UNDERSTANDING TERRORISM RISK PERCEPTION AND IMPROVING RISK COMMUNICATION: FINAL REPORT

Transcription

1 UNDERSTANDING TERRORISM RISK PERCEPTION AND IMPROVING RISK COMMUNICATION: FINAL REPORT Compiled and Edited by: William J. Burns Ph.D. Decision Research, CSUSM October 25, 2010 The following report describes a project that evolved out of a workshop hosted by CREATE in June Leading risk perception and communication researchers were invited to share their insights about where the field was and should be heading. They were invited to submit proposal ideas, which were then assembled into a white paper. Detlof von Winterfeldt and Bill Burns drafted a proposal that was funded in April Many participants, excited to move forward initiated their research in summer 2006 with no budget and effectively donating their time. This project would eventually involve the direct collaboration of investigators from three DHS University Centers (CREATE, START, NCFPD), Pacific Northwest National Lab (PNNL) and six university or research institutes (USC, ASU, UKY, UCI, RTI and Decision Research). Three subsequent conferences were held that collectively involved over 80 participants from leading universities, governmental agencies, research institutes and policy consulting firms. These conferences along with the work of the project s principle investigators have produced over 30 papers and 55 presentations. A special issue in Risk Analysis currently has 12 of these papers under review. Five nationwide surveys were conducted. Two of these longitudinal surveys tracked public response to real time crises. We acknowledge and have been grateful for the support of CREATE and the Pacific Northwest National Lab (PNNL) through this project. This project was supported by the United States Department of Homeland Security through the National Center for Risk and Economic Analysis of Terrorism Events (CREATE) under grant number 2007-ST Parts of this project were also supported by the National Science Foundation under grant number SES However, any opinions, findings, and conclusions or recommendations in this document are those of the authors and do not necessarily reflect views of the United States Department of Homeland Security or the National Science Foundation.

2 SUMMARY INSIGHTS UNDERSTANDING PUBLIC RESPONSE AMIDST CRISES A number of experiments and surveys have been conducted over the duration of this project that seek to understand how the public might respond to different types of disasters including natural disasters, terrorist attacks and even the financial crisis. Experiments have manipulated types and levels of disasters to understand which characteristics are associated with the highest perceived risks, strongest emotional response and most determined intentions to avoid the particular risk. These experiments have included disasters such as a major earthquake and biological, radiological and MANPADS attacks. On occasion, events have allowed surveys in the field to capture how people respond in real time (e.g., financial crisis, Christmas Day and Times Square terrorist attempts, Haiti earthquake and BP oil spill). These longitudinal surveys have provided insight into how perceptions, emotions and behaviors change over time in response to adverse events. Consistent across this work is that terrorism looms large in terms of perceived risk, emotions like fear and anger and avoidance behavior relative to natural disasters. Respondents offered explanations related to uncertainty and a lack of perceived control in such circumstances. They remark that it is difficult to get closure on such events because terrorist groups continue to plan and have the ability to wait for opportunity. There is an increased sense of vulnerability. Regarding technological accidents and terrorists attacks, biological and radiological events appear more freighting than do explosions though both are of great concern. The reasons offered refer to the unseen nature of biological and radiological agents and their potential long-term effects. While it is true that experts can measure by sampling or by instrument levels of these contaminants, still it requires a great deal of trust on the part public. Likewise, it is difficult to know whether certain levels of exposure lead to health problems in the future. Conversely, if a person was not injured in the explosion the risk is over (apart from a future mishap). Negative emotions like fear, anger, worry and anxiety appear positively correlated with perceived risk and avoidance behavior. There is evidence, in highly controlled experiments where these emotions have been induced, that anger is negatively correlated with perceived risk. However, throughout this project anger has been positively related to perceived risk, especially in real time events in the field. The predominant emotions expressed with respect to terrorism were first anger then fear than sadness. For natural disasters like Haiti it was almost exclusively sadness. Perceived risk and negative emotions clearly decay over time and they decline most quickly directly following the peak of an event. Perceived risk and fear surrounding air travel after the Christmas Day terrorist attempt declined markedly within a month. Sadness dropped dramatically within a month of the earthquake in Haiti. Following the mass sell-off of stocks in late September 2008 perceived risk, 2

