral reaction when the future environmental and health impacts of the near-miss accident remained uncertain compared to when all future impacts are completely dismissed. This is reasonable in that people would feel confident about the control and management of a LOCA event when assured that there will be no negative future impacts from the incident. The elimination of future negative consequences (certain outcome) might also increase public trust in nuclear power safety. Similarly, internal near-miss attributions could also be associated with more confidence in government regulatory control and trust in the nuclear power industry. Variables such as confidence, trust, and perceived controllability could be important in understanding my findings on causal attributions and outcome uncertainty (e.g., mediating the effects on public responses) and warrant further investigation. My findings point to dependencies among negative affect, perceived risk, and intended avoidance behaviors reported over the event sequence. Consistent with the affect heuristic (Slovic, Finucane, Peters, & MacGregor, 2007) and previous research that has examined these dependencies (Slovic et al., 2005; Terpstra, 2011; Loewenstein, Weber, Hsee, & Welch, 2001; NEAR-MISS APPRAISALS IN DISASTER EVENTS 30 Loewenstein & Lerner, 2003), I found that negative affect could influence risk perception and avoidance behavioral intention, and risk perception could influence behavioral intention. Moreover, I found that a large proportion of the variance in public responses could be explained by the same response measure reported during the previous phase. The current research suggests a framework of how the three repeated measures change over time during an ongoing incident scenario in which respondents were exposed to the three phases sequentially. Future research is needed to extend these results to a “real-time” context. I also found that individual characteristics, including sex, age, and income, predicted public response to the LOCA event. Female respondents exhibited more negative affect, perceived greater risk and avoidance behavior intentions than did males. This is consistent with previous findings that women generally report greater perceived risk than men (DeJoy, 1992; Flynn, Slovic, & Mertz, 1994). Younger respondents reported more negative affect, risk perception, and intention to pursue more avoidance behavior. Older respondents have more life experience and might have a long-term view of the risks and benefits of nuclear power and therefore are more resilient to a single event. Respondents with higher incomes perceived less risk associated with nuclear power. It may be that those with higher incomes focus more on the economic benefits associated with nuclear power and consequently minimize perceptions of risk from nuclear power. Further research is needed on the role of all three individual difference variables in predicting responses to nuclear power and near-miss power plant related disasters, such as the LOCA scenario used in the current study. In addition, I found that negative affect, risk perception, and avoidance behavioral intentions weakened as the nuclear incident was resolved over the three phases. This finding is consistent with previous studies. Cui et al. (2016) demonstrated that reactions to cumulative NEAR-MISS APPRAISALS IN DISASTER EVENTS 31 terrorist events are influenced by the trajectory of the events. Rosoff et al. (Rosoff, John, & Prager, 2012) found that respondents’ reactions (emotional, cognitive, and behavioral) to a flu epidemic intensified as the disaster escalated. One limitation of the present study is the artificial nature of the news reports. People may simply not accurately report how they would react when confronted with an actual potentially disastrous nuclear event. While it would be interesting to study the role of attributions in reactions to actual near-miss nuclear accidents, such research would not allow systematic manipulation of causal attributions for event initiation and event halting. The paradigm used for the current study sacrifices some degree of mundane realism in exchange for high experimental realism (participant engagement) and internal validity. NEAR-MISS APPRAISALS IN DISASTER EVENTS 32 Chapter 3 : Development of the Near-miss Appraisal Scale (NMAS) Using a Polytomous Item Response Theory Model Abstract Near-miss experiences have been identified as a contributing factor in responses to risk of disaster events. Researchers have found that specific characteristics of a near-miss event could lead individuals to interpret the risk as either “vulnerable” or “resilient”; moreover, these interpretations can lead to quite different decisions regarding future protective behavior. I developed a Near-miss Appraisal Scale (NMAS) to assess an individual’s tendency to interpret near-misses as vulnerable (or resilient). I developed an initial item pool of 21 items and recruited a sample of 298 respondents through Amazon Mechanical Turk. The final version of the NMAS is based on 10 of these items, following psychometric analysis for dimensionality, scale reliability, and item functioning. I establish discriminant validity of the NMAS by correlating the NMAS with scales of locus of control and risk taking, and predictive validity by using the NMAS to predict individual responses to a near-miss disaster scenario. The current study demonstrates that responses to near-misses are not only determined by the nature of the event itself, but also related to decision makers’ near-miss appraisal tendencies. Keywords: near-miss appraisal; risk; decision making; Item Response Theory; validity NEAR-MISS APPRAISALS IN DISASTER EVENTS 33 Development of the Near-miss Appraisal Scale (NMAS) Using a Polytomous Item Response Theory Model While there is considerable literature on the effect of near-miss event characterization on risk perception, limited attention has been given to individual differences in near-miss interpretation. Dillon et al. (2014) suggested that individuals could respond differently to the same near-miss event because of prior experiences serving as reference points in their interpretation of the current event. When varying the degree of vulnerability of previous near-misses, researchers found that responses to the same near-miss event differed. The current research, however, hypothesizes that the tendency to interpret a near-miss as vulnerable (or resilient) and to take precautionary measures following the near-miss is a fairly stable trait of an individual. For example, a person may more consistently view near-miss situations as vulnerable. This person’s tendencies not only depend upon her prior near-miss experience, but also are determined by a broad range of factors such as personality, socioeconomic status, personal experience, etc. I argue that a near-miss is a unique and ambiguous event that is different from a disaster or an ordinary event. It could deliver different messages depending on how close it is to being disastrous. The response to a near-miss event is also unique and involves two components. In the first component, the decision maker evaluates the near-miss information in terms of how likely a disaster could have happened. In the second component, the decision maker evaluates the connections between herself and the risk in terms of how vulnerable she is and how proactively she needs to protect herself against future disasters. I call the two-component process near-miss appraisal. The current paper describes the development of a scale to measure individual near-miss appraisal. I report on the psychometric properties of the scale, and I demonstrate that near-miss NEAR-MISS APPRAISALS IN DISASTER EVENTS 34 appraisal is a trait that determines to some extent individual tendency to interpret a near-miss as vulnerable and to take precautionary measures following a near-miss. The next section motivates near-miss appraisal as an individual trait and describes how the Near-miss Appraisal Scale is developed and evaluated. Near-miss Appraisal I propose that the near-miss appraisal trait is a type of cognitive appraisal. Cognitive appraisal refers to how an individual interprets an event and produces a response based on the interpretation (Lazarus & Folkman 1984; Lazarus, 1991, Frijda, 1986). In the primary stage, an event is interpreted as either positive or negative. If the event is viewed as negative, a coping evaluation is made in the secondary stage. For instance, when a colleague is promoted, the promotion may be interpreted as positive feelings of happiness for a friend. Conversely, the promotion could be interpreted as negative feelings resulting from disappointment in personally not receiving a deserved promotion (primary stage). If the promotion is interpreted as negative, one needs to determine how to respond. For instance, one could view the negative situation as a challenge, expect to be promoted in the future, and respond positively by working harder. However, one could also view it as a threat, expect to be fired in the future, and respond negatively by disengaging (secondary stage). Similarly, surviving a hurricane, a home owner would be expected to appraise the near-miss in two steps. She would first assess how likely the hurricane would have damaged the house. Second, the owner would evaluate how vulnerable her house is to a hurricane and whether she should relocate, purchase flood insurance, or do nothing. Tinsley et al. (2012) suggested a cognitive chain between near-miss experiences and response behaviors (Figure 3.1, dashed lines). When evaluating the risk of a disaster following a near-miss, people combine information at hand, including threat probability and consequence, NEAR-MISS APPRAISALS IN DISASTER EVENTS 35 with their near-miss experience to form an evaluation of the risk using a subjective expected utility (SEU) framework (von Winterfeldt & Edwards, 1986). This evaluation drives response behavior. I propose that each individual has her own tendencies when viewing near-miss events, which I have termed near-miss appraisal. These tendencies are expected to affect how one interprets the near-miss event, which would further influence risk assessment and behavior. My revised cognitive chain is depicted in Figure 3.1. Figure 3.1. A Framework of the determinants of response following a near-miss Researchers have identified cognitive appraisal as a subjective process. Individual differences in cognitive appraisal can be assessed using various psychometric scales (Peacock & Wong 1990; Kessler, 1998; Fromme, Katz, & Rivet, 1997) such as Stress Appraisal Measure (SAM) and Cognitive Appraisal of Health Scale (CAHS). Researchers also found that cognitive appraisal is correlated with different personality scales (Cooper, 1998; Hojat, Gonnella, Erdmann, & Vogel, 2003). For instance, appraisals of stress and coping were found to correlate with self- esteem (Rector & Roger, 1997). I believe near-miss appraisal is also a subjective process and for this reason developed the Near-miss Appraisal Scale (NMAS) to assess individual differences in interpreting near-miss events. Threat' Probability'and' Consequence' Subjec7ve' Risk' Near<miss' Event' Near<miss' Appraisal' Ac7on' NEAR-MISS APPRAISALS IN DISASTER EVENTS 36 Scale Development I assess individual appraisals of near-miss events by measuring willingness to take a protective measure after experiencing the near-miss. The items in the NMAS included a wide range of near-miss events. The items varied in three aspects: (1) nature of the risk, (2) near-miss message, and (3) possible future countermeasure behavior. First, the items involve only personal physical risks that one may encounter in daily life including transportation, health, and personal safety risk. For instance, transportation risk could involve a traffic accident following drinking alcohol; health risk could be suffering from the flu; personal safety risk could relate to an assault on the street. Second, the description of a near-miss message varies in terms of the magnitude of the consequence of the near-miss. A more severe scenario could be that you live near a seismic fault and your neighbor’s home was seriously damaged during a major earthquake while you had none; a more moderate scenario could be that there were several home break-ins in your neighborhood recently and your home has not been burglarized. Finally, a countermeasure behavior could be radical, involving perhaps moving away from a specific area; while a moderate countermeasure might be finding a companion. A more radical countermeasure indicates a more protective attitude, which indicates a more vulnerable view of the near-miss. For each item, I briefly present a situation that involves a near-miss experience and ask respondents about the strength of their intention to take a certain countermeasure in the future (or next time). For instance, in a flood risk scenario, the near-miss message is “The last flood was forecasted to have a high rise and it was recommended that you evacuate. Some structures in your neighborhood were severely damaged, but your home was fine.” The protective countermeasure for this item is to evacuate during the next flood warning. If a person intends to choose a protective action, the person views the near-miss as more vulnerable rather than more NEAR-MISS APPRAISALS IN DISASTER EVENTS 37 resilient. To assess the psychometric properties of the scale, I address issues of dimensionality using Factor Analysis (FA), scale reliability using Classical Test Theory (CTT), and individual item functioning using Item Response Theory (IRT). IRT (Embretson & Reise, 2000) allows us to analyze the performance of individual items in terms of how well each item could differentiate respondents (i.e., discrimination of the item) on the latent trait, i.e., near-miss appraisal, and how the likelihood of a respondent endorsing the item relates to her score on the latent trait (i.e., difficulty of the item). IRT also provides a score for each respondent on the underlying latent trait, near-miss appraisal. The IRT estimates of item difficulty and discrimination also allow us to evaluate Differential Item Functioning (DIF) (Holland & Wainer, 1993) for specific groups of respondents, e.g., males vs. females; old vs. young. I am interested also in the discriminant validity of the near-miss appraisal scale with respect to other personality measures, such as locus of control and risk taking. Locus of control (Rotter, 1954, 1966) describes the extent to which people believe life is contingent on what they do (internality) or on environmental factors such as fate, chance, or powerful others (externality). Levenson (1974) distinguished powerful others and chance as separate external factors and developed a three-factor locus of control scale. Many studies have found a relationship between reactions to the threat in a situation and locus of control (Anderson, 1977; Brown, Mulhern, & Joseph, 2002; Gianakos 2002). For example, Anderson (1977) found that internals were associated with less stress and more task-centered coping behaviors than were externals. In the case of a near-miss appraisal, a person who believes the disaster was avoided by sheer luck (scores high in chance factor) could be more likely to interpret a near-miss as vulnerable and engage in protective behaviors than would a person who believes she is resilient to the disaster NEAR-MISS APPRAISALS IN DISASTER EVENTS 38 risk. Risk-taking behaviors have been explained by individual differences in risk taking, which can be measured by various self-report scales (Lejuez et al., 2002; Weber, Blais, & Betz, 2002; Nicholson, Soane, Fenton-OĆreevy, & Willman, 2005). Protective (vs. risk taking) behavior following near-miss events could be related to general propensity to take risks. A person who is willing to take risks could be more likely to engage in risky behaviors following near-miss experiences than a person who is risk-averse. However, risk taking propensity and near-miss appraisal are different because the latter involves a process of dealing with new information. In this study, I evaluate the psychometric properties of the scale including dimensionality, reliability, individual item functioning, discriminant validity, and predictive validity. The rest of the paper describes the method of data collection and results from the evaluation of psychometric properties. The paper closes with a discussion of the implications and limitations of the study. Method Measures Near-miss Appraisal Scale (NMAS). The initial NMAS item pool encompassed 21 items evenly distributed across three domains: transportation (7), health (7), and personal safety (7). Respondents were asked to assess the degree of intention to adopt a particular countermeasure after experiencing a near-miss. I embedded possible countermeasures as response alternatives, and respondents indicated their intentions to adopt using a 4-point scale ranging from 1 (definitely will not) to 4 (definitely will). A full description of the 21 items is included in Table 3.1. NEAR-MISS APPRAISALS IN DISASTER EVENTS 39 Table 3.1. Near-miss Appraisal Scale Items NMAS items Transportation 1 Suppose you decided to drive home after sharing a bottle of wine at dinner. You ended up crossing the double yellow lines into oncoming traffic. You returned to your lane before hitting a car. The next time I have wine with dinner: I definitely will find another way home that does not involve my driving. 2 Suppose you did not fully stop at a stop sign and a woman was crossing the road. You almost, but did not, hit her. The next time I am at a stop sign: I definitely will stop completely. 3 Suppose you were driving on the freeway and checking your email at the same time. The car in front of you slammed on the brakes and you almost collided. The next time I am driving: I definitely will wait to check my email until I arrive at my destination. 4 Suppose as you were crossing a busy intersection you nearly got clipped by a car racing around the corner that did not see you. The next time I am at that intersection: I definitely will avoid crossing the street. 5 Suppose while walking down the street a car pulled out quickly in front of you from a blind driveway. You were not hit. The next time I am out for a walk: I definitely will stop and check driveways for cars. 6 Suppose on a recent flight you experienced terrible turbulence throughout the entire trip. The next time I plan to travel by plane: I definitely will fly on another airline. 7* Suppose you were riding a bike without wearing a helmet and a car hit you. You suffered from no injuries because the car was moving very slowly at impact. The next time I am out for a bike ride: I definitely will be sure to wear a helmet. Health 8* Suppose you and your friends did not get flu shot. All of your friends suffered from severe cases of the flu, while you only got a mild cold. Next year: I definitely will get a flu shot. 9* Suppose the last time you were sexually intimate you did not use contraception and did not acquire a sexually transmitted disease (STD). The next time I am sexually intimate: I definitely will use contraception. 10* Suppose you went out for Italian food with a friend and you both ate the same dish. Later that night he got food poisoning, but you did not. The next time I have a craving for Italian food: I definitely will go to another restaurant. 11* Suppose when petting a dog on the street it snaps at you and your friend. Your friend gets bitten, but you do not. The next time I see a dog on the street: I definitely will walk on the other side of the road. 12* Suppose you were eating at your favorite restaurant and just before biting into your food you noticed a fingernail on your plate. The waiter apologized and replaced your dish. When going out to dinner: I definitely will stop going to your favorite restaurant. 