3 negative emotions and behaviors related to savings, stocks and retirement declined markedly in the first six weeks of the crisis. This latter point is remarkable considering news of the economy continued to get worse until March of This has caused researchers in this project to speculate that there may be mechanisms that operate during a crisis that nudge people back to baseline levels of perceived risk and negative emotions. These mechanisms may be sociological (community offers increased aid and reassurances until fear and perceived risk decreases to acceptable levels). They may also be psychological (emotions in particular return to pre-event levels as proposed by Solomon s opponent process theory). Typically these mechanisms help people get back to normal activities. Sometimes they cause people to take unnecessary risks, as when they allow their insurance policies to lapse following a disaster or fail to continue to take medications once their symptoms disappear. Rapid escalation of perceived risk, negative emotion and risk-related behaviors typically follows a dramatic risk event that comes as a complete surprise (e.g., 9/11, earthquake in Haiti and to a lesser extent the two most recent terrorist attempts) but not always. It may be the case that some threats may not come as a complete surprise but rather their risk signal cumulates until it reaches a level of criticality in which some event triggers a large response. Researchers chronicled the events leading up to the massive sell-off of stocks on Wall Street at the height of the financial crisis. They noted that the public and business community had growing but guarded concerns until Lehman Brothers was allowed to collapse. This finding suggests that growing concerns over cyber threats may eventually reach a tipping point in which the public responds in dramatic fashion even though the threat comes as no surprise. Fear and avoidance behavior appear to increase as a crisis escalates-at least up to a point. Respondents were exposed to hypothetical scenarios involving multiple MANPADS attacks. As the number of attacks increased so did the reported fear and intentions to cancel travel plans. These experimental findings suggest that the public may not habituate to the danger depicted in multiple events (e.g., terrorist attacks). Instead, multiple attacks may interfere with the public s ability to recover and resume normal activities. Hence, the decay rate for negative emotions and perceived risk reported in other studies may be markedly different when multiple events are involved. Level of avoidance behavior appears to depend on the type of behavior. When investigating pubic response to a single MANPADS attack researchers reported that respondents were less likely to cancel travel plans to a close friend s wedding or for an important job interview than they were for a vacation. While studying a RDD attack on the financial district of a major city, researchers asked about willingness to interact with the effected district over time. They found respondents were more willing to use professional services (especially for services conducted remotely) and to buy electronic goods located in the once contaminated district then they were to buy food items from there or to take a vacation nearby. Respondents did report their delay in consumption would probably decline over time. Also, incentives (discounts and increased wages) did appear to convince many to associate with the once contaminated area sooner. For some however, no amount of incentives would cause them to do business in the effected district again. 3

4 Individuals vary greatly in their perceptions of risk and negative emotions with respect to most hazards and this variation may be related to gender, numeracy, worldviews and possibly income. Many studies noted such differences and observed correlations with certain demographic characteristics. Most consistent was the finding that women have higher perceptions of risk than men across most threats. At the peak of the financial crisis there was wide variation in people s initial levels of perceived threat. Income and numeracy (ability to understand numbers) was inversely related to perceived risk. It is not clear whether people begin with different baseline perceptions and emotions or they differ in terms of their reactivity to the event. The economic impact of technological accidents or terrorist attacks may be critically effected by public perceptions and risk-related behaviors. Researchers developed a detailed scenario of a RDD attack on the financial district of Los Angeles as well as a scenario of a 6.5 magnitude earthquake located near Los Angeles. They questioned respondents nationwide about their use of products and services as well as their willingness to vacation and willingness to work in Los Angeles following these events. They found that the RDD attack caused much more concern and more reticence to associate in any way with Los Angeles than did the earthquake. The researchers also imported these survey findings along with pertinent economic data into a Computable Generalized Equilibrium (CGE) model to estimate the economic impact of a RDD attack on the financial district. They were particularly interested to discover the degree to which behavioral factors contributed to the overall economic consequences. It turned out that for every dollar attributed to direct losses (e.g., loss of life, property damage, reduction of GDP due to business interruption) about 15 dollars could be attributed to behavioral factors (e.g., reduced consumption, required wage increases, increased business risk premiums). THE ROLE OF RISK AND CRISIS COMMUNICATION Nationwide surveys, simulation experiments, focus groups and case studies were used to examine how best to communicate with the public during a crisis. Types of crises included terrorist attacks (i.e. RDD, food contamination, anthrax, IED attacks), a flu outbreak, a 7.8 earthquake affecting Los Angeles and the financial crisis. Risk and crisis communication messages need to be tested well in advance of their intended use. Typically this testing must be done initially under controlled conditions in a lab to examine both content interpretation and emotional impact. These messages also need to be tailored to specific audiences (and tested on these audiences) and appropriate in their timing. NCFPD has developed these capabilities and implemented them throughout this project (see Sellnow under Specific Projects). For any risk there will be a gap between expert understanding and public understanding of this risk. For a particular risk domain a model of expert understanding should be developed (e.g., developing an influence diagram). Following this the public should be interviewed and then surveyed regarding their understanding of the risk domain. Risk communication messages should be developed to close the gap between expert and public understanding. This approach has been well documented and implemented by researchers at CMU. 4

5 There are certain characteristics of messages, messengers and audience that make risk and crisis communication more compelling. If a message depicts an event that has close proximity to the intended audience it will be more compelling. Likewise, stories personalizing the crisis (e.g., a short feature on someone who died as a result of the outbreak) will be more salient. Messages discussing how the risk can be avoided and how medical treatment can be obtained generate more interest. Details about the origin and nature of the risk are less interesting. Messages emphasizing numerical assessments of risk (e.g., probability estimates) make sense for highly numerate people but are far less effective for those who are not especially numerate. Vulnerable subpopulations may look to information sources other than traditional media-risk messages must make use of appropriate channels. Likewise, cultural worldviews influence perceptions of risk and preference for risk policy. Local media and representatives from governmental agencies appear to be given more credibility. Risk and crisis communication need to consider how initial and evolving levels of uncertainty will be expressed during a crisis. Being forthright about what is known and not known about the risk is preferred. In the initial stages of a crisis, understating the level of risk may be worse than overstating the risk. Decisions about the level of disclosure regarding general threats (e.g., terrorist threats) also need to be considered. Public preference for full disclosure of the risk and potential consequences appears to depend on the risk domain. For example, risks to airline travel may require full disclosure to the public whereas risk to the financial system may require less. MODELING AND METHODOLOGICAL CHALLENGES A number of different modeling approaches were employed to summarize data (multilevel models, latent growth curve models), structure decisions (influence diagrams, multi-attribute utility models- MAU), simulate public response to different adverse events (system dynamics models) and simulate economic impacts (computable general equilibrium-cge models). The purpose of these efforts has been to describe what people are likely to do during a crisis and to prescribe what risk and policy managers would like people to do. Related to this objective has been an effort to understand how perceptions, emotions and risk-related behaviors translate into economic impacts. Public response to large-scale disasters displays a level of complexity that goes beyond our ability to intuit what consequences might result or what risk management policies might be helpful. In this project we employed system dynamics modeling to simulate how rapidly perceived risk and fear might escalate and return to normal levels. We used CGE modeling to investigate the economic impacts of a terrorist attack on the financial district of Los Angeles. Modeling the trajectory of real time events required latent growth curve modeling and data collected over several time periods. Apart from studying actual events, understanding how people might react to different crises requires developing compelling scenarios to engage respondents. We experimented with different modalities to accomplish this aim. Written scenarios work well enough for largely static depictions of a crisis. To capture the dynamic aspects of a crisis we first crafted scenarios that evolved over time. We later added audio and video components to increase the emotional engagement in the scenario. Story line matters, 5

6 crises have a beginning, middle and end. This point is critical because it is challenging indeed to have people imagine how they might feel and behave in a hypothetical crisis and to forecast how they recover afterward. What we do know is that people do not do this task well and that they tend to underestimate their ability to adapt to adversity. Forecasts of public perceptions of risk, emotional reactions and behavioral intentions that stretch much beyond immediate circumstances should be regarded as upper bounds. Validating models of public reaction presents a formidable challenge moving forward. Model predictions were compared to the findings from longitudinal surveys that tracked the financial crisis as well as the Christmas Day and Times Square terrorist attempts. Likewise, model output was compared to ridership data following the London subway bombings. In these comparative cases, system dynamics modeling did reasonably well describing the general pattern of public response over time following these pivotal events (e.g., rapid rise in concern followed by an immediate marked decline and a leveling off of perceived risk, emotional reaction and risk-related behavior). Even so, most such modeling, including the modeling done in this project, is only approximately corroborated. Looking forward, careful attention must be given to data collection that can lead not only to estimation of model parameters but falsification of model predictions. EVENTS AND TASKS COMPLETED I. Initial Risk Perception Workshop. As mentioned above, during June 2006 nationally recognized risk perception and communication researchers were invited to participate in a workshop at CREATE to develop a research agenda for the study of risk perception and its implication for policy regarding terrorism. Prior to attending, participants were sent a questionnaire asking them to articulate what they thought were the most important research questions as well as to recommend best methodological practices to achieve these research objectives. Please see Slovic and Burns below under Specific Projects for the workshop s recommendations. II. Simulated Behavior Study During California s Great Shake-Out Earthquake Drill. Receiving encouragement that our PNNL INC project would be funded researchers from CREATE and Decision Research took advantage of California s planned 7.8 magnitude earthquake drill. It was promoted as the largest earthquake drill in U. S. history. Our data collection occurred in November 2008 during the actual drill on the USC campus. Students from the University of California participated in a simulated exercise lasting about two hours long and were paid for their involvement. Essentially a scenario was created in which students were initially trapped inside a building. They were then guided through a series of evolving events consistent with what might happen during a 7.8 magnitude earthquake. The idea was to have them talk out loud as a group about what they were thinking, feeling and planning to do as a means of surviving a crisis such as this. Particularly important was what information they would pay attention to and how they would seek to obtain it. Initially participants were concerned more about the safety of family and friends and wished to receive assurances about these matters. As the scenario evolved though an urgency developed and more concern was expressed as to their own safety and ability to survive the disaster. Women had higher perceived risk than men. Details of the study have 6