13 Suppose while you were in the bathroom at a bar, a stranger slipped a roofie into your drink. The bartender happened to see what happened and told you. The next time I am at a bar: I definitely will finish my drink before going to the bathroom. 14 Suppose your doctor found an irregular skin lesion on your arm. He ran the necessary tests and determined that the lesion was benign and non-cancerous. The next time I am out in the sun: I definitely will use sunscreen and try to stay in the shade. Personal Safety 15* Suppose you heard a news report of a mugging near your home. In the future: I definitely will plan to walk with a companion in that area. 16 Suppose a man was caught carrying a small bomb at the bus station you use to get to work every day. In the future: I definitely will use another mode of transportation (e.g., drive or ride a bike) to get to work. 17* Suppose there were several home break-ins in your neighborhood recently. Your home has not been NEAR-MISS APPRAISALS IN DISASTER EVENTS 40 burglarized. I definitely will plan to set up an alarm system in my home. 18 Suppose you live near a seismic fault and you just experienced a major earthquake. Your neighbor's home was seriously damaged, while you had none. I definitely will move away from the area near the seismic fault. 19* Suppose you live in an area subject to floods. The last flood was forecasted to have a high rise and it was recommended that you evacuate. Some structures in your neighborhood were severely damaged, but your home was fine. During the next flood warning: I definitely will evacuate. 20 Suppose the crime rate is up in your neighborhood, however you have not been personally affected. I definitely will start looking for a safer area to live. 21* Suppose when walking to your car after midnight, you notice a figure following you in the shadows. You fortunately get to your car and drive off before the stranger caught up to you. The next time I am walking to my car after midnight: I definitely will make sure to be with a friend. Note: * item retained in the final 10-item scale Measures of discriminant validity. I assessed discriminant validity between the NMAS and other established related constructs, including locus of control and risk taking measures. Locus of control scale included internality and chance subscales (17 items) from Levenson’s IPC (Internality, Powerful Others, and Chance) Scale (Levenson, 1981) 8 . The response ranged from 1 (strongly disagree) to 6 (strongly agree). The risk taking measure contained a subset (15 items) of the Domain-Specific Risk-Taking (DOSPERT) scale (Weber et al., 2002, Blais & Weber, 2006), including 12 items from health/safety and recreational domains in the revised version of DOSPERT (Blais & Weber, 2006) along with 3 items from the original version (Weber et al., 2002). Another risk taking measure contained a subset (10 items) of the Passive Risk Taking (PRT) scale (Keinan & Bereby-Meyer, 2012), including 3 items from the resources domain and 7 items from the medical domain. Responses for DOSPERT and PRT ranged from 1 (extremely unlikely) to 6 (extremely likely). Measure of predictive validity. Predictive validity was assessed as the extent to which the NMAS can predict a response from an existing near-miss scenario. I used the scenario description stimuli from Study 4 in Tinsley et al. (2012). In the scenario, respondents were 8 I did not use the subscale of power because power is obviously not relevant to near-miss appraisal. NEAR-MISS APPRAISALS IN DISASTER EVENTS 41 randomly assigned to one of the three versions of the scenario (control, resilient near-miss, vulnerable near-miss). In the control condition no near-miss experience was given; in the resilient near-miss condition, a prior near-miss experience with no disaster was presented; in the vulnerable near-miss condition, a prior near-miss experience with a near disaster was presented. Tinsley et al. (2012) found that people with resilient near-miss information were significantly less willing to forgo the cruise than people without near-miss information. People with vulnerable near-miss information were significantly more willing to forgo the cruise than people with resilient near-miss information. In the current paper, respondents were presented with the same scenario and were asked to decide whether to go on a cruise that is subject to a hurricane (response scaled from 1 = definitely will not go to 4 = definitely will go) and rate how likely a hurricane would impact their ship (response scaled from 0 = no likelihood to 100 = very likely). Respondents and Procedure Three hundred and twelve US residents were recruited through Amazon Mechanical Turk (AMT) to participate in the experiment. Each respondent earned $1 for completing the study. All respondents completed the four scales: NMAS (21 items), IPC scale (17 items), DOSPERT scale (15 items), and PRT scale (10 items). The items on each scale were randomly presented to the respondents. In addition, two items from NMAS were repeated to check for random responding. I assumed that a respondent was not focusing on the test if she switched more than one ordered category (e.g., from “definitely will not” to “probably will”) for the repeated items. Fourteen respondents were removed for failing at least one of these attention check items. The remaining sample included 298 respondents (53% female, mean age = 35.6 years). Table 3.2 summarizes demographic information of the sample. Among the 298 respondents, a subset (N = 191) were NEAR-MISS APPRAISALS IN DISASTER EVENTS 42 presented with the hurricane scenario stimuli from Tinsley et al. (2012) before responding to the 63 scale items and were included in the predictive validity analysis. Table 3.2. Demographic Information of the Sample (N = 298) Category Number and Percentage Sex Male 141 (47.3%) Female 157 (52.7%) Age 19-28 84 (28.2%) 29-34 73 (24.5%) 35-43 66 (22.1%) 44-76 74 (24.8%) Income Below $20,000 87 (29.2%) $20,000-$29,999 49 (16.4%) $30,000-$39,999 42 (14.1%) $40,000-$49,999 31 (10.4%) $50,000-$59,999 25 (8.4%) $60,000-$69,999 23 (7.7%) $70,000-$79,999 9 (3.0%) $80,000-$89,999 10 (3.4%) $90,000 or more 21 (7.0%) Education Less than high school 1 (0.3%) High school 79 (26.5%) 2-year college 59 (19.8%) 4-year college 125 (41.9%) Master’s degree 26 (8.7%) PhD degree 8 (2.7%) Race White 225 (75.5%) Hispanic 20 (6.7%) Black or African American 14 (4.7%) Asian or Pacific Islander 28 (9.4%) Two or more races 10 (3.4%) Other 1 (0.3%) NEAR-MISS APPRAISALS IN DISASTER EVENTS 43 Results Dimensionality and Reliability I conducted an exploratory factor analysis for the 21-item NMAS using the entire sample. I used Parallel Analysis (PA) (Turner, 1998) to determine the number of factors, since researchers have suggested that PA provides the most accurate approach among the rules including eigenvalue > 1 and scree plot (Henson & Roberts, 2006). Two factors were retained from parallel analysis using package ‘nFactors’ in R. Fifteen out of the 21 items had loadings > 0.30 on the primary factor (items 5, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21). The 15 items had a Cronbach’s 𝛼 of 0.79. Only one item had loadings > 0.30 on the secondary factor (item 13). Therefore a secondary dimension was not pursued. The 15 items were retained in the item response analyses for further inspection; as such, I pursued my analysis on these 15 items loading on a single dimension. Item Response Analysis I next applied Item Response Theory to examine the remaining items. I submitted the 15 unidimensional NMAS items to item response analysis using package ‘ltm’ in R (Rizopoulos, 2006). I fitted a graded response model (GRM) (Samejima, 1969) to the data, which is often used on likert-scale responses. The model is defined as follows log 𝑟 !" 1−𝑟 !" = 𝛽 ! 𝑧− 𝛽 !" where 𝑟 !" denotes the likelihood of a response in the 𝑘th category (in the current case, 𝑘 = 1, 2, 3, 4) or above (cumulative probability) to the 𝑖th item, given the latent trait 𝑧. 𝛽 ! is the discrimination parameter and 𝛽 !" is the difficulty parameter. 𝛽 ! denotes the discrimination parameter, 𝛽 !" / 𝛽 ! denotes the difficulty parameter. The unconstrained model fit significantly better than the constrained GRM with equal discrimination parameters across items (𝛽 ! = 𝛽) NEAR-MISS APPRAISALS IN DISASTER EVENTS 44 using likelihood ratio test (𝑝 < 0.001). I then inspected the 15 item characteristic curves (ICCs) generated from the unconstrained graded response model. Each ICC shows the probability of endorsing an option (e.g., each of the four options in my NMAS) at a certain latent trait level (characterized with z-score). The point where ICCs cross marks the difficulty threshold, which describes at what level of near-miss appraisal respondents would have a .5 probability of endorsing the response category or above. The steepness of the curves where they cross indicates the discrimination level, which represents how well the item could distinguish respondents at different levels of the latent trait, i.e., near-miss appraisal. Items were eliminated if they did not provide evidence of multiple efficient discriminations (i.e., two discrimination thresholds) within 5 th and 95 th percentile. Using this criterion, I dropped 5 items, items 5, 14, 16, 18, 20. Figure 3.2 illustrates an item (item 5) that failed this criterion and an item that met the criterion (item 15). Item 5 was eliminated because it had one level of discrimination where response 4 was more likely to be endorsed than all other options among 60% of respondents and responses 1 and 2 were never more likely to be endorsed than option 3. In other words, the item itself was very vulnerable and most respondents would interpret the near-miss as vulnerable with that item; therefore, it is hard to discriminate respondents on the latent trait of near-miss appraisal. In contrast, item 15 was retained because it had two points of discrimination at the 17 nd and the 72 th percentiles (standard scores = -0.95 and 0.57). Thus, the item provided information about individuals across a broad range of near-miss appraisal. It was able to discriminate respondents who were resilient to near-misses from those who were vulnerable to near-misses. The retained 10 items formed a unidimensional and efficient measure of near-miss appraisal (𝛼 = 0.75). Two sets of NMAS scores were computed: (1) IRT trait scores and (2) total scores from summing responses across the 10 items. The two sets of scores were highly correlated (r = 0.89); hence, NEAR-MISS APPRAISALS IN DISASTER EVENTS 45 the IRT trait scores were used for further analysis. Item response category characteristic curves – Item 5 Item response category characteristic curves – Item 15 Figure 3.2. ICCs of two items (item 5 and item 15) from the 21-item NMAS Differential Item Functioning (DIF) I tested DIF using the ‘mirt’ package from R (Chalmers, 2012)for different gender groups, i.e., whether the measurement instrument functioned the same way for men and women. I tested -3 -2 -1 0 1 2 3 0.0 0.2 0.4 0.6 0.8 1.0 Item Response Category Characteristic Curves - Item: X5 latent trait Probability 1 2 3 4 -3 -2 -1 0 1 2 3 0.0 0.2 0.4 0.6 0.8 1.0 Item Response Category Characteristic Curves - Item: X13 latent trait Probability 1 2 3 4 NEAR-MISS APPRAISALS IN DISASTER EVENTS 46 whether the slopes (i.e., discriminations) were the same for men and women using Wald tests with Benjamini-Hochberg adjustment (Benjamini & Hochberg, 1995). Results indicated that no items were significantly different for men and women. This suggested that the probabilities of receiving different scores across different latent trait levels of near-miss appraisal did not differ between males and females. I then tested whether the locations (i.e., difficulty thresholds) were the same for men and women. Results indicated that 5 items (item 7, 9, 15, 19, and 21) were significantly different for men and women (𝑝 < 0.05). Inspecting the 5 items, I found that at a given latent trait level for near-miss appraisal, men would likely to receive higher scores than women. I also tested DIF for different age, education, income, and race groups. I identified two age groups (younger: age ≤ 33 years; older: age > 33 years), which partitioned the sample into two approximately equal group (97 and 93 respondents respectively), two income groups (low income: income below $30,000; high income: income above $30,000), two education groups (low education: education below 4-year college; high education: education above 4-year college), and two race groups (white and non-white). Results indicated that for both the slopes and locations tests, none of the items were significantly different for respondents in different age, education, income, and race groups. This suggested that the probabilities of receiving different latent trait scores for near-miss appraisal did not differ across age, education, income, and race groups. At a given near-miss appraisal level, younger and older respondents, respondents with low income and high income, respondents with low education and high education, white and non-white did not receive different scores on these items. NEAR-MISS APPRAISALS IN DISASTER EVENTS 47 Effects of Demographic Variables Since DIF is an indicator that people of the same latent trait in different groups receive different trait scores, lack of substantial evidence of DIF indicates that the trait score is a reliable representation of the latent trait. Therefore, I used the NMAS to compare individuals in different gender, age, income, education and race ranges, i.e., to explore whether people have different near-miss appraisal traits. I applied regressions to investigate the effects of demographic variables on the NMAS with the entire sample. The NMAS IRT score was used as the dependent variable; gender, age, income, education and race were included as binary predictors. Results showed that females had a higher score on the NMAS (β = 0.40, 𝑝 < 0.001), indicating that females were more likely to view a near-miss as vulnerable and to take protective action. Respondents with high income had higher NMAS scores than those with low income (β = 0.12, 𝑝 = 0.02), indicating that they were more likely to view a near-miss as a signal of vulnerability and engage in protective behaviors. Results also showed that non-white respondents had a higher score of NMAS (β = 0.19, 𝑝 < 0.001), indicating that non-white respondents tended to view a near-miss as evidence of vulnerability, and thus engage in avoidant behavior. NMAS scores did not differ between young and old (β = -0.04, 𝑝 = 0.31), or between those with low education vs. high education (β = -0.10, 𝑝 = 0.07). Discriminant Validity I assessed discriminant validity by examining the correlations between the final 10-item NMAS and scales for other related constructs, including risk taking (DOSPERT), passive risk taking (PRT), and locus of control (Levenson IPC). The descriptive statistics for various measures are summarized in Table 3.3. The NMAS was negatively correlated with the DOSPERT score (15 items) (r = -0.43, 𝑝 < 0.001), indicating that a higher NMAS score NEAR-MISS APPRAISALS IN DISASTER EVENTS 48 (vulnerable appraisal) was associated with greater risk-averse attitude. The medium-level correlation (accounting for less than 20% of the variance) suggested that there were other factors that account for individual differences on the NMAS other than risk taking. Similarly, the NMAS was positively correlated with PRT (10 items) (r = 0.30, 𝑝 < 0.001), accounting for less than 10% of the variance. The NMAS was also correlated with the chance subscale (9 items) of the Levenson IPC scale (r = 0.17, 𝑝 = 0.002), accounting for less than 4% of the variance. The NMAS was not correlated with the internality subscale (8 items) of Levenson IPC scale (r = 0.06, 𝑝 = 0.26). This suggested that near-miss appraisal was weakly related to the chance dimension of IPC and was linearly unrelated to the internality dimension of IPC. Table 3.3. Mean and Standard Deviation of the Measures Measure Mean S.D. NMAS -0.02 0.88 Internality 4.42 0.68 Chance 3.14 0.81 DOSPERT 2.63 0.86 PRT 3.6 0.73 Predictive Validity I next examined whether NMAS could predict response to a simulated near-miss scenario involving a hurricane (Tinsley et al., 2012) with 191 respondents 9 . I assessed the extent to which the NMAS score could predict the intention of going on a cruise that might sail into a hurricane (as a proxy of behavioral intention) and the rating of belief that a hurricane would impact the ship (as a proxy of risk perception). I regressed behavioral intention and risk perception on the NMAS score and the manipulated variable (control vs. resilient near-miss vs. vulnerable near- 9 The data was collected in two separate waves and only the second group was asked to respond to the near-miss scenario in Tinsley et al. (2012). NEAR-MISS APPRAISALS IN DISASTER EVENTS 49 miss). Results indicated that the NMAS score was a significant predictor of both behavioral intention (β = -0.17, 𝑝 = 0.03) and risk perception (β = 0.16, 𝑝 = 0.02). However, the manipulation of the near-miss experience did not predict behavioral intention or risk perception over and above the individual NMAS scores of near-miss appraisal. I also assessed the extent to which the DOSPERT and PRT scores could predict responses in the hurricane scenario since DOSPERT and PRT were moderately correlated with NMAS score. DOSPERT significantly predicted behavioral intention (β = 0.01, 𝑝 = 0.02) but did not predict risk perception (β = -0.001, 𝑝 = 0.83), while PRT predicted neither behavioral intention (β = 0.003, 𝑝 = 0.64) nor risk perception (β = 0.002, 𝑝 = 0.77). Overall, I constructed a 10-item Near-miss Appraisal Scale with a polynomous Item Response Theory model. Analysis of Differential Item Functioning indicated that the scale is a reliable representation of the latent trait, near-miss appraisal. I found that female, higher income and non-white respondents tended to have higher NMAS scores compared to male, lower income, white respondents. Furthermore, near-miss appraisal is a separate construct, demonstrating discriminant validity from other established scales of locus of control and risk taking, and was predictive of responses to an independent near-miss scenario used in previous published research. Discussion The present study demonstrates that individuals differ in interpreting, evaluating and responding to near-miss events that involve physical risks by establishing the efficacy of a psychological scale – Near-miss Appraisal Scale (NMAS) --- to measure the degree that a person interprets near-misses as vulnerable and responds protectively. I established that responses to near-miss events involve a unique cognitive appraisal process, overlapping with but distinct from measures of locus of control and risk taking. The scale is able to discriminate individuals at NEAR-MISS APPRAISALS IN DISASTER EVENTS 50 different trait levels of near-miss appraisal. The scale predicts respondents’ risk perception and behavioral intention when encountering a potential hazard, even if the near-miss messages could not differentiate respondents’ responses. The current study established a framework describing the cognitive process of how near- miss events could influence the decisions regarding a disaster. Tinsley et al. (2012) suggested a cognitive chain in which people use their personal knowledge of the disaster, essentially near- miss experiences, to form the assessments of risk and then choose the behavior to engage. The current framework extends that cognitive chain, suggesting that people interpret near-miss experiences with a near-miss appraisal process that differs across individuals and results in different subjective risk formations and behaviors. The current scale