7 been compiled in a paper entitled Scenario Simulation Group Reactions to the Aftermath of the Great Shake-out Magnitude 7.8 Earthquake. III. DHS Brown Bag. In June 2009 Tim Sellnow (University of Kentucky, NCFPD), Kathleen Vidoloff (University of Kentucky, NCFPD), Richard John (University of Southern California, CREATE) and Bill Burns (Decision Research, CREATE) made presentations of their current work on this project during a brown bag presentation at the Department of Homeland Security. IV. Risk Perception Workshop. In August 2009 CREATE, Decision Research and Pacific Northwest National Lab (PNNL) hosted a workshop that was attended by risk researchers from a number of different disciplines. The workshop was entitled Public Response to Threat: Cross-Disciplinary Contributions and Collaboration. Nine researchers from three DHS university centers (CREATE, START and NCFPD), three researchers from PNNL and scientists from leading academic institutions attended. Marilyn Morgan from DHS university programs participated in the workshop as well. Please see conference agenda, participants and presentations at the following link: V. Symposium: Society for Risk Analysis. In December 2009 Richard John (USC, CREATE), Heather Rosoff (USC, CREATE), Tim Sellnow (UKY, NCFPD), Carol Mansfield (RTI, CREATE) and Bill Burns (Decision Research, CREATE) made four presentations at a symposium during the annual Society for Risk Analysis conference in Baltimore. This symposium is focused on their respective work on this project and is being sponsored by the Risk Communication Special Group (RCSG). The symposium is entitled Perceived Risk: Causes, Consequences and Communication. The four presentations were as follows: 1) Comparing the economic consequences of three disasters: Accounting for fear and perceived risk (Bill Burns); 2) How does government risk communication and social norm affect fear, perceptions of risk, and behavioral intentions following terrorist attacks (Richard John, Heather Rosoff); 3) Risk Communication as a Mitigating Factor in Crisis Situations: Audience Perception and Preference (Tim Sellnow) and 4) The Impact of Profession on Risk Perceptions and Attitudes Towards Potential Homeland Security Programs (Carol Mansfield). VI. Risk Perception and Risk Communication Conference. In March 2010 CREATE hosted a conference that was attended by risk researchers from a number of different disciplines and a representative from the community of local emergency managers. The conference was entitled Risk Perception and Risk- Related Behaviors: Anticipated and Responding to Crisis. Twenty seven researchers and practitioners participated. Three DHS university centers (CREATE, START and NCFPD) and one national lab (PNNL) were represented. Members from three different policy or academic research institutes (RAND, Institute for Alternative Futures, Decision Research) attended. The CEO of the Los Angeles Emergency Preparedness Foundation participated. Scholars representing University of Southern California, Texas A&M, UPenn (Wharton), University of Oregon, University of Maryland, Cornell, Monash University, University of Kentucky, Georgetown University, Yale, University of Colorado, University of Pittsburgh, Cal State Los Angeles, Cal State San Marcos and Carnegie Mellon attended. Additionally, two doctoral students and two post-doctoral students from the University of Kentucky and the University of Sothern 7

8 California participated. Errol Southers (CREATE) who was President Obama s first pick to head the TSA gave a dinner talk entitled Emerging Threats. Please see Slovic and Burns below under Specific Projects for conference agenda, participant biographical information. VII. Special Issue: Risk Analysis. Emerging from the above March 2010 conference has been a special issue of Risk Analysis (paper reviews in progress) tentatively entitled Risk Perception and Risk-Related Behaviors: Anticipating and Responding to Crisis. This special issue has already been approved and Bill Burns is the guest special issue editor with Paul Slovic acting in an advisory role. This special issue represents an important opportunity because Risk Analysis is a highly regarded journal in the area of risk and uncertainty. Currently thirteen papers have undergone at least one round of reviews. Please see Slovic and Burns below under Specific Projects for paper titles. VIII. Homeland Security Policy Analysis Conference. In September, 2010 RTI International hosted a conference in Washington D.C. entitled Estimating the Benefits of Homeland Security Policies. Drs. Carol Mansfield (RTI) and Kerry Smith (Arizona State University) organized this two-day event. This conference brought together leading researchers that examined the benefits of Homeland Security policies. In addition to scholarly presentations there were three round table discussions. Forty-eight scientists and policy makers participated which included representatives from the Department of Homeland Security, the Environmental Protection Agency, U. S. Coast Guard, leading universities and research institutes. Please see Mansfield and Smith under Specific Projects below for the conference agenda and participants. IX. Five Nationwide Risk Perception Surveys and Data Sets: A. Nationwide Longitudinal Survey of Public Response to the Financial Crisis: An Examination of How Perceived Risk Changes Amidst a Crisis. In September 2008 researchers at Decision Research began to survey a nationwide panel about their perceptions of risk to their jobs, savings, investments and retirement emanating from the economic crisis. Data collection took place for several days following September 29, 2008, October 8, 2008, November 5, 2008, December 6, 2008, March 21, 2009, June 30, 2009 and October 6, The original intent was to understand how risk perception might change over time in response to an on-going crisis like a terrorist attack (because terrorist attacks are rare the economic crisis was a proxy). Emotions such anger, fear, sadness, worry, anxiety and stress were tracked to note their possible correlation with perceived risk. Both perceived risk and emotional response were related to people s actions (e.g., withdrawing money from savings, reallocating portfolios, postponing spending). The researchers also inquired about how perceptions of the financial crisis compared to perceptions of terrorism, natural disasters and global warming. As it turned out concern about the ramifications of the financial crisis dwarfed any worry about the threat of terrorism, natural disasters or global warming. Risk perception declined quickly and then leveled off. Details are provided later under specific projects (Burns and Slovic). Also the paper Risk Perception and the Economic Crisis: A Longitudinal Study of the Decay of Perceived Risk is available upon request by contacting Bill Burns at bburns@csusm.edu. 8

9 B. Simulated Earthquake and RDD Attack in Los Angeles. In April 2010, two scenarios were developed by researchers at CREATE and Decision Research with help from Los Angeles Emergency Management. They depicted a 6.5 magnitude earthquake with epicenter near Los Angeles and a dirty bomb attack on the financial district of Los Angeles. Over six hundred people nationwide were shown both disasters (the order of scenario presentation was varied). The central questions addressed: perceptions of risk, trust in local first responders and government officials and emotional and behavioral response. Specifically, the researchers were interested to know how these two events would affect people s willingness to buy products, use professional services, vacation or work in Los Angeles. They specifically looked at a distribution of heightened wage premiums or discounted product prices required to get consumers or workers to resume normal activities with respect to Los Angeles. The RDD attack produced much higher perceptions of risk and fear than the earthquake. The dirty bomb scenario also caused people to report that they would delay their purchases much longer than for the earthquake. A Computable General Equilibrium economic model was used to look at the economic impacts of the RDD attack. Details are provided later under specific projects (Burns and Slovic). Also, the paper Regional Economic Damage From Catastrophic Events: Evaluation and Comparison of Resource Loss and Fear Effects Under a Hypothetical RDD Attack Scenario is available upon request by contacting James Giesecke at james.giesecke@buseco.monash.edu.au or Bill Burns at bburns@csusm.edu. C. Nationwide Longitudinal Survey of Public Response to the Christmas Day and Times Square Terrorist Attempts as well as the Haiti Earthquake and BP Oil Spill. Prompted by the Christmas Day terrorist attempt five waves of survey data from a national sample were collected on December 31, 2009, January 21, 2010, February 24, 2010, May 12, 2010 and September 21, This data queried people about perceptions, emotions and behaviors in response to the Christmas Day and Times Square terrorists attempts as well as the Haiti earthquake and BP oil spill. The surveys inquired about people s willingness to still board airlines or visit major U. S. cities after the two terrorist attempts. It also asked people their confidence in the Department of Homeland Security and willingness to following it security measures. Especially important about this data set is that it tracks the same panel of people over time and hence is able to examine how people s perceptions, emotions and behaviors change. Important insights were gained regarding public response to near misses. Like in the response to the financial crisis perceived risk declined quickly and then leveled off. Perceptions of risk and fear were related to intentions to avoid airlines. More detail will be provided later in the report about these findings. Analysis is in its beginning stages so no paper is available as of this writing. Contact Bill Burns regarding progress with this analysis bburns@csusm.edu. D. A Nationwide Survey that Looks at Public Policy Preferences Regarding a Dirty Bomb Attack in Urban Areas. Conducted , the survey provides a description of a dirty bomb attack and the extent of contamination from a dirty bomb using a map of the respondent s city. Respondents were asked about their preferences over three plans: (1) install radiation monitors and cameras in major threat areas in cities, (2) support the allocation of sheltering space and supplies in retail and office facilities in cities for people to shelter in place in the event of an attack, and (3) fund the organization and implementation of practice evacuations to improve the responses to an attack in cities. The survey also included questions 9