focuses only on near-misses involved in physical risks. Hence, the scale is used for characterizing the individual differences in interpreting near-misses when a disaster could have happened; I did not consider other types of near-misses such as those found in gambling (Sundali, Safford, & Croson, 2012). Furthermore, I suspect that individual tendencies in interpreting near-misses involved in physical risks are different from those involved in other risks. For instance, interpreting a near-miss in a financial domain (e.g., betting behavior) might be more related to risk-taking propensity. In fact, Blais and Weber (2006) found that intrapersonal variation in risk taking across different domains (e.g., financial, safety) is much larger than interpersonal variation. Further research is warranted to explore intrapersonal variation in interpreting near-misses for different types of risks. I believe near-miss appraisal is a person’s overall tendency in interpreting near-misses. I found the effect of demographic variables on the scale score such that female, high-income and non-white respondents tended to have higher NMAS scores. It would be useful to examine the NEAR-MISS APPRAISALS IN DISASTER EVENTS 51 relationship between near-miss appraisal and other individual factors. For instance, studies on recovery after traumatic events found that those with higher coping self-efficacy are more resilient (Kessler, Galea, Jones, & Parker, 2006; Bonanno, Brewin, Kaniasty, & Greca, 2010). Self-efficacy refers to one’s belief in her own competence in specific situations (Bandura, 1977), whereas locus of control relates to more general beliefs about control. It is important to establish that near-miss appraisal is a separate construct, and not simply another means of conceptualizing and measuring self-efficacy. I used the behavioral response following a near-miss as a proxy of the extent to which the person interpreted the event as vulnerable vs. resilient. There could be some elements that also contribute to the behavioral response other than near-miss interpretation, such as self-efficacy and motivation for taking actions. More research could be done to assess these possible dependencies. Although the predictive validity of the scale is tested by responses to a common near- miss scenario, the scenario is hypothetical and the responses are self-reported. Future studies should validate the NMAS in a variety of other realistic settings. For example, I could examine the scale by transportation workers’ near-miss appraisal tendencies and examine how they could influence their interpretations of past near-misses and actual reactions to new near-misses. General Implications Major organizational decisions under risk and uncertainty are often made by a small group of individuals who gather information from multiple resources. The decision could be based on scientific assessments of the risk and the decision maker’s own judgments that can be influenced by past experience as well as individual tendencies to appraise the experience. The current research provides a useful measure to assess the subjective evaluation of common near- miss experiences. My finding suggests that decisions following a near-miss experience may be NEAR-MISS APPRAISALS IN DISASTER EVENTS 52 better predicted by individual differences in how a decision maker typically appraises a near- miss than the specific characteristics of the near-miss event. Organizations responsible for risk communication should be cautious in broadcasting a message regarding a near-miss event. People may not respond uniformly to a description of a past near-miss event. Specifically, a near-miss description that is interpreted as vulnerable only by a small group of people (e.g., an item with a high threshold of discriminating people with high NMAS scores from those with low NMAS scores) may not make people vigilant. In contrast, a near-miss description that is largely interpreted as vulnerable may terrify too many people and impact the economy. Moreover, a near-miss description could vary in multiple aspects including category of the risk, vulnerability of the outcome, and suggested countermeasures. Future research is needed to clarify the relationship between these aspects of near-miss messages and near-miss interpretations. NEAR-MISS APPRAISALS IN DISASTER EVENTS 53 Chapter 4 : Effects of Psychological Distance on Near-miss Appraisals Abstract Previous studies have demonstrated that near-miss events could lead individuals to interpret the risk as either vulnerable (i.e., disaster that almost happened), requiring precautionary measures, or resilient (i.e., disaster that could have happened but did not), resulting in no protective action. The current study hypothesized that responses to near-misses are determined by the psychological distance (PD) between the decision maker and near-miss events. I conducted an experiment that evaluated responses to a single near-miss event (a near-miss terrorist attack during a public event) with a 2 (spatial PD: proximal vs. distal) by 2 (social PD: high vs. low) between-subjects design. Each respondent from New York or Los Angeles was randomly presented with one scenario that took place in one of the two cities. Results indicated that spatial PD significantly predicted risk perception and behavior, and social PD significantly predicted risk perception. Results also indicated that respondents perceived less risk and more concrete benefits of a public event when a near-miss attack was spatially far away. Keywords: near-miss, psychological distance, risk perception, behavioral intention NEAR-MISS APPRAISALS IN DISASTER EVENTS 54 Effects of Psychological Distance and Cumulative Sequences on Near-miss Appraisals In this study, I am interested in a subjective factor, psychological distance, and whether and how it could affect near-miss appraisals. Psychological distance is “a subjective experience that something is close or far away from the self, here, and now” (Trope & Liberman, 2010). Along with the actual attributes of an object or event, psychological distance affects the way people construe the object or event. For example, a disaster that happened either yesterday or last year, or in one’s current city or another country, or to one’s relative or someone unknown, could make the person feel differently about the disaster, even though the actual attributes such as cause and consequences of the disaster unchanged. Psychological distance could affect people’s judgments and behaviors through their construals of an event or an action, which is illustrated in Construal Level Theory (CLT) (Trope, Liberman, & Wakslak, 2007; Trope & Liberman, 2010). CLT identified four types of psychological distance (PD): temporal, spatial, social, and hypothetical. Temporal PD means how far from now the event is to the decision maker (e.g., “tomorrow” or “next year”). Spatial PD denotes the physical/geographical distance between the event and the decision maker (e.g., “here” or “there”). Social PD is the similarity between the person who experiences the event and the decision maker (e.g., “us” or “others”). Hypothetical PD elicits how probable the decision maker believes the event will happen (e.g., “sure” or “maybe”). Studies consistently found that people would construe an event that is psychologically proximal (e.g., events that happen tomorrow, here, to us, for sure) at an abstract and superordinate level, whereas they construe an event that is psychologically distant (e.g., events that happen next year, there, to others, maybe) at a concrete and subordinate level. For example, Henderson, Fujita, Trope, and Liberman (2006) suggested that people would focus more on the global trend (abstract level) as psychological NEAR-MISS APPRAISALS IN DISASTER EVENTS 55 distance increases. Respondents from New York University viewed graphs of years of information about either the university campus in Manhattan (spatially proximal) or in Florence, Italy (spatially distant) over years. The graphs had either upward or downward global trends with the last year deviating from the global trend. Respondents estimated how likely the next year would follow the global trend or not. Researchers found that respondents in the spatially distant condition exhibited a greater tendency to follow the global trend. Moreover, CLT suggested that people were more likely to generate pros rather than cons toward a psychologically distant event and thereby had more favorable attitudes. As cons are subordinate to pros, the latter become more salient as psychological distance increases. For example, Eyal, Liberman, Trope, and Walther (2004) showed that when asked to generate the pros and cons of proposed actions, respondents generated more arguments in favor of distant future actions and produced more arguments against near future actions. Herzog, Hansen, and Wanke (2007) also found that respondents had more favorable attitudes toward a proposed action when it was to happen in the distant future. Similarly, risk perception associated with a psychological distant action decreases since positive features become more salient as the psychological distance increases. For instance, Chandran and Menon (2004) demonstrated that temporal distance could influence the concreteness and proximity of health risks, which would further affect risk perception and behavioral intention. Respondents were told that people died of certain diseases every day (vs. every year. They reported greater concern and anxiety about the health risks and were more willing to engage in precautionary measures. In another study, Li et al. (2010) looked at the overconfidence of respondents who lived in a disaster-hit area in 2008 China and those who lived in other areas. They found that respondents in the disaster-hit areas (i.e., spatially proximal) NEAR-MISS APPRAISALS IN DISASTER EVENTS 56 estimated their rank of being infected by a disease significantly lower than those in the non- disaster area (i.e., spatially distant), indicating that respondents in the spatially proximal condition were more negative. In the case of near-misses, I anticipated that people would be more likely to view the positive side (how the disaster did not happen) when the near-miss is psychologically distant, and therefore be more resilient to the near-miss. Furthermore, they would perceive less risk about the near-miss event and less willing to take proactive measures following the event. This could also be explained as follows. When the near-miss is psychologically close, people are more likely to construe the risk related to the near-miss event more concretely and therefore perceive greater risk. When the near-miss is psychologically distant, people are more likely to construe the benefit more concretely and therefore perceive greater benefit. The current study examined the effect of psychological distance on responses to a one- time near-miss event. Using a scenario of a near-miss terrorist attack, I focused on two types of psychological distance, social and spatial. In previous studies, social PD was manipulated as interpersonal similarity between the respondent and a target person (Liviatan, Trope, & Liberman, 2008). Subjects received a description of a target person who was either similar or dissimilar to them and were asked to predict the target person’s attitudes and behaviors. For example, Liviatan et al. (2008) showed respondents a list of classes the target person attended. In the similar condition, respondents had taken most of the classes on the list; in the dissimilar condition, respondents had taken none of the classes on the list. Researchers found that respondents in the similar condition were more likely to identify behaviors performed by the target person in a low-level construal (i.e., a concrete description of “how” instead of an abstract description of “why”). Similarly, in the NEAR-MISS APPRAISALS IN DISASTER EVENTS 57 current study I described a hypothetical target person who experienced a near-miss terrorist attack and I manipulated the social similarity between the target person and the respondents. I anticipated that when the target was socially dissimilar to the respondents, respondents would view the near-miss to be more distant and resilient. As a result, they would think more abstractly of the target’s risk and predict that the target would perceive less risk and would be less likely to engage in mitigation behaviors. At the same time, I manipulated spatial PD as the physical distance between the target person and the terrorist attack. When the target person was geographically close to the attack, respondents would view the target person to be more vulnerable to the near-miss event and therefore perceive more risk and more likely to engage in mitigation behaviors. Method Each respondent was randomly assigned to one of four conditions in a 2 (spatial PD: high or low) by 2 (social PD: high or low) between-subjects design. Spatial PD was represented by whether the target person and the terrorist attack were at the same place. Social PD was represented by whether the respondent was socially close to the target person. The respondent and target were socially close if they were located at the same place, had the same gender (a male target named “Chris Williams” or a female target named “Laura Johnson”), and were within the same age range. Each respondent was randomly presented with one of the two scenarios (movie theater or concert) that took place in either New York City or Los Angeles. The two scenarios and manipulations are shown in Table 4.1. For instance, if a respondent was from Los Angeles and received a scenario of an attack taking place in a movie theatre in the city, the high level of spatial PD and high level of social PD would be that the target person was located NEAR-MISS APPRAISALS IN DISASTER EVENTS 58 in New York City, and with the opposite gender and in a different age range compared to the respondent. Table 4.1. Descriptions of two scenarios and manipulations Movie Theater Concert Scenario On November 13, a mass shooting occurred in theater 9 at the Century 21 movie theater, located in the Town Amo shopping center in {Los Angeles, CA/New York city}, during a screening of the film The Lightman Returns. A gunman, dressed in tactical clothing, set off tear gas grenades and shot into the audience with multiple firearms. Twelve people were killed and seventy others were injured. The sole assailant, Eagan Harris, was arrested in his car parked outside the cinema minutes later. It was the deadliest mass shooting by a single gunman in the U.S. since the September 11 attacks in 2001. The Town Amo shopping center is a three-level regional shopping center, featuring three department stores, more than 200 retailers, multiple full-service restaurants, and a Century 21 Theaters multi-complex. On the evening of November 13, a series of coordinated terrorist attacks occurred in {Los Angeles, CA/New York city}. Beginning at 09:20 pm, three suicide bombers struck outside the Belays Center during a concert by the band Moley Cruell, followed by suicide bombings and mass shootings at cafes and restaurants. There were about 10,000 people in the arena when the bombings happened. The attacks killed 26 civilians and injured an estimated 127 others. The attacks were the deadliest terrorist attack in the U.S. since the September 11 attacks in 2001. The Belays Center is a multi-purpose indoor arena located in Los Angeles, hosting events including basketball games, hockey games, concerts, and many other entertainment events. Spatial PD: low {Chris Williams/Laura Johnson}, {age}, who lives in {Los Angeles/New York city}, goes to a movie theater almost every Saturday night. {He/she} was watching another movie in theater 1 at the same movie theater when the shooting happened on the night of November 13. The multiplex's fire alarm system began sounding soon after the attack began and {He/she} left the theater safely following the security’s guidance. {Chris Williams/Laura Johnson}, {age}, who lives in {Los Angeles/New York city}, is a fan of the band Moley Cruell. On the night of November 13, {he/she} went to Moley Cruell's concert at the Belays Center with {his/her} friends. The concert was interrupted by the suicide bombings exploded outside the arena. They left the arena safely following the security's guidance after the attacks. Spatial PD: high {Chris Williams/Laura Johnson}, {age}, who lives in {New York city/ Los Angeles}, goes to a movie theater almost every Saturday night. {He/she} was watching the film The Lightman Returns in a movie theater in {NY/LA} on the night of November 13. {He/she} heard the news of the shooting at Century 21 in {LA/NY} when {he/she} got back home {Chris Williams/Laura Johnson}, {age}, who lives in {New York city/Los Angeles}, is a music fan. {He/she} often goes to concerts with {his/her} friends. On the night of November 13, {He/she} went to the concert of Selena Adagio with {his/her} friends at Staphs Center in {NY/LA}. {He/she} heard news of the bombings in {Los NEAR-MISS APPRAISALS IN DISASTER EVENTS 59 from the movie. Angeles/New York} when {he/she} got back home from the concert. Procedure and Measures I used Turkprime, a toolkit for Amazon Mechanical Turk that runs MTurk studies (Litman, Robinson, & Abberbock, 2017), to recruit respondents from the state of California and New York. Respondents were asked to click on a survey link to learn whether they were qualified to participate in the experiment. Respondents’ locations at the time were determined by their IP addresses and only those who were located in the Los Angeles and New York City regions could continue to take the survey. Qualified respondents were then asked to report their demographic information. A similar or dissimilar target was then generated based on the respondent’s gender, age, and location. Respondents then read about a scenario of an attack. Following the scenario they were asked where the scenario took place. Respondents then read a near-miss experience of the target person (either Laura Johnson or John Smith). They were then asked where the target person was located. Subjects were then asked to predict the target’s behavior and perceived risk. For instance, respondents who received a scenario of an attack in a movie theater were asked 5 questions about how likely they thought the target person would not go to a movie theater (1=extremely likely to 6 = extremely unlikely) and 5 questions about how likely the target person would believe there would be a terrorist attack (1=extremely likely to 6 = extremely unlikely). Table 4.2 lists the 10 questions. The two sets of questions formed two factors and were internally consistent (𝛼 = 0.93 and 0.87). Therefore I created composite scores for predicted behavior and risk perception by averaging the scores in each factor. NEAR-MISS APPRAISALS IN DISASTER EVENTS 60 Table 4.2. Questions about predicted target’s behavior and risk perception Predicted target’s behavior 1 How likely do you think Laura Johnson /John Smith would NOT go to a movie theater in the next month? 2 How likely do you think Laura Johnson /John Smith would NOT go to a movie theater in the next year? 3 How likely do you think Laura Johnson /John Smith would NEVER go to a movie theater? 4 How likely do you think Laura Johnson /John Smith would go to a movie theater less frequently than before over the next month? 5 How likely do you think Laura Johnson /John Smith would go to a movie theater less frequently than before over the next year? Predicted target’s risk perception 1 How likely would Laura Johnson /John Smith believe there will be a terrorist attack at a movie theater in the United States over the NEXT MONTH? 