10 about situations where the government might keep information about the risk of terrorist attacks from the public. Respondents were then asked their preferences for government disclosure. The survey was administered to a sample from a web panel (panel members are selected randomly originally using random-digit dialing and now through address based sampling and supplied with a computer and internet access). The respondents were drawn from 33 large cities in the United States Please see Mansfield and Smith under Specific Projects for details. Also, contact Carol Mansfield and Kerry Smith for progress on this analysis (Carol Mansfield <carolm@rti.org>; Kerry Smith <kerry.smith@asu.edu>). E. A Nationwide Survey and Experiment Involving a Terrorist Flu Outbreak Scenario. A series of studies involving a terrorist-inspired flu outbreak were carried out in Video simulations were introduced as a way to investigate perceived risk and behavior change as threat situations escalate in the context of biological attacks. Researchers were interested in investigating the independent and interactive effect of proximity and cause of attack over time on emotions, cognitions, risk perception, and intended avoidance behavior. These experiments involved over 600 college students in both Los Angeles and Washington D.C. Please see John, Rosoff and Weiss under Specific Projects for details. Also contact Richard John, Heather Rosoff and David Weiss for progress on this analysis (Richard John <richardj@usc.edu>; Heather Rosoff <rosoff@usc.edu>; David Weiss <dweiss@calstatela.edu>). X. Specific Projects. A. Richard John (USC, CREATE), Heather Rosoff (USC, CREATE) and David Weiss (Cal State LA). Public Response to Disaster Events. 1. Summary. The objective of this research was to formulate a better understanding of how the public perceives the risk associated with disaster events (both terror and non-terror) and the influence this has on behavioral decision making, both in the immediate and in the long-term. Over the course of the PNNL project we conducted the following six studies: 2. The first study evaluated the independent and interactive effect of government response and public reaction to a MANPADS attack upon a local airport, resulting in a 3x3x3 design (Gov by Pub X FlyBen). Subjects were asked to evaluate their perceived fear, worry, risk perception, and intention to fly, with a particular emphasis on varying the benefits of flying. The subjects recruited for the study were college students in California and Spain and sex was considered as a moderator. A total of 360 subjects were recruited, 180 from each study location and 10 within each of the manipulated study cells. The study showed that there was little difference between the responses in the two countries. Fear and flight plans were unaffected by the government announcements or the social norm. Almost all respondents said they would not change a planned flight to attend a close friend s wedding or important job interview, but a substantial number said they would postpone a vacation or would prefer to drive to a different vacation destination. Women expressed more fear than men. 3. The second study was designed as a follow on to the original MANPADS study findings. MANPADS 2 assessed the independent and interactive effect of government action and public behavior to escalating MANPADS attacks, with a particular emphasis on the escalation of attacks over time. The study design is 10

11 a 3x3x3 (Gov X Pub X Time). As with the previous study, subjects were asked to evaluate their perceived fear, worry, risk perception, and intention to fly. The sample was made up of college students in California and Israel, and sex was considered as a moderator. A total of 504 subjects were recruited, 252 from each study location and 28 within each of the manipulated study cells. The study found that fear increased and more trips were canceled as the attacks escalated. Government protective actions and ticket sales had little impact on either fear or planned flying. Women expressed substantially more fear than men, but canceled more trips only after the second attack. 4. In the remaining studies, video simulation was introduced as a way to investigate perceived risk and behavior change as threat situations escalate in the context of biological attacks. In the third study, public reaction to a flu outbreak was investigated. More specifically, in the FLU 1 study researchers studied the independent and interactive effect of proximity and cause of attack over time on emotions, cognitions, risk perception, and intended avoidance behavior. The study design is a 2x2x7 (Proximity X Cause X Time). The study sample consisted of 600 college students in Los Angeles and Washington DC, including 300 from each study location, with equal numbers in each of the 12 manipulated study cells. Sex and personality variables were tested for moderation effects. Based on the findings from the FLU 1 study, FLU 2 was conducted. The study design was the same as FLU 1 with additional dependent variables and fewer tests for moderators. The study sample was recruited through Qualtrics, an online survey host and respondent recruitment resource. The final sample included 600 respondents, including 300 from each study location (Los Angeles and Washington DC), with equal numbers of males and females in each of the 6 manipulated study cells. The FLU 2 study showed that respondent s reactions to the flu epidemic increase as the disaster escalates. The reactions of respondents closer to the epidemic increased more rapidly and with greater intensity than their distant counterparts. For DC respondents, distance moderated the relationship between epidemic causal origin and respondents pattern of temporal responses to the epidemic. As the flu epidemic escalated, unknown causal origins were perceived as threatening to respondents local to the epidemic, where the known accident causal origin concerned respondent s distant from the epidemic most. In addition, a third flu study was conducted (FLU 3) to test FLU 2 in primarily Spanish speaking adults. The video and survey instrument developed for FLU 2 were translated and re-recorded. Spanish speaking subjects were recruited through Qualtrics from two groups, Los Angeles and the Southwest United States. The final sample included 180 respondents, including 90 from each study location, with equal numbers of males and females in each of the 6 manipulated study cells. The data from this study is currently under analysis. 5. Lastly, we constructed a video simulation of an escalating anthrax attack in Los Angeles. The scenario was written to escalate over 7 time periods in the immediate aftermath of an anthrax release. Reactions to the video scenario were obtained during a 2 hour focus group exercise at the Workshop on Risk Perception and Risk-Related Behaviors: Anticipating and Responding to Crisis. Lessons-learned from the workshop exercise are being used to revise the video simulation and conduct focus groups to learn more about the public s reactions and decision making following an anthrax release. 11

12 6. Publications Submitted Papers Rosoff, H., John, R., and Prager, F. (2010). Flu, risks, and videotape: Escalation of fear and avoidance behavior. Manuscript submitted for publication. Weiss, D. J., John, R., and Rosoff, H. (2010). Fear of flying: Emotional and behavioral responses to escalating terrorist threat. Manuscript submitted for publication. Presentations Baumert, T., Weiss, D. J., Buesa, M., Valino, A., John, R. S., Rosoff, H., & Hovsepian, M. Terrorists scare us, but will they interfere with our plans? Paper presented at the Bi-annual Conference on Subjective Probability, Utility, and Decision Making, Rovereto, August Rosoff, H., and John, R. S. (March 2010). The effects of proximity and cause on the dynamics of affect, threat behavior, risk perception and behavioral intention during a two week escalating deadly flu outbreak. Invited paper presented at the Fourth Annual DHS University Network Summit, Washington, D.C. Rosoff, H., and John, R. S. (March 2010). Effects of propinquity and causal origin on the dynamics of affect and behavioral intention during a two week video simulation of an escalating and deadly flu outbreak. Invited paper presented at the Workshop on Risk Perception and Risk-Related Behaviors: Anticipating and Responding to Crisis, Los Angeles, California. Rosoff, H. and John, R. S. (August 2009). Fear and loathing in Los Angeles and Hollywood. Invited paper presented at The Subjective, Probability, Utility and Decision Making (SPUDM) 22 Biannual Meeting, Rovereto, Italy. Rosoff, H., and John, R. S. (August 2009). Risk perception driving factors. Invited paper presented at the Workshop on Public Response Threat: Cross-Disciplinary Contributions and Collaboration, Eugene, Oregon. Rosoff, H., and John, R. S. (January 2009). Using terrorism risk perception to better understand the public s response to disasters. Paper presented at the 47th Annual Bayesian Research Conference, Fullerton, California. Weiss, D. J., John, R. S., Rosoff, H., Shavit, T., and Rosenboim, M. (March 2010). The effects of multiple attacks on fear and flight cancellation in California and Israel. Paper presented at the Spring Workshop on Risk Perception and Risk-Related Behaviors, Los Angeles, California. Weiss, D. J., John, R. S., Rosoff, H., and Hovsepian, M. (November 2008). Emotional and behavioral responses to terrorism threats. Paper presented at the International Meeting of the Brunswik Society, Chicago, IL. 12