2 How likely would Laura Johnson /John Smith believe there will be a terrorist attack at a movie theater in the United States over the NEXT YEAR? 3 How likely would Laura Johnson /John Smith believe there will be a terrorist attack at a shopping center in the United States over the NEXT MONTH? 4 How likely would Laura Johnson /John Smith believe there will be a terrorist attack at a shopping center in the United States over the NEXT YEAR? 5 How likely would Laura Johnson /John Smith believe there will be a terrorist attack over the NEXT YEAR? Subjects were also asked to what extent they thought the target person would agree with 10 statements about the risks and benefits regarding the scenario (1 = strongly disagree to 6 = strongly agree). The 10 statements for the concert scenario are shown in Table 4.3. The first six statements are about the benefits of going to a concert with the first three more personal and the other three more general. The remaining four statements are about the risks or concerns about going to a concert with two of them more personal and the other two more general. The order of the 10 statements was randomized. The 6 statements about benefits and 4 statements about risks were internally consistent (𝛼= 0.80 and 0.85). I created composite scores for perceived benefits and risks by averaging the scores in each. NEAR-MISS APPRAISALS IN DISASTER EVENTS 61 Table 4.3. Statements about predicted target’s perceived risks and benefits for the concert scenario Risks and benefits statements for the concert scenario 1 I can hear my favorite music LIVE at a concert. 2 I can meet cool people who share the same interest as me at a concert. 3 I can see someone live at a concert after listening to them for a long time. 4 People go to a concert for live music. 5 People make cool friends at a concert. 6 People go to a concert to see celebrities. 7 I feel I am in danger if I go to a concert because there is not enough security. 8 I feel I am in danger if I go to a concert because people are not bag-checked. 9 Concerts have become targets of terrorist attacks. 10 Public entertainment and activities have become targets of terrorist attacks. I also created a composite manipulation-check score based on two questions about how socially close respondents were to the target person: “How similar do you think Chris Williams/Laura Johnson) is to yourself?” and “How close do you feel to Chris Williams/Laura Johnson?” (1 = a great deal to 5 = not at all, r = 0.82). After answering the two manipulation check questions, respondents answered four questions about their experience: How often do you go to a movie theater/shopping center/concert/sports game? (1 = never to 4 = almost every week). Respondents then took the Near-miss Appraisal Scale (NMAS) developed in Chapter 3. A composite score was calculated to represent each respondent’s near-miss appraisal. Finally, respondents answered the question “in what city do you live over the past two years?” Respondents A total of 409 respondents were recruited from Amazon Mechanical Turk. Since four- minutes is the expected minimum amount of time respondents should spend on the survey, 33 respondents were removed for completing the survey in under 4 minutes. In addition, seventeen respondents were removed by failing the first attention check question (location of the scenario), and another seven respondents were removed by failing the second attention check question NEAR-MISS APPRAISALS IN DISASTER EVENTS 62 (location of the target). Another three respondents were removed because they checked “never” to all four experience questions. One respondent was removed because her location at the time was NYC but she had spent the previous two years in California. The final sample consisted of 348 respondents in total. Table 4.4 summarizes the demographic information of the sample. Table 4.4. Demographic information of the sample (N = 348) Variables Values Frequencies and Percentages Sex Male 163 (46.8%) Female 185 (53.2%) Income < $20,000 70 (20.1%) $20,000-$40,000 89 (25.6%) $40,000-$60,000 81 (23.3%) $60,000-$80,000 44 (12.6%) > $80,000 54 (15.5%) Age 21-30 80 (23.0%) 31-36 97 (27.9%) 35-42 84 (24.1%) > 42 87 (25.0%) Education High school and below 77 (22.1%) 2-year college 69 (19.8%) 4-year college 136 (39.1%) Master’s degree and above 65 (18.7%) Location Los Angeles 202 (58.0%) New York 146 (42.0%) Results An independent sample t-test on the social distance composite score verified that respondents perceived the target person who lived in another city, of the same gender and similar age, as socially farther away than the target person who lived in the another city, of the opposite gender and dissimilar age (𝑝 = 0.050). A 2 (spatial PD) x 2 (social PD) between-subjects analysis NEAR-MISS APPRAISALS IN DISASTER EVENTS 63 of covariance was performed on predicted behavior and risk perception. Adjustment was made for the NMAS score. Results indicated that spatial PD significantly predicted risk perception (𝐹(1,338) = 37.88, 𝑝 < 0.001, 𝜂 ! = 0.10) and behavior (𝐹(1,338) = 97.54, 𝑝 < 0.001, 𝜂 ! = 0.22). When the target avoided an attack that was geographically farther away, respondents predicted the target would perceive less risk and be more likely to continue the activity (going to a concert or a movie theater) than would the target who avoided an attack that was geographically close. Social PD significantly predicted risk perception (𝐹(1,338) = 4.41, 𝑝 = 0.04, 𝜂 ! = 0.01) but did not significantly predict behavior (𝐹(1,338) < 1, 𝑝 = 0.44, 𝜂 ! = 0.002). When the target experiencing a near-miss was socially close to the respondent, the respondent predicted the target would perceive more risk than would the socially distant target. Moreover, respondents’ near- miss appraisal significantly influenced their evaluations of the target’s risk perception (𝐹(1,338) = 8.34, p = 0.004, 𝜂 ! = 0.02) and behavior (𝐹(1,338) = 8.82, p = 0.003, 𝜂 ! = 0.03). Respondents who were more vulnerable to near misses predicted the target person would perceive more risk and be more likely to stop the activity. Figure 4.1 shows plots of mean predicted risk perception and behavior by spatial PD and social PD. Figure 4.1. Mean predicted risk perception and behavioral intention to avoid the target location by spatial PD and social PD Since all respondents evaluated the target person’s perceived benefits and risks of going to a concert or movie theater, I performed a mixed-design analysis of variance with perception of 3" 3.5" 4" 4.5" 5" 5.5" 6" low" high" Risk%Percep+on%(106)% Spa+al% Mean%Risk%Percep+on% low" high" Social% 3" 3.5" 4" 4.5" 5" 5.5" 6" low" high" Behavioral*Inten.on*(116)* Spa.al* Mean*Behavioral*Inten.on* low" high" Social* NEAR-MISS APPRAISALS IN DISASTER EVENTS 64 the activity (benefits vs. risks) as the within-subjects measure and spatial PD and social PD as between-subjects measures. Results indicated that perceived benefits were significantly higher than perceived risks (𝐹(1,342) = 144.11, 𝑝 < 0.001, 𝜂 ! = 0.30). In addition, results indicated that there was a significant interaction effect between spatial PD and perceived benefits and risks (𝐹(1,342) = 36.98, 𝑝 < 0.001, 𝜂 ! = 0.10). When the near-miss experience was geographically far away, respondents indicated greater identification with the benefits compared to the risks of the activity (as shown in Figure 4.2). I also performed mixed-design ANOVAs with perceived benefits and risks (concrete vs. abstract) as within-subjects measures and spatial PD and social PD as between-subjects measures. Results indicated that there was a significant interaction effect between spatial PD and concreteness of perceived benefits (𝐹(1,342) = 36.98, 𝑝 < 0.001, 𝜂 ! = 0.02). Respondents indicated greater agreement with concrete rather than general benefits when the near-miss was geographically far away (as shown in Figure 4.3). Figure 4.2. Mean scores of perceived benefits and risks by spatial PD. 3" 3.5" 4" 4.5" 5" 5.5" 6" low" high" Percep&on)(1,6)) Spa&al) Mean)Perceived)Benefits)and)Risks)) benefit" risk" Percep&on) NEAR-MISS APPRAISALS IN DISASTER EVENTS 65 Figure 4.3. Mean scores of perceived benefits by spatial PD and concreteness. Discussion Near-miss events are far more prevalent than a disaster event. Responses to near-miss events are often ambiguous since people do not always react in a protective way when knowing that a disaster could have happened. Previous studies of near-misses introduced the resilient vs. the vulnerable near-miss but did not specify the characteristics of a resilient (or vulnerable) near- miss. This study proposed that a psychologically proximal near-miss could make people feel vulnerable to the event, while a psychologically distant near-miss with high psychological distance could make people feel resilient to the event. I tested this hypothesis with a behavioral experiment that simulated a hypothetical near-miss event. I found that respondents’ evaluations of a near-miss terrorist attack were influenced by both spatial and social psychological distance. Respondents’ predictions of the likelihood of an upcoming terrorist attack and of the target person taking the safe option significantly decreased as the geographic distance increased (i.e., previous terrorist attack happened in a different city from where the target person was located). Risk perception is also influenced by social psychological distance. When the target person who experienced a near-miss terrorist attack and the respondents were socially similar, respondents predicted a higher likelihood of an upcoming terrorist attack. 3" 3.5" 4" 4.5" 5" 5.5" 6" low" high" Percep&on)(1,6)) Spa&al) Mean)Perceived)Benefits) concrete" abstract" Perceived)Benefits) NEAR-MISS APPRAISALS IN DISASTER EVENTS 66 As argued by CLT that cons are subordinate to pros, risks are subordinate to benefits. As a result, people would think more about the positive aspects than the negative aspects when considering whether to engage in an action. Indeed, results showed that when the near-miss experience was geographically far away, respondents were more likely to consider benefits of the activity rather than risks. Furthermore, results indicated that respondents were more likely to consider concrete rather than general benefits as the geographic distance of a near-miss increased. I believe the change in psychological distance on any dimension (spatial, social, hypothetical) could move respondents’ attitudes (more positive or more negative) toward a near- miss (i.e., placement in the range of near-miss appraisal, as shown in Figure 4.4) and become more vulnerable or more resilient to a near-miss, and therefore change respondents’ perceived risks and behavioral intentions. Figure 4.4. An illustration of how psychological distance could change near-miss appraisal As suggested by CLT, different dimensions of psychological distance are interrelated (Boroditsky, 2007; Bar-Anan, Liberman, Trope, & Algom, 2007; Stephan, Liberman, & Trope, 2010). Psychological distance in one dimension should bring to mind the distance of another dimension. In the context of a near-miss, spatial and social proximity could imply that a disaster is more probable (i.e., hypothetical proximity) and any change in the three dimensions toward proximity could make respondents more sensitive to the risks and become more vulnerable. Moreover, I can consider that a negative outcome is temporally distant to a person who has PD PD Resilient* Vulnerable* Near1Miss* Apprisal* NEAR-MISS APPRAISALS IN DISASTER EVENTS 67 experienced many near-miss events. This is because a disaster becomes less and less probable (i.e., hypothetically distant) as the number of experienced near-misses increases and people tend to think positively and become more and more resilient. A vulnerable near-miss introduced by Tinsley et al. (2012) that delineated possible bad outcomes could also be considered as a near- miss with a lower hypothetical psychological distance compared to a resilient near-miss that contained no details of possible outcomes. Various methods could alter psychological distance, which would essentially manipulate level of construal. In a study about climate change, Jones, Hine, and Marks (Jones, Hine, & Marks, 2016) demonstrated that the effect of manipulated psychological distance on public concerns and behavioral intentions associated with climate change could be fully mediated by respondents’ general construal levels of climate change risk. In the context of a near-miss, further research is warranted to test whether psychological distance would change factors other than the construal level that could also influence risk perception and behavioral intention. For example, a manipulation of how likely bad outcomes could happen to an individual might also change the individuals’ subjective probability assessment of a disaster. NEAR-MISS APPRAISALS IN DISASTER EVENTS 68 Chapter 5 : Near-miss Appraisals in Sequential Near-miss Events Abstract The current study includes three experiments on responses to sequential near-misses. Respondents were exposed to a sequence of 20 events. The 20 events varied in outcomes of non- disaster, disaster, and near-miss. In Experiments I and II, each sequence included near-misses in which two dimensions of psychological distance (PD) were manipulated using a 2 (spatial PD: low vs. high) by 3 (hypothetical PD: low vs. medium vs. high) within-subjects design. Results indicated that respondents predicted less risk of a disaster and were less likely to engage in protective measures when a near-miss event was psychologically distant (either geographically or probabilistically) to the decision maker. Results also indicated that a disaster would be viewed as more resilient with the accumulation of near-misses over time. In Experiment III, the 20-event sequence included near-misses in which spatial and hypothetical PD were manipulated with a 2 by 3 between-subjects design. Results indicated that respondents updated their probability beliefs of disaster events following near-misses. Respondents who experience vulnerable near-misses would be more likely to count the near-misses as hits. Keywords: near-miss, sequence, psychological distance, Beta-binomial NEAR-MISS APPRAISALS IN DISASTER EVENTS 69 Near-miss Appraisals in Sequential Near-miss Events People often need to make predictions or decisions based on a sequence of past outcomes. In the current study, I am interested in responses to sequential near-misses. For example, do people continue purchasing flood insurance after escaping multiple floods that nearly resulted in disaster in the past. The past experienced sequence of outcomes could direct expectations of future outcomes and their behaviors in preparation of future outcomes. People could expect that the opposite outcome would follow a previous event (negative recency, or gambler’s fallacy) or expect that the same outcome would follow a previous event (positive recency, or hothand). Research has demonstrated that without the knowledge of genuine randomness (e.g., coin tosses, birth), positive recency is a psychological default (Wilke & Barrett, 2009; Scheibehenne, Wilke, & Todd, 2011; Tyszka, Zielonka, Dacey, & Sawicki, 2008). For example, Wilke and Barrett (2009) found that respondents expected positive recency on sequences of various resources, including seeing a bird in a nest or finding parking spots in a city. Lacking information pertaining to the underlying process, people predominately perceive a sequence as nonrandom (Tyszka et al., 2008). They are more likely to detect the positive recency pattern in a sequence than they are negative recency, and to predict and learn according to positive recency (Scheibehenne et al., 2011). Studies have demonstrated that responses to rare events (i.e., events with a low probability, such as disasters) also exhibit positive recency. If positive recency holds, a rare event will be considered less likely to occur when it is not recent; a rare event will be more likely to reoccur once it has occurred in a given area. Barron and Yechiam (2009) asked respondents to choose between a safe option with a small sure loss and a risky option with a small probability (0.15) of a big loss for 100 trials. The probability of respondents choosing the risky option NEAR-MISS APPRAISALS IN DISASTER EVENTS 70 following no loss was significantly bigger than 0.5. In contrast, the probability of choosing the risky option dropped significantly after the rare outcome appeared. In particular, Newell, Rakow, Yechiam, and Sambur (2015) asked respondents to choose to live in either a safe region with no chance of disaster or a risky region with a small chance of disaster in a sequence of 400 rounds. Respondents could learn one of the following alternatives: (1) whether their current dwelling was damaged, (2) the number of dwellings damaged in their own village when a catastrophe hit their village; (3) the number of dwellings damaged in all villages when a catastrophe hit any village. Although near-miss events were not specified, circumstances in conditions 2 and 3 involve near-miss experiences. When the respondents’ own houses were not damaged, they were able to see whether other houses were damaged, which indicated that their own houses could have been damaged. As a result, respondents in condition 3 were more likely to choose to live in a risky region than those in conditions 1 and 2. The authors suggested that “the round-by-round feedback providing full information about all regions reinforces the fact that ‘most of the time, nothing “bad” happens in the risky areas’” (Newell et al., 2015). This interpretation is consistent with observed reactions to a resilient near-miss experience as suggested in previous near-miss studies (Dillon & Tinsley, 2008): people tended to take risky actions when a disaster did not happen. Therefore, I anticipated that responses to sequential near-misses will exhibit positive recency, i.e., people will perceive less risk and will more likely choose a safe option or deviate from the status quo when they experience more and more near-misses. Moreover, the near-miss experiences in condition 2 can be considered as spatially closer than those in condition 3 since respondents in condition 2 can only view the outcomes in their own village whereas respondents in condition 3 can view outcomes in all regions. Therefore, the NEAR-MISS APPRAISALS IN DISASTER EVENTS 71 result that respondents in condition 3 chose to live in a risky region more often compared to condition 2 also implied that people might be more likely to choose a risky option when near- misses are psychologically proximal. Therefore, I anticipated that psychological distance could influence near-miss appraisals in a sequence of near-misses, i.e., greater risk perception and avoidance behavioral intention associated with psychological proximity. In this study, I designed a game that involved sequential near-misses. I utilized a scenario of a near-miss natural disaster (similar to the scenario in Study 2 in Tinsley et al., 2012) in which a near-miss is defined as houses other than the decision maker’s own were damaged when a natural disaster struck the region. I anticipated that the protective reactions to near-misses would attenuate through a cumulative sequence of near-misses, i.e., respondents would become more and more resilient to near-misses as they experienced a sequence of such events. As a result, they would be less likely to believe that their houses would get damaged, and would be more willing to leave the current houses. I also manipulated two dimensions if psychological distance (PD). Spatial PD was manipulated as whether the respondent was close to the location where houses were hit by the natural disaster. Hypothetical PD, which illustrates how likely an event is imagined by the respondent, was manipulated by how likely a house could be hit by a natural disaster. In particular, when a large proportion of houses were damaged during the natural disaster, the likelihood that a house is hit by a disaster is higher (i.e., more vulnerable to a near-miss). Therefore, hypothetical PD was represented by how many houses in a town were damaged (a low vs. medium vs. high number). When the respondent’s dwelling was located in that town, spatial PD is low; when the respondent’s dwelling was located in the other town, spatial PD is high. I anticipated that respondents would feel more vulnerable to a near-miss natural disaster NEAR-MISS APPRAISALS IN DISASTER EVENTS 72 (i.e., did not get hit during a disaster) when the spatial PD was lower and the likelihood of getting damaged in a disaster was higher (lower hypothetical PD). As a result, they would be more likely to believe that their dwellings would be hit, and they would be more willing to leave their dwellings. Previous studies have suggested that people update subjective probabilities based on observed data, applying Bayes Theorem (Howard, 1970; Martignon & Krauss, 2003; Peterson & Beach, 1967). In particular, Yi and Bier (1998) proposed a Bayesian model indicating that people should update their probability beliefs of risk events with experienced near-misses. However, little research has tested the Bayesian framework in near-misses empirically. In this study, I tested whether people update their judgment of probability of a hit following near-misses. As the number of hits experienced increases, people should update probability of hit with a higher probability. I anticipated that people who experience vulnerable near-misses would be more likely to count the near-misses as hits and update their probability judgment correspondingly. I conducted three experiments using the same game paradigm. In the first two experiments, I evaluated whether respondents’ judged probability of damage and intentions to leave were influenced by spatial and hypothetical PD and the number of near-misses experienced with two different samples. In the third experiment, I tested whether respondents updated their probability judgment differently after experiencing near-misses in different PD levels. The chapter concludes with a discussion of the results from the three experiments. NEAR-MISS APPRAISALS IN DISASTER EVENTS 73 Experiment I Method Procedure and Measures. Subjects were asked to imagine that they lived in a region that is subject to natural disasters. The region has two towns, A and B. There are 100 houses in each town and respondents were asked to imagine their current house is in town A, as shown in Figure 5.1. Subjects were given the base rate of a house in the region being hit when a disaster strikes, such that on average each of the 200 houses in the region was damaged once out of the past four disasters. Each round represents a fixed period of time. In each round, respondents were informed whether a disaster happened, whether their present dwellings were damaged, and how many other dwellings were damaged if their dwellings were not. After receiving the information, they judged the probability that their current houses would be damaged in the next round (0% to 100%), indicated the likelihood that they would leave the current house for the next round (1 = extremely likely to 5 = extremely unlikely), and chose a place (current house, a house in town A or a house in town B) to live for the next round. They were asked to assume there was no cost of moving (i.e., time, effort, money, etc.), and that a house could be renovated immediately after being damaged. I did not provide any further information about the disaster or houses, such as the direction of the disaster, the location of the houses or their level of seismic resistance, etc. Following 20 rounds, respondents took the NMAS and DOSPERT scales introduced in Chapter 3. They also reported their sex at the end of the survey. NEAR-MISS APPRAISALS IN DISASTER EVENTS 74 Figure 5.1. A picture of the scenario. Design. The information that respondents received in each of the 20 rounds contained three types of outcomes: hit, miss, and near-miss. The hit outcome was worded as “A disaster happened. Your house in Town A/B is damaged.” The miss outcome was worded as “A disaster did not happen.” The near-miss outcome was worded as “A disaster happened. Your house in Town A/B is not damaged.” However, X (hypothetical psychological distance: low vs. medium vs. high) houses in Town A/B (spatial psychological distance: same town as the current house or the other town) are damaged. A random number from 1 to 20, 40 to 60, or 80 to 99 was selected as the number of houses damaged for low, medium, and high hypothetical psychological distance, respectively. To avoid the confounding effect of the order of different types of events and test responses to sequential near-misses, each respondent was given a fixed order of outcomes: 2 hits, 2 misses, 6 near-misses, 2 hits, 6 near-misses, and 2 misses. The event sequence is summarized in Table 5.1. NEAR-MISS APPRAISALS IN DISASTER EVENTS 75 Table 5.1. Design of the 20 outcomes in Experiment I R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 H H M M NM1 NM2 NM3 NM4 NM5 NM6 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 H H NM1 NM2 NM3 NM4 NM5 NM6 M M Each respondent was exposed to all 6 types of near-misses in a 2 (spatial PD) by 3 (hypothetical PD) within-subjects design. The order of the six near-misses was counterbalanced with a balanced Latin Square, shown in Table 5.2. For the first sequence of 6 near-misses (rounds 5 to 10), respondents were randomly assigned to one of the six orders. For the second sequence of 6 near-misses (rounds 13 to 18), respondents were randomly assigned to one of the other five orders. As a result, each respondent received all 6 types of near-misses twice. In the first part of the survey, respondents read the scenario and worked on rounds 1-10, in the second part of the survey, respondents continue working on rounds 11-20 and completed the two scales. Respondents were redirected to the second part of the survey once they finished the first part. Table 5.2. Within-subjects Design of the six near misses in Experiment I NM1 NM2 NM3 NM4 NM5 NM6 Condition1 1 2 3 4 5 6 Condition2 2 3 4 5 6 1 Condition3 3 4 5 6 1 2 Condition4 4 5 6 1 2 3 Condition5 5 6 1 2 3 4 Condition6 6 1 2 3 4 5 Respondents. A total of 100 respondents were recruited from USC Psychology Subject Pool. Each student earned course credit for participating in the survey. Among the 100 respondents, 90 finished the survey. Respondents were expected to spend at least two minutes on the first half of the survey and three minutes on the second half, including the two scales at the NEAR-MISS APPRAISALS IN DISASTER EVENTS 76 end. As a result, one respondent was removed for finishing the first part of the survey in less than two minutes and an additional three respondents were removed for taking less than three minutes for the second part of the survey. Eighty-six respondents were included in the data analysis. Seventy-one respondents (82.6%) were female. The median completion time was 8 minutes (IQR = 6 – 11). Table 5.3 shows the descriptive statistics of respondents’ NMAS and DOSPERT scores and their correlations with demographic variable. Table 5.3. Means, standard deviations, and correlations with sex for respondents’ NMAS and DOSPERT scores in Experiment I Mean S.D. NMAS Sex 1. NMAS 3.00 0.36 0.06 2. DOSPERT 3.13 0.81 -0.31* -0.07 Note: * 𝑝 < .05 Results I first examined responses to the 12 near-misses with a two-way repeated measures ANOVA. The measure of leaving intention was reverse-coded so a higher score indicates more likely to leave the current house. For the first 6 near-misses (rounds 5 to 10), a near-miss in the low spatial PD condition (i.e., houses in the same town were damaged) was associated with a significantly higher probability judgment that the current house would be damaged in the next round (𝐹(1,67) = 19.04, 𝑝 < 0.001, 𝜂 ! = 0.22) compared to a near-miss in the high spatial PD condition. A near-miss in the low-probable condition (i.e., a small number of houses were damaged, high hypothetical PD) is associated with a significantly lower probability judgment of damage in the next round compared to a near-miss in the medium or high-probable condition (𝐹(1,67) = 6.87, 𝑝 = 0.01, 𝜂 ! = 0.09). There was no significant difference between medium and high-probable conditions (𝐹(1,67) < 1, 𝑝 = 0.83, 𝜂 ! = 0.001). In addition, a near-miss in the low NEAR-MISS APPRAISALS IN DISASTER EVENTS 77 spatial PD condition was associated with a significantly greater intention to leave in the next round (𝐹(1,86) = 43.29, 𝑝 < 0.001, 𝜂 ! = 0.34) compared to a near-miss in the high spatial PD condition. Subjects were more willing to leave the current house for the next round if more houses were damaged in the current round (𝐹(1,86) = 9.35, 𝑝 = 0.003, 𝜂 ! = 0.10). For the second 6 near-misses (rounds 13 to 18), a near-miss in the low spatial PD condition was associated with a significantly higher probability judgment that the current house would be damaged in the next round (𝐹(1,68) = 22.20, 𝑝 < 0.001, 𝜂 ! = 0.25) compared to a near- miss in the high spatial PD condition. A near-miss in the high-probable condition is associated with significantly higher probability judgment of damage in the next round compared to a near- miss in the medium or high-probable condition (𝐹(1,68) = 6.19, 𝑝 = 0.02, 𝜂 ! = 0.08). There was no significant difference between medium and high-probable conditions (𝐹(1,68) < 1, 𝑝 = 0.92, 𝜂 ! < 0.001). In addition, a near-miss in the low spatial PD condition was associated with a significantly greater leaving intention in the next round (𝐹(1,85) = 34.46, 𝑝 < 0.001, 𝜂 ! = 0.29) compared to a near-miss in the high spatial PD condition. Hypothetical PD did not significantly predict an intention to leave for the second near-miss sequence (𝐹(1,85) < 1, 𝑝 = 0.45, 𝜂 ! = 0.01). Figure 5.2 shows the means of probability of damage in the next round and leaving intention across different psychological distance levels for the first and second sequence of near-misses. (a) 20# 25# 30# 35# 40# 45# 50# low# med# high# Probability*(0-100)* Hypothe4cal** Mean*Probability*Es4ma4on* low# high# Spa4al* 1.5$ 1.7$ 1.9$ 2.1$ 2.3$ 2.5$ 2.7$ 2.9$ 3.1$ 3.3$ 3.5$ low$ med$ high$ Leaving(Inten+on((1/5)( Hypothe+cal(( Mean(Leaving(Inten+on( low$ high$ Spa+al( NEAR-MISS APPRAISALS IN DISASTER EVENTS 78 (b) Figure 5.2. Mean of probability judgment and leaving intention by spatial and hypothetical psychological distance for the first (a) and second (b) near-miss sequence. I further conducted a growth curve analysis to examine how responses to near-misses changed over time when respondents experienced near-misses continuously and whether this change differed across different levels of spatial and hypothetical psychological distance. The analysis was conducted in R using the ‘nlme’ package (Pinheiro, et al., 2017). Results indicated that responses changed significantly over time such that as respondents experienced more near- misses, probability judgments that their current dwellings would be damaged in the next round decreased (b = -0.63, 𝑝 < 0.001) and they were less willing to leave the current house in the next round (b = 0.04, 𝑝 < 0.001). This change was affected by psychological distance. Respondents predicted a higher probability that the current house would be damaged in the next round (b = - 7.82, 𝑝 < 0.001) and were more willing to leave the current house when houses in the same town were damaged (b = 0.62, 𝑝 < 0.001). Probability judgment of damage in the next round (b = 1.55, 𝑝 = 0.003) and leaving intention (b = -0.08, 𝑝 = 0.01) were also higher when a large number of houses were damaged. As expected, the scales of individual differences did not predict growth curves intercepts. Figure 5.3 shows the means of probability judgments of damage in the next round and leaving intention over time across different levels of psychological distance. 20# 25# 30# 35# 40# 45# 50# low# med# high# Probability*(0-100)* Hypothe4cal** Mean*Probability*Es4ma4on* low# high# Spa4al* 1.5$ 1.7$ 1.9$ 2.1$ 2.3$ 2.5$ 2.7$ 2.9$ 3.1$ 3.3$ 3.5$ low$ med$ high$ Leaving(Inten+on((1/5)( Hypothe+cal(( Mean(Leaving(Inten+on( low$ high$ Spa+al( NEAR-MISS APPRAISALS IN DISASTER EVENTS 79 (a) (b) Figure 5.3. Mean of (a) probability judgment of damage in the next round and (b) leaving intention by spatial and hypothetical psychological distance over the 12 rounds of near-misses in Experiment I. In Experiment II, I replicated Experiment I with a sample from Amazon Mechanical Turk (AMK). AMK sample is different from Psychology undergraduate students (Buhrmester, Kwang, & Gosling, 2011) due to greater diversity of the sample, including age, sex, education, etc. Moreover, AMK sample is motivated financially by participating in the study, whereas students were not. The goal of Experiment II is to test whether the findings found in Experiment I holds for a different group of respondents. 30 40 8 12 16 time prob Spatial PD low high 30 40 50 8 12 16 time prob Hypothetical PD low probable high probable 2.8 3.2 3.6 4.0 4.4 8 12 16 time leavelik Spatial PD low high 3.2 3.6 4.0 8 12 16 time leavelik Hypothetical PD low probable high probable NEAR-MISS APPRAISALS IN DISASTER EVENTS 80 Experiment II Method Design. Respondents first reported demographic information. They then participated in the same 20-round game introduced in Experiment I. Following the 20 rounds, respondents completed the NMAS, DOSPERT, and PRT scales introduced in Chapter 3. Respondents. A total of 202 respondents were recruited from Amazon Mechanical Turk. Each respondent was paid $0.75 for participating in the survey. Twelve respondents were removed for finishing the first half of the survey in less than two minutes. Since respondents took one more scale and reported more demographic information than those in Experiment I, they were expected to take longer in the second half of the survey. Therefore, an additional 31 respondents were removed for taking less than four minutes for the second half of the survey. A total of 159 respondents were included in the data analysis. The median completion time was 10 minutes (IQR = 9 – 14). Table 5.4 summarizes demographic information of the sample. Table 5.5 shows the descriptive statistics of respondents’ NMAS, DOSPERT, and PRT scores and their correlations with demographic variables. Table 5.4. Demographic Information of the Sample (N = 159) Variables Values Frequencies and Percentages Sex Male 78 (49.1%) Female 80 (50.3%) Income < $20,000 34 (21.4%) $20,000-$40,000 50 (31.5%) $40,000-$60,000 42 (26.4%) $60,000-$80,000 15 (9.4%) > $80,000 18 (11.3%) Age 18-26 42 (26.4%) 27-33 43 (27.1%) NEAR-MISS APPRAISALS IN DISASTER EVENTS 81 34-43 37 (23.2%) > 44 37 (23.3%) Education High school and below 41 (25.8%) 2-year college 33 (20.8%) 4-year college 70 (44.0%) Master’s degree and above 15 (9.4%) Note: Frequencies do not always sum to 159 due to non-responses Table 5.5. Means, standard deviations, and correlations with demographic variables for respondents’ NMAS, DOSPERT, and PRT scores in Experiment II Mean S.D. NMAS DOSPERT Sex Age Education Income NMAS 2.89 0.46 0.04 -0.04 0.01 0.06 DOSPERT 2.67 1.03 -0.42* -0.02 -0.04 0.08 0.06 PRT 3.04 0.52 0.05 -0.15 -0.02 0.19** -0.02 0.07 Notes: ** 𝑝 < .05; * 𝑝 < .1. Results I first examined responses to the 12 near-misses with two-way repeated measures ANOVA for the Amazon Turk sample. For the first 6 near-misses (rounds 5 to 10), a near-miss in the low spatial PD condition (i.e., houses in the same town were damaged) was associated with a significantly higher probability judgment that the current house would be damaged in the next round (𝐹(1,151) = 4.57, 𝑝 = 0.03, 𝜂 ! = 0.03) compared to a near-miss in the high spatial PD condition. A near-miss in the low-probable condition (i.e., high psychological PD, a smaller number of houses were damaged) is associated with a significantly lower probability judgment of damage in the next round compared to a near-miss in the medium or high-probable condition (𝐹(1,151) = 7.88, 𝑝 = 0.01, 𝜂 ! = 0.05). There was no significant difference between medium and high-probable conditions (𝐹(1, 157) = 1.83, 𝑝 = 0.18, 𝜂 ! = 0.01). In addition, spatial PD did not predict intention to leave in the next round (𝐹 < 1, 𝑝 = 0.45, 𝜂 ! < 0.01). Subjects were more willing to leave the current house for the next round if more houses were damaged in the current round (𝐹(1,157) = 5.79, 𝑝 = 0.02, 𝜂 ! = 0.04). NEAR-MISS APPRAISALS IN DISASTER EVENTS 82 For the second 6 near-misses (rounds 13 to 18), however, results indicated that respondents’ judgment of the probability that the current house would be damaged in the next round was not influenced by spatial PD (𝐹 < 1, 𝑝 = 0.43, 𝜂 ! < 0.01) or hypothetical PD (𝐹(1,150) = 1.29, 𝑝 = 0.26, 𝜂 ! < 0.01). Subjects’ intention to leave the current house for the next round was also not influenced by spatial PD (𝐹 < 1, 𝑝 = 0.88, 𝜂 ! < 0.001) or hypothetical PD (𝐹(1,155) = 2.62, 𝑝 = 0.11, 𝜂 ! = 0.02). Figure 5.4 shows the means of probability of damage in the next round and leaving intention across different psychological distance levels for the first sequence of near-misses. Figure 5.4. Mean of probability judgment of damage in the next round and leaving intention by spatial and hypothetical psychological distance for the first 10-trial near-miss sequence. As before, I also ran a growth curve analysis to examine how responses to near-misses changed over time when respondents experienced multiple near-misses and whether this change varied across different levels of spatial and hypothetical PD. Results indicated that a near-miss in a less probable condition (i.e., a smaller number of houses other than their own were damaged) is associated with a significantly lower probability judgment of damage in the next round (b = 2.11, 𝑝 = 0.01) and less willingness to leave (b = 0.10, 𝑝 = 0.05) compared to a near-miss in a more probable condition. Spatial PD did not predict probability of damage in the next round (b = -0.55, 𝑝 = 0.44), and leaving intention (b = -0.03, 𝑝 = 0.43). Moreover, time influenced intention to leave the current house (b = -0.01, 𝑝 = 0.09). Respondents were less willing to leave the current 20# 22# 24# 26# 28# 30# 32# 34# 36# low# med# high# Probability*(0-100)* Hypothe4cal** Mean*Probability*Es4ma4on* low# high# Spa4al* 1.5$ 1.7$ 1.9$ 2.1$ 2.3$ 2.5$ 2.7$ 2.9$ low$ med$ high$ Leaving(Inten+on((1/5)( Hypothe+cal(( Mean(Leaving(Inten+on( low$ high$ Spa+al( NEAR-MISS APPRAISALS IN DISASTER EVENTS 83 house as the number of near-misses experienced increased. Probability judgments of damage in the next round did not change as respondents experienced more near-misses (b = -0.09, 𝑝 = .62). As expected, the scales of individual differences were not predictive of the growth curves intercepts. Figure 5.5 depicts the means of leaving intention over time by low vs. high hypothetical PD. Figure 5.5. Mean of leaving intention by hypothetical psychological distance over the 12 rounds of near-misses in Experiment II. Findings from Experiments I and II suggested that respondents updated beliefs of a disaster as they experienced multiple near-misses. In Experiment III, I further tested whether the updates of subjective probability after near-misses follow a Bayesian model using the same game paradigm. Assuming respondents’ probability belief follows a Beta-binomial model, I anticipated that respondents who received vulnerable near-misses (i.e., psychologically close) sequentially would be more likely to count the events as hits. Distributions of subjective probabilities under different PD conditions were compared. 25 30 35 8 12 16 time prob Hypothetical PD low probable high probable 2.0 2.2 2.4 2.6 2.8 8 12 16 time leavelik Hypothetical PD low probable high probable NEAR-MISS APPRAISALS IN DISASTER EVENTS 84 Experiment III Method As in Experiments I and II, respondents were presented with a sequence of 20 events and evaluated probability of a future hit and their behavioral intention following each event. The only difference from the previous experiments is regarding the sequence of the 20 outcomes. Design. Each respondent was presented with a fixed order of 20 outcomes: 2 hits, 6 near- misses, 4 misses, 2 hits, and 6 near-misses. The first near-miss sequence (rounds 3 to 8) follows a 2 (spatial PD: low vs. high) by 3 (hypothetical PD: low vs. medium vs. high) between-subjects design. Subjects were randomly assigned to one of the six PD levels. Specifically, a respondent would receive near-misses in the same condition for rounds 3, 4, 7, and 8 and receive near- misses with no information about other houses (i.e., “A disaster happened. Your house is not damaged.”). The second near-miss sequence follows a within-subjects design as before: each respondent received all 6 psychological distance levels; order is counterbalanced by a balanced Latin square. Table 5.6 displays the outcome of each round. Table 5.6. Design of the 20 outcomes in Experiment III R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 H H NM NM NM (no) NM (no) NM NM M M R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 M M H H NM1 NM2 NM3 NM4 NM5 NM6 Procedure. Respondents first participated in the 20-round game. They then completed the NMAS, DOSPERT, and PRT scales introduced in Chapter 3. They also reported their sex at the end of the survey. NEAR-MISS APPRAISALS IN DISASTER EVENTS 85 Respondents. A total of 296 respondents were recruited from USC Psychology Subject Pool. Among the 296 respondents, 261 completed the survey. Eight respondents were removed for finishing the first half of the survey in less than two minutes; an additional sixteen respondents were removed for taking less than four minutes for the second half of the survey. A total of 237 respondents were included in the data analysis. One hundred and sixty (67.5%) respondents were female. The median completion time was 11 minutes (IQR = 8 – 14). Table 5.7 shows the descriptive statistics of respondents’ NMAS, DOSPERT, and PRT scores and their correlations with demographic variable. Table 5.7. Means, standard deviations, and correlations with sex for respondents’ NMAS, DOSPERT, and PRT scores in Experiment III Mean S.D. NMAS DOSPERT Sex NMAS 2.97 0.39 0.13* DOSPERT 3.10 0.87 -0.31* -0.06 PRT 3.15 0.44 -0.05 0.12 -0.11* Note: * 𝑝 < .05 Results I investigated the process of updating subjective probability in a Bayesian approach with a Beta-binomial model. Since there are two outcomes in each event, i.e., a house is either hit by a disaster or not, the probability of being hit at round k over a sequence of n rounds follows a Binomial distribution ! ! 𝑝 ! (1−𝑝) !!! . Respondents can gain information of the process and update their perceived probability of being hit (𝑝) over time. Therefore, I assume perception of the probability of being hit (𝑝) follows a Beta distribution, i.e., 𝑝 ~ 𝐵𝑒𝑡𝑎(𝑝|𝑎,𝑏), where 𝑎 is the number of hits and 𝑎+𝑏 is the number of trials. Since on average each of the 200 houses in the region was damaged once out of the past four disasters, the prior 𝑝 is 0.25 where 𝑎 is 1 and 𝑎+𝑏 is 4. Applying Bayes Theorem, the posterior 𝑝 after 𝑛 rounds also follows the Beta NEAR-MISS APPRAISALS IN DISASTER EVENTS 86 distribution, i.e., the conjugate posterior 𝑝 ~ 𝐵𝑒𝑡𝑎(𝑝|𝑟+𝑎,𝑛−𝑟+𝑏), where 𝑟 is the number of hits in the 𝑛 rounds. When the outcome of a round is a near-miss, I hypothesized that the respondents who experience a proximal near-miss are more likely to interpret the outcome as a hit, whereas respondents who experience a distant near-miss are more likely to interpret the outcome as a miss. Thus, I anticipated that respondents who continuously received near-misses that are psychologically close would perceive r to be higher than those who continuously received near-misses that are psychologically distant. I can then compare the estimated parameters in the posterior Beta distribution 𝐵𝑒𝑡𝑎(𝑝|𝑟+𝑎,𝑛−𝑟+𝑏). Specifically, the fraction of (𝑟+𝑎)/ (𝑛−𝑟+𝑏) is expected to be higher for the group in the low PD condition. Also, since 𝑛, 𝑎, 𝑏 are the same across different conditions at a given round, 𝑟+𝑎 is expected to be higher, and 𝑛−𝑟+𝑏 is expected to be lower for the group in the low PD condition. Perceived posterior probability of being hit is represented by respondents’ estimated probability that their current house would be damaged in the next round. Because the order of near-misses at each level is mixed from round 15 to round 20 due to the within-subjects design, I evaluated probability judgments only in the first 14 rounds. Assuming that the heterogeneity of beliefs in a group captures the random process of Beta distribution, I fit the probability beliefs in each round for the low and high PD condition separately with Beta distributions. I conducted Kolmogorov-Smirnov tests to assess whether the two groups of low and high spatial PD were drawn from the same distribution for each round. Table 5.8 shows the Kolmogorov-Smirnov statistics and p-values from round 1 to round 14. Results indicated that I can reject the null that the two datasets were drawn from the same distribution for rounds 3, 4, 7, 8, which were the rounds that respondents received the information of a near-miss, as well as for round 9 and round 13. NEAR-MISS APPRAISALS IN DISASTER EVENTS 87 I then fit the two datasets to Beta distributions separately for rounds 3, 4, 7, 8, 9, and 13 and estimated the parameters with mme (method of moments estimation) using package ‘fitdistrplus’ in R. The estimated parameters of the posterior Beta distributions at each spatial PD level were also shown in Table 5.8. Results indicated that 𝑟+𝑎 was higher, 𝑛−𝑟+𝑏 was lower, and (𝑟+𝑎)/ (𝑛−𝑟+𝑏) was higher in the low spatial PD condition compared to the high spatial PD condition for rounds 3, 4, 7, and 8, but not for rounds 9 and 13. 10 Table 5.8. Kolmogorov-Smirnov test results and estimated parameters of the posterior Beta distributions Kolmogorov-Smirnov test Posterior Beta D 𝑝-value 𝑟+𝑎 𝑛− 𝑟+𝑏 𝑟+𝑎 𝑛− 𝑟+𝑏 Round 1 0.12 0.36 Round 2 0.11 0.50 Round 3 0.20 0.02* Low 1.19 1.75 0.68 High 1.02 2.08 0.49 Round 4 0.22 0.006* Low 1.13 1.57 0.72 High 0.95 2.07 0.46 Round 5 0.17 0.06 Round 6 0.09 0.75 Round 7 0.25 0.002* Low 1.17 1.77 0.66 High 0.84 2.05 0.41 Round 8 0.28 0.0002* Low 1.07 1.48 0.72 High 0.70 1.85 0.38 Round 9 0.18 0.04* Low 0.86 2.16 0.40 High 0.58 1.95 0.30 Round 10 0.08 0.87 Round 11 0.10 0.65 Round 12 0.09 0.72 Round 13 0.20 0.02* Low 0.87 1.62 0.54 High 1.08 2.90 0.37 Round 14 0.17 0.08 Note: * 𝑝 < .05 10 I acknowledge that a Beta distribution might not provide the best fit to the data. In fact, the Kolmogorov-Smirnov goodness-of-fit test suggested that the Beta distribution did not fit the data well. NEAR-MISS APPRAISALS IN DISASTER EVENTS 88 Discussion The current study is one of the few studies to study responses over a sequence of near- miss events in an experiment that manipulated the vulnerability of the near-misses. My results confirmed the hypothesis that psychological distance could influence risk perception and behavioral intention when multiple near-miss events were experienced sequentially. In a simulated natural disaster context, spatial distance was measured by whether the damaged houses were near or far away from the respondent’s own house; hypothetical distance was measured by how many houses other than respondent’s present house suffered from a disaster in the current time period. Respondents expected that their present dwellings would more likely be damaged and expressed more willingness to move to a different dwelling in the next time period as spatial distance was low and hypothetical distance was low (disaster more probable). Added emphasis on the vulnerability associated with near-misses could sensitize the public to the importance of adopting protective measures. In addition, results showed that responses following near-misses decreased as respondents experienced near-misses continuously, following the psychological default of an expectation of positive recency. Respondents were more likely to expect the non-disastrous outcome to continue, as they experienced a sequence of near-misses. Positive recency (i.e., hot hand fallcy) can be explained by the law of small numbers (Tversky & Kahneman, 1971); people believe that an observed small sample is representative of a large population. Observing a sequence of the same outcome leads to the rejection of independent Bernoulli trials, and the more frequent outcome is expected to continue (Gilovich, Vallone, & Tversky, 1985). This would give rise to assumptions that observations do not follow a “random” Bernoulli process. In fact, the current research suggests that if assuming probability of a disaster follows a Bernoulli NEAR-MISS APPRAISALS IN DISASTER EVENTS 89 distribution, subjective probability of a disaster (p in the Bernoulli distribution) could be influenced by near-miss category. It is also possible that subjective probability is changing dynamically. For instance, people could have knowledge and beliefs about how a sequence works before observing the sequence (Oskarsson, van Boven, McClelland, & Hastie, 2009). For example, a mental model that global warming will intensify over time could make people believe that the probability of droughts will increase over time. Future studies should explore further the systematic change of probability beliefs. The current study is also one of the first to empirically test the extent to which the updates of probability beliefs after near-misses follow a Bayesian model. In a behavioral experiment, Dillon & Tinsley (2008, Study 3) told respondents that a 40% probability of hit was estimated based on 30 experimental trials. They then told respondents there were either three near-misses or three misses. They suggested that if respondents were updating probability belief in a Bayesian way, respondents in the near-miss condition should have a probability judgment of 37%, which is close to 40%, and therefore should not make different decision from those in the miss condition. Since they found a significant difference between the two groups, they concluded that people did not use the number of near-miss events to update probability assessments. The current study is different from the previous study in several ways. First, respondents “experienced” multiple near-misses one by one instead of being told that three near-misses happened before. Second, respondents provided subjective probabilities rather than solely expressing behavioral intention. Third, I compared responses following a near-miss in different psychological distance levels instead of a near-miss vs. a miss. The current study suggested that people updated their perception of being hit by a disaster following near-misses. They are more NEAR-MISS APPRAISALS IN DISASTER EVENTS 90 likely to count near-misses that are psychologically close as hits rather than do near-misses that are psychologically far away. The effects of psychological distance and cumulative sequence of a near-miss studied with hypothetical scenarios in the current paper should be tested in a more naturalistic setting in future studies. In reality, we should consider the effects of various covariates, for example, respondents’ past experience with similar disasters, on near-miss appraisals. We also should consider respondents’ decisions following near-misses in other measures, for example, intention to purchase flood insurance, which might be less aggressive and more realistic than leaving a current dwelling. NEAR-MISS APPRAISALS IN DISASTER EVENTS 91 Chapter 6 : Conclusions This dissertation presented four empirical studies that extend the literature on risk perception in near-miss events as factors that could influence appraisals of near-misses. I determined that individuals’ interpretations of a near-miss event as vulnerable or resilient can be affected by the causal attribution of the event (Chapter 2), an individual’s near-miss appraisal trait (Chapter 3), the psychological distance between a decision maker and the near-miss (Chapter 4), and the cumulative sequences of multiple near-misses (Chapter 5). In particular, the first study (Chapter 2) is one of very few to investigate public responses at different phases of an unfolding near-miss nuclear accident. Throughout the scenario- simulated three-phase incident, negative affect, risk perception, and avoidance behavioral intention decreased as the incident was revealed to be a near-miss and resolved. The study addresses whether appraisal of a near-miss nuclear accident as vulnerable vs. resilient is influenced by the causal attributions for either the initiation of the incident sequence or the halting of the incident progression. Uncertainty regarding the future consequences of a near-miss disaster was found to be another important predictor of public response to the near-miss event. I believe the findings could apply to other incidents involving a chain of related events and causes. It is important to assess public responses to disaster events as they progress over time. Attention to dynamic aspects of a disaster event is particularly important to assess the role of causal attributions for why the incident happened, and why it was averted. The second study (Chapter 3) further predicts that near-miss appraisal is a personality trait such that each individual has her own tendency to interpret a near-miss message. The study developed a Near-miss Appraisal Scale (NMAS) to measure the degree that a person interprets near-misses as vulnerable and responds protectively. The study also describes a framework of the NEAR-MISS APPRAISALS IN DISASTER EVENTS 92 cognitive chain in which reactions to near-misses are determined by the near-miss event itself as well as people’s own appraisal processes. I believe the NMAS could be used in organizations to assess their decision makers’ near-miss appraisal tendency. When communicating potential risks, organizations should also consider individual differences in interpreting the near-miss message. The third and fourth studies (Chapters 4 and 5) further contribute to the literature on near- miss events by focusing on a subjective factor, the psychological distance between a decision maker and the near-miss event. Both studies found that a psychologically proximal near-miss could make people feel vulnerable to a near-miss, while a psychologically distant near-miss could make people feel resilient to the event. The finding was tested with different types of psychological distance (spatial, social, and hypothetical) with a single near-miss event as well as sequential near-miss events. Chapter 5 focuses on the cumulative effect of sequential near-miss events. Consistent with literature on perceptions of sequential extreme events, interpretations of sequential near-miss events exhibit positive recency. People become more and more resilient as near-misses were accumulated. I believe when assessing public response to disaster events it is important to pay attention to individual characteristics that could form different dimensions of psychological distance as well as individuals’ past near-miss experience. Findings in this dissertation suggest that near-miss appraisals are affected by characteristics of both the near-miss events and individuals who are exposed to the events. It is important to consider both individual appraisal traits and characteristics of events when communicating disaster risks. Focusing on characteristics that could make near-misses more vulnerable could also incentivize those at risk to undertake loss reduction measures and encourage individuals, firms and communities to invest in loss mitigation prior to a disaster. NEAR-MISS APPRAISALS IN DISASTER EVENTS 93 One limitation of the current research is it evaluates reactions to disaster events that are deemed to be complicated and dramatic in a simulated experiment setting. While future research should extend and validate the findings reported in this dissertation in naturalistic field study settings, the current research allows systematic manipulation of various factors and control over event characteristics. In a more naturalistic setting, studies should consider various characteristics of events and individuals, as well as potential interactions between event and individual characteristics. The current dissertation evaluated the responses to near-misses in different contexts of disasters, such as nuclear accidents, terrorist attacks, and natural disasters. It is worth noting that the predicted effects of some characteristics could be context dependent. For instance, I found that people became more resilient as they experienced more near-misses over time when encountering natural disasters. This could be different in the context of a nuclear disaster. The public might become more vigilant about nuclear power as the number of near-miss nuclear accidents increases. Thus, further research is needed to compare different types of disasters, e.g., nature vs. human-caused (Siegrist & Sütterlin, 2014), controllable vs. uncontrollable. Furthermore, the current dissertation focuses only on near-misses involved in physical risks. A near-miss could be broadly defined as “something bad could have happened but did not,” such as a near-loss from an investment, or almost missing a bus. The findings in the current research may not apply to domains such as financial decisions or social decisions. Moreover, the near-misses studied in the current dissertation are distinguished from those in gambling behavior in which a near-miss is defined as “near-wins” or “near-losses.” Studies on near-misses in betting behavior have found that gamblers would prolong gambling following near-misses due to expectations of winning (Côté, Caron, Aubert, Desrochers, & Ladouceur, 2003), internal locus- NEAR-MISS APPRAISALS IN DISASTER EVENTS 94 of-control (Clark, Lawrence, Astley-Jones, & Gray, 2009), and physiological arousal (Clark et al., 2011; Dixon et al., 2011). In particular, Sundali et al. (2012) found heterogeneity in whether to bet on near-miss numbers in roulette play. The distinctions and connections between near- misses in disaster events and near-misses in betting behavior could be investigated further in future research. NEAR-MISS APPRAISALS IN DISASTER EVENTS 95 References Anderson, C. R. (1977). Locus of control, coping behaviors, and performance in a stress setting: a longitudinal study. Journal of Applied Psychology, 62(4), 446-451. Bandura, A. (1977). Self-efficacy: toward a unifying theory of behavioral change. Psychological Review, 84(2), 191-215. Bar-Anan, Y., Liberman, N., Trope, Y., & Algom, D. (2007). Automatic processing of psychological distance: evidence from a Stroop task. Journal of Experimental Psychology: General, 136(4), 610-622. Barron, G., & Yechiam, E. (2009). The coexistence of overestimation and underweighting of rare events and the contingent recency effect. Judgment and Decision Making, 4(6), 447- 460. Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289-300. Blais, A.-R., & Weber, E. U. (2006). A domain-specific risk-taking (DOSPERT) scale for adult populations. Judgment and Decision Making, 1(1), 33-47. Bonanno, G. A., Brewin, C. R., Kaniasty, K., & Greca, A. M. (2010). Weighing the costs of disaster: Consequences, risks, and resilience in individuals, families, and communities. Psychological Science in the Public Interest, 11(1), 1-49. Boroditsky, L. (2007). Comparison and the development of knowledge. Cognition, 102(1), 118- 128. NEAR-MISS APPRAISALS IN DISASTER EVENTS 96 Boynton, D. M. (2003). Superstitious responding and frequency matching in the positive bias and gambler’s fallacy effects. Organizational Behavior and Human Decision Processes, 91(2), 119-127. Brown, J., Mulhern, G., & Joseph, S. (2002). Incident-related stressors, locus of control, coping, and psychological distress among firefighters in Northern Ireland. Journal of Traumatic Stress, 15(2), 161-168. Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon's Mechanical Turk: A new source of inexpensive, yet high-quality, data?. Perspectives on Psychological Science, 6(1), 3-5. Carroll, J. S. (2004). Knowledge management in high-hazard industries: Accident precursors as practice. In J. R. Phimister (Ed.), Accident Precursor Analysis and Management: Reducing Technological Risk through Diligence. Washington, DC: National Academies Press. Chalmers, R. P., & others. (2012). mirt: A multidimensional item response theory package for the R environment. Journal of Statistical Software, 48(6), 1-29. Chandran, S., & Menon, G. (2004). When a day means more than a year: Effects of temporal framing on judgments of health risk. Journal of Consumer Research, 31(6), 375-389. Clark, L., Lawrence, A. J., Astley-Jones, F., & Gray, N. (2009). Gambling near-misses enhance motivation to gamble and recruit win-related brain circuitry. Neuron, 61(3), 481-490. Cooper, C. L. (1998). Theories of organizational stress. Oxford University Press. Côté, D., Caron, A., Aubert, J., Desrochers, V., & Ladouceur, R. (2003). Near wins prolong gambling on a video lottery terminal. Journal of Gambling Studies, 19(4), 433-438. Cui, J., Rosoff, H., & John, R. S. (2016). Cumulative response to sequences of terror attacks varying in frequency and trajectory. Risk Analysis, 36(12), 2272-2284. NEAR-MISS APPRAISALS IN DISASTER EVENTS 97 Cui, J., Rosoff, H., & John, R. S. (2017). A polytomous Item Response Theory model for measuring near-miss appraisal as a psychological trait. Decision Analysis, 14(2), 75-86. De Boer, C., & Catsburg, I. (1988). A report: The impact of nuclear accidents on attitudes toward nuclear energy. The Public Opinion Quarterly, 52(2), 254-261. DeJoy, D. M. (1992). An examination of gender differences in traffic accident risk perception. Accident Analysis & Prevention, 24(3), 237-246. Denning, R., & Mubayi, V. (2016). Insights into the societal risk of nuclear power plant accidents. Risk Analysis, 37(1), 160-172. Dillon, R. L., & Tinsley, C. H. (2005). Interpreting near-miss events. Engineering Management Journal, 17(4), 25-29. 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., & Burns, W. J. (2014). Evolving risk perceptions about near-miss terrorist events. Decision Analysis, 11(1), 27-42. Dillon, R. L., Tinsley, C. H., & Cronin, M. (2011). Why near-miss events can decrease an individualś protective response to hurricanes. Risk Analysis, 31(3), 440-449. Dixon, M. J., Harrigan, K. A., Jarick, M., MacLaren, V., Fugelsang, J. A., & Sheepy, E. (2011). Psychophysiological arousal signatures of near-misses in slot machine play. International Gambling Studies, 11(3), 393-407. Embretson, S. E., & Reise, S. P. (2000). Item response theory. Hillsdale, NJ: Lawrence Erlbaum Associates. NEAR-MISS APPRAISALS IN DISASTER EVENTS 98 Enders, C. K., & Bandalos, D. L. (2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling, 8(3), 430-457. Eyal, T., Liberman, N., Trope, Y., & Walther, E. (2004). The pros and cons of temporally near and distant action. Journal of Personality and Social Psychology, 86(6), 781-795. Flynn, J., Slovic, P., & Mertz, C. K. (1994). Gender, race, and perception of environmental health risks. Risk Analysis, 14(6), 1101-1108. Folkman, S. (1984). Personal control and stress and coping processes: A theoretical analysis. Journal of Personality and Social Psychology, 46(4), 839-852. Folkman, S., Lazarus, R. S., Dunkel-Schetter, C., DeLongis, A., & Gruen, R. J. (1986). Dynamics of a stressful encounter: Cognitive appraisal, coping, and encounter outcomes. Journal of Personality and Social Psychology, 50(5), 992-1003. Frijda, N. H. (1986). The emotions: Studies in emotion and social interaction. New York: Cambridge University Press. Fromme, K., Katz, E. C., & Rivet, K. (1997). Outcome expectancies and risk-taking behavior. Cognitive Therapy and Research, 21(4), 421-442. Gianakos, I. (2002). Predictors of coping with work stress: The influences of sex, gender role, social desirability, and locus of control. Sex Roles, 46(5), 149-158. Gilovich, T., Vallone, R., & Tversky, A. (1985). The hot hand in basketball: On the misperception of random sequences. Cognitive Psychology, 17(3), 295-314. Griffin, R. J., Yang, Z., ter Huurne, E., Boerner, F., Ortiz, S., & Dunwoody, S. (2008). After the flood: Anger, attribution, and the seeking of information. Science Communication, 29(3), 285-315. NEAR-MISS APPRAISALS IN DISASTER EVENTS 99 Henderson, M. D., Fujita, K., Trope, Y., & Liberman, N. (2006). Transcending the" here": the effect of spatial distance on social judgment. Journal of Personality and Social Psychology, 91(5), 845-856. Henson, R. K., & Roberts, J. K. (2006). Use of exploratory factor analysis in published research: Common errors and some comment on improved practice. Educational and Psychological Measurement, 66(3), 393-416. Herrero, A. M., Ramírez-Maestre, C., & González, V. (2008). Personality, cognitive appraisal and adjustment in chronic pain patients. The Spanish Journal of Psychology, 11(2), 531- 541. Herzog, S. M., Hansen, J., & Wänke, M. (2007). Temporal distance and ease of retrieval. Journal of Experimental Social Psychology, 43(3), 483-488. Hojat, M., Gonnella, J. S., Erdmann, J. B., & Vogel, W. H. (2003). Medical students ́cognitive appraisal of stressful life events as related to personality, physical well-being, and academic performance: A longitudinal study. Personality and Individual Differences, 35(1), 219-235. Holland, P. W., & Wainer, H. (1993). Differential item functioning. Hillsdale, NJ: Lawrence Erlbaum Associates. Hooper, D., Coughlan, J., & Mullen, M. (2008). Structural equation modelling: Guidelines for determining model fit. Electric Journal of Business Research Methods, 6(1), 53-60. Howard, R. A. (1970). Decision analysis: Perspectives on inference, decision, and experimentation. Proceedings of the IEEE, 58(5), 632-643. NEAR-MISS APPRAISALS IN DISASTER EVENTS 100 Hughey, J. B., & Sundstrom, E. (1988). Perceptions of Three Mile Island and acceptance of a nuclear power plant in a distant community. Journal of Applied Social Psychology, 18(10), 880-890. International Atomic Energy Agency. (2013). INES: The International Nuclear and Radiological Event Scale user’s manual, 2008 edition. Retrieved from http://www- pub.iaea.org/books/iaeabooks/10508/INES-The-International-Nuclear-and-Radiological- Event-Scale-User-s-Manual-2008-Edition International Atomic Energy Agency. (2017, April). Nuclear share of electricity generation in 2016. Retrieved from https://www.iaea.org/PRIS/WorldStatistics/NuclearShareofElectricityGeneration.aspx Jones, C., Hine, D. W., & Marks, A. D. (2016). The future is now: reducing psychological distance to increase public engagement with climate change. Risk Analysis, 37(2), 331- 341. Kahlor, L., Dunwoody, S., & Griffin, R. J. (2002). Attributions in explanations of risk estimates. Public Understanding of Science, 11(3), 243-257. Kahneman, D., & Varey, C. A. (1990). Propensities and counterfactuals: The loser that almost won. Journal of Personality and Social Psychology, 59(6), 1101-1110. Kasperson, R. E., Renn, O., Slovic, P., Brown, H. S., Emel, J., Goble, R., . . . Ratick, S. (1988). The social amplification of risk: A conceptual framework. Risk Analysis, 8(2), 177-187. Katsuya, T. (2001). Public response to the Tokai nuclear accident. Risk Analysis, 21(6), 1039- 1046. NEAR-MISS APPRAISALS IN DISASTER EVENTS 101 Katz, E. C., Fromme, K., & Dámico, E. J. (2000). Effects of outcome expectancies and personality on young adults ́illicit drug use, heavy drinking, and risky sexual behavior. Cognitive Therapy and Research, 24(1), 1-22. Keinan, R., & Bereby-Meyer, Y. (2012). " Leaving it to chance"-Passive risk taking in everyday life. Judgment and Decision Making, 7(6), 705-715. Kelley, H. H., & Michela, J. L. (1980). Attribution theory and research. Annual Review of Psychology, 31(1), 457-501. Kessler, R. C., Galea, S., Jones, R. T., & Parker, H. A. (2006). Mental illness and suicidality after Hurricane Katrina. Bulletin of the World Health Organization, 84(12), 930-939. Kessler, T. A. (1998). The cognitive appraisal of health scale: Development and psychometric evaluation. Research in Nursing & Health, 21(1), 73-82. Khakzad, N., Khan, F., & Paltrinieri, N. (2014). On the application of near accident data to risk analysis of major accidents. Reliability Engineering & System Safety, 126, 116-125. Kunreuther, H., Easterling, D., Desvousges, W., & Slovic, P. (1990). Public attitudes toward siting a high-level nuclear waste repository in Nevada. Risk Analysis, 10(4), 469-484. Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press. Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer. Lejuez, C. W., Read, J. P., Kahler, C. W., Richards, J. B., Ramsey, S. E., Stuart, G. L., . . . Brown, R. A. (2002). Evaluation of a behavioral measure of risk taking: the Balloon Analogue Risk Task (BART). Journal of Experimental Psychology: Applied, 8(2), 75-84. Lerner, J. S., Gonzalez, R. M., Small, D. A., & Fischhoff, B. (2003). Effects of fear and anger on perceived risks of terrorism a national field experiment. Psychological Science, 14(2), 144-150. NEAR-MISS APPRAISALS IN DISASTER EVENTS 102 Levenson, H. (1974). Activism and powerful others: Distinctions within the concept of internal- external control. Journal of Personality Assessment, 38(4), 377-383. Levenson, H. (1981). Differentiating among internality, powerful others, and chance. In H. M. Lefcourt (Ed.), Research with the Locus of Control Construct (Vol. 1, pp. 15-63). New York: Academic Press. Li, S., Li, J.-Z., Chen, Y.-W., Bai, X.-W., Ren, X.-P., Zheng, R., . . . Liu, H. (2010). Can Overconfidence be Debiased by Low-Probability/High-Consequence Events? Risk Analysis, 30(4), 699-707. Lindell, M. K., & Perry, R. W. (1990). Effects of the Chernobyl accident on public perceptions of nuclear plant accident risks. Risk Analysis, 10(3), 393-399. Liviatan, I., Trope, Y., & Liberman, N. (2008). Interpersonal similarity as a social distance dimension: Implications for perception of others’ actions. Journal of Experimental Social Psychology, 44(5), 1256-1269. Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. (2001). Risk as feelings. Psychological Bulletin, 127(2), 267-286. Loewenstein, G., & Lerner, J. S. (2003). The role of affect in decision making. In R. A. Davidson (Ed.), Handbook of Affective Science (pp. 619-642). New York: Oxford University Press. Madsen, P., Dillon, R. L., & Tinsley, C. H. (2016). Airline safety improvement through experience with near-misses: A cautionary tale. Risk Analysis, 36(5), 1054-1066. Martignon, L., & Krauss, S. (2003). Can l’homme eclaire be fast and frugal? Reconciling Bayesianism and bounded rationality. In S. Schneider & J. Shanteau (Ed.), Emerging NEAR-MISS APPRAISALS IN DISASTER EVENTS 103 Perspectives on Judgment and Decision Research (pp. 108-122). Cambridge: Cambridge University Press. Mayer, L. S., & Carroll, S. S. (1987). Testing for lagged, cotemporal, and total dependence in cross-lagged panel analysis. Sociological Methods & Research, 16(2), 187-217. McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal data. Annual Review of Psychology, 60, 577-605. McDaniels, T. L. (1988). Chernobylś effects on the perceived risks of nuclear power: A small sample test. Risk Analysis, 8(3), 457-461. Meyer, R. J. (2012). Failing to learn from experience about catastrophes: The case of hurricane preparedness. Journal of Risk and Uncertainty, 45(1), 25-50. Newell, B. R., Rakow, T., Yechiam, E., & Sambur, M. (2015). Rare disaster information can increase risk-taking. Nature Climate Change, 6, 158-161. Nicholson, N., Soane, E., Fenton-OĆreevy, M., & Willman, P. (2005). Personality and domain- specific risk taking. Journal of Risk Research, 8(2), 157-176. Parkes, K. R. (1984). Locus of control, cognitive appraisal, and coping in stressful episodes. Journal of Personality and Social Psychology, 46(3), 655-668. Peacock, E. J., & Wong, P. T. (1990). The stress appraisal measure (SAM): A multidimensional approach to cognitive appraisal. Stress and Health, 6(3), 227-236. Peters, E. M., Burraston, B., & Mertz, C. K. (2004). An emotion-based model of risk perception and stigma susceptibility: Cognitive appraisals of emotion, affective reactivity, worldviews, and risk perceptions in the generation of technological stigma. Risk Analysis, 24(5), 1349-1367. NEAR-MISS APPRAISALS IN DISASTER EVENTS 104 Peters, E., & Slovic, P. (1996). The role of affect and worldviews as orienting dispositions in the perception and acceptance of nuclear power. Journal of Applied Social Psychology, 26(16), 1427-1453. Peterson, C. R., & Beach, L. R. (1967). Man as an intuitive statistician. Psychological Bulletin, 68(1), 29-46. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., Heisterkamp, S., Van Willigen, B., & Maintainer, R. (2017). Package ‘nlme’: Linear and nonlinear mixed effects models. Retrieved from http://cran.icesi.edu.co/CRAN/web/packages/nlme/nlme.pdf Prati, G., & Zani, B. (2013). The effect of the Fukushima nuclear accident on risk perception, antinuclear behavioral intentions, attitude, trust, environmental beliefs, and values. Environment and Behavior, 45(6), 782-798. Rector, N. A., & Roger, D. (1997). The stress buffering effects of self-esteem. Personality and Individual Differences, 23(5), 799-808. Rickard, L. N. (2014). Perception of risk and the attribution of responsibility for accidents. Risk Analysis, 34(3), 514-528. Rizopoulos, D. (2006). ltm: An R package for latent variable modeling and item response theory analyses. Journal of Statistical Software, 17(5), 1-25. Rosoff, H., Cui, J., & John, R. S. (2013). Heuristics and biases in cyber security dilemmas. Environment Systems and Decisions, 33(4), 517-529. Rosoff, H., John, R. S., & Prager, F. (2012). Flu, risks, and videotape: escalating fear and avoidance. Risk Analysis, 32(4), 729-743. Rosseel, Y. (2012). An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1-36. NEAR-MISS APPRAISALS IN DISASTER EVENTS 105 Rotter, J. B. (1954). Social learning and clinical psychology. Prentice-Hall, Inc. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80(1), 1-28. Rubin, G. J., Amlôt, R., Page, L., & Wessely, S. (2009). Public perceptions, anxiety, and behaviour change in relation to the swine flu outbreak: cross sectional telephone survey. British Medical Journal, 339, b2651. Rundmo, T. (2002). Associations between affect and risk perception. Journal of Risk Research, 5(2), 119-135. Samejima, F. (1969). Estimation of latent ability using a response pattern of graded scores. Psychometrika Monograph Supplement, 34(4), 100-114. Scheibehenne, B., Wilke, A., & Todd, P. M. (2011). Expectations of clumpy resources influence predictions of sequential events. Evolution and Human Behavior, 32(5), 326-333. Siegrist, M., & Sütterlin, B. (2014). Human and nature-caused nazards: The affect heuristic causes biased decisions. Risk Analysis, 34(8), 1482-1494. Slovic, P., Finucane, M. L., Peters, E., & MacGregor, D. G. (2007). The affect heuristic. European Journal of Operational Research, 177(3), 1333-1352. Slovic, P., Layman, M., & Flynn, J. H. (1991). Risk perception, trust, and nuclear waste: Lessons from Yucca Mountain. Environment: Science and Policy for Sustainable Development, 33(3), 6-30. Slovic, P., Peters, E., Finucane, M. L., & MacGregor, D. G. (2005). Affect, risk, and decision making. Health Psychology, 24(45), S35--S40. NEAR-MISS APPRAISALS IN DISASTER EVENTS 106 Smith, B. W., Kay, V. S., Hoyt, T. V., & Bernard, M. L. (2009). Predicting the anticipated emotional and behavioral responses to an avian flu outbreak. American Journal of Infection Control, 37(5), 371-380. Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in emotion. Journal of Personality and Social Psychology, 48(4), 813-838. Steiger, J. H., Shapiro, A., & Browne, M. W. (1985). On the multivariate asymptotic distribution of sequential chi-square statistics. Psychometrika, 50(3), 253-263. Stephan, E., Liberman, N., & Trope, Y. (2010). Politeness and psychological distance: a construal level perspective. Journal of Personality and Social Psychology, 98(2), 268- 280. Sundali, J. A., Safford, A. H., & Croson, R. (2012). The impact of near-miss events on betting behavior: An examination of casino rapid roulette play. Judgment and Decision Making, 7(6), 768-778. Telch, M. J., Brouillard, M., Telch, C. F., Agras, W. S., & Taylor, C. B. (1989). Role of cognitive appraisal in panic-related avoidance. Behaviour Research and Therapy, 27(4), 373-383. Terpstra, T. (2011). Emotions, trust, and perceived risk: Affective and cognitive routes to flood preparedness behavior. Risk Analysis, 31(10), 1658-1675. 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. Trope, Y., & Liberman, N. (2010). Construal-level theory of psychological distance. Psychological Review, 117(2), 440-463. NEAR-MISS APPRAISALS IN DISASTER EVENTS 107 Trope, Y., Liberman, N., & Wakslak, C. (2007). Construal levels and psychological distance: Effects on representation, prediction, evaluation, and behavior. Journal of Consumer Psychology, 17(2), 83-95. Turner, N. E. (1998). The effect of common variance and structure pattern on random data eigenvalues: Implications for the accuracy of parallel analysis. Educational and Psychological Measurement, 58(4), 541-568. Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers. Psychological Bulletin, 76(2), 105. Tyszka, T., Zielonka, P., Dacey, R., & Sawicki, P. (2008). Perception of randomness and predicting uncertain events. Thinking & Reasoning, 14(1), 83-110. Van Der Pligt, J. (1985). Public attitudes to nuclear energy: Salience and anxiety. Journal of Environmental Psychology, 5(1), 87-97. Verplanken, B. (1989). Beliefs, attitudes, and intentions toward nuclear energy before and after Chernobyl in a longitudinal within-subjects design. Environment and Behavior, 21(4), 371-392. Visschers, V. H., & Siegrist, M. (2013). How a nuclear power plant accident influences acceptance of nuclear power: Results of a longitudinal study before and after the Fukushima disaster. Risk Analysis, 33(2), 333-347. Von Winterfeldt, D., & Edwards, W. (1986). Decision analysis and behavioral research. New York: Cambridge University Press. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: the PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063-1070. NEAR-MISS APPRAISALS IN DISASTER EVENTS 108 Watson, L., & Spence, M. T. (2007). Causes and consequences of emotions on consumer behaviour: A review and integrative cognitive appraisal theory. European Journal of Marketing, 41(5/6), 487-511. Weber, E. U., Blais, A.-R., & Betz, N. E. (2002). A domain-specific risk-attitude scale: Measuring risk perceptions and risk behaviors. Journal of Behavioral Decision Making, 15(4), 263-290. Weiner, B. (1985). An attributional theory of achievement motivation and emotion. Psychological Review, 92(4), 548-573. Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Guilford Press. Wilke, A., & Barrett, H. C. (2009). The hot hand phenomenon as a cognitive adaptation to clumped resources. Evolution and Human Behavior, 30(3), 161-169. Wittneben, B. B. (2012). The impact of the Fukushima nuclear accident on European energy policy. Environmental Science & Policy, 15(1), 1-3. World Nuclear Association. (2017, January). Nuclear power in the world today. Retrieved from http://www.world-nuclear.org/information-library/current-and-future-generation/nuclear- power-in-the-world-today.aspx Yi, W., & Bier, V. M. (1998). An application of copulas to accident precursor analysis. Management Science, 44(12), S257-S270.

Abstract (if available)

Abstract Near-miss experiences have been identified as a contributing factor in responses to risk of disaster events. Previous research suggested that different near-miss events could lead individuals to interpret the risk as either “vulnerable” or “resilient”

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Choice biases in making decisions for oneself vs. others

PDF

The cost of missing objectives in multiattribute decision modeling

PDF

Using classification and regression trees (CART) and random forests to address missing data

PDF

Individual differences, science communication, and critical thinking for emergent risks

PDF

Cumulative risk as a moderator of multisystemic therapy effects for juvenile offenders

PDF

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

PDF

A robust model for disaster risk and decision analysis

PDF

On the latent change score model in small samples

PDF

Trade-offs among attributes of authentication

PDF

Evaluating aleatory uncertainty assessment

PDF

Effect of mindfulness training on attention, emotion-regulation and risk-taking in adolescence

PDF

Expressing values and group identity through behavior and language

PDF

Modeling human bounded rationality in opportunistic security games

PDF

Regularized structural equation modeling

PDF

Making terrorism risk assessments more useful for decision-making

PDF

A longitudinal study of sexual minority stress and behavioral health in sexual minority adolescents

PDF

Identity, perceived discrimination, and attenuated positive psychotic symptoms among college students

PDF

A risk analysis methodology to address human and organizational factors in offshore drilling safety: with an emphasis on negative pressure test

PDF

“What difficulty means for me”: predictors and consequences of difficulty mindsets

Asset Metadata

Creator Cui, Jinshu (author)

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

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Publication Date 12/11/2017

Defense Date 09/08/2017

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

Tag Decision making,disaster,near-miss,OAI-PMH Harvest,risk perception

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor John, Richard S. (committee chair), Dehghani, Morteza (committee member), Monterosso, John (committee member), von Winterfeldt, Detlof (committee member)

Creator Email jinshucu@gmail.com,jinshucu@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-460579

Unique identifier UC11266414

Identifier etd-CuiJinshu-5948.pdf (filename),usctheses-c40-460579 (legacy record id)

Legacy Identifier etd-CuiJinshu-5948.pdf

Dmrecord 460579

Document Type Dissertation

Rights Cui, Jinshu

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