13 Weiss, D. J., John, R. S., Rosoff, H., Baumert, T., Shavit, T., Rosenboim, M., Buesa, M., and Valino, A. (July 2010). Emotional and behavioral responses to terrorism threat. Paper presented at the International Congress of Psychology, Melbourne, Australia. Workshop Exercises Table-top Exercise: Anthrax attack on Los Angeles in Risk Perception and Risk-Related Behaviors: Anticipating and Responding to Crisis Conference at CREATE in March B. Vaughan, Extreme Events Communication and Risk Perceptions in Socially Diverse Populations, University of California, Irvine (in collaboration with Dr. Alan Brothers at PNNL for the model building phase of the project) 1. Summary. This project employed several methodologies to explore crucial risk communication challenges presented by dynamic extreme risk events that, despite uncertainties, require effective public involvement and sustained trust in order to mitigate the public health, social and economic consequences of a major event. Original studies sought to analyze the psychosocial dynamics of significant acts of bioterrorism perpetrated in densely populated urban areas, and to identify and understand variables that foster or threaten community resilience, effective actions and trust in those agencies that implement risk/crisis communication strategies as part of response and recovery. Missteps in risk or crisis communication are especially likely when extreme events: are sudden, complex, have a wide impact, or significantly escalate in severity over time; present ongoing uncertainty as to the scope and severity of consequences; raise questions about the effectiveness of preparedness, response and recovery decisions by leaders and involved government agencies; and affect socially vulnerable or ethnically diverse populations. Research activities examined these specific characteristics of bioterrorism incidents. In addition, the project aimed to identify and understand possible underlying causes of the longstanding and well-documented problem of unequal effectiveness of crisis and risk communication across society with an emphasis on how certain cultural orientations, particularly a precautionary perspective, influence risk perceptions and responses to formal risk communications during an unfolding crisis. The multimethod approach utilized and yielded insights from: case studies of past effective and failed extreme event communications (e.g., Hurricane Katrina, Anthrax bioterrorism, Mumbai terrorist attacks of December 2008); review and analysis of the scientific literature; focus groups conducted in ethnically diverse urban communities, including those with a relatively high rate of poverty; and an online survey (with over 220 participants) that quantitatively measured associations among trust, risk perceptions, precautionary perspectives, risk information processing styles and willingness to adopt certain difficult protective behaviors. Moreover, results from these studies provided input for a system dynamics model that represented complex interactions over time among various cultural, psychological, risk communication and public health factors that may influence the absolute level, rate and direction of change of risk perceptions, trust, interpretations of communications and willingness to engage in recommended self-protective behaviors during an emerging event with considerable initial uncertainty. The project was divided into three phases: 1) Qualitative inquiry, 13

14 background research and exploratory phase, 2) Construction and validation of quantitative measurement instruments of key constructs and 3) model-building and testing. The project yielded several new insights and findings related to bioterrorism risk and crisis communication for those ethnically and socioeconomically diverse populations residing in high priority urban areas (i.e., those areas determined to be likely targets of future terrorist acts). Preliminary project findings from focus groups suggested that during early stages of terrorism incidents, under many circumstances individuals can accept and expect that there may be uncertainties in initial or ongoing official characterizations of the risk situation. However, trust is threatened if formal communications ignore or minimize uncertainty, commit particular communication errors or fail to reference important protected cultural values (e.g., fairness, and protection of the most vulnerable groups in society). These communication gaps have consequences for individuals emotions and judgments that may be evoked during a significant extreme event. For example, among some individuals in vulnerable communities (and other communities, as well), false negative communication errors (premature reassurances) are much more damaging to the immediate and long-term credibility of homeland security and other government officials than false positives (warnings that never translate into predicted serious events), and these effects can persist biasing subsequent interpretations of risk information. This may be because false negative errors violate an underlying cultural expectation or norm about the values that should provide the framework for decision-making and crisis communications. Other findings suggested likely ethnic or cultural differences in orientations toward risk and uncertainty, including the prevalence of a precautionary principle and use of a worst-case scenario criterion to evaluate and formulate attitudes toward homeland security risk decisions and crisis management strategies. Data from focus groups and the online survey revealed that a precautionary perspective is a multidimensional construct that is composed of values (e.g., the importance of taking action to protect public health even in the event of highly uncertain information), emotions (e.g., likelihood that errors in crisis communications will trigger anger) and cognitions related to uncertainty (e.g., attributions about and interpretation of admissions of mistakes in previous risk/crisis communications). This cultural orientation among affected populations potentially can play a significant role in determining reaction to changing risk communications during a dynamic and unfolding bioterrorism incident or other extreme event, and was a major component of the system dynamics model for Phase 3 of the project. This original model depicted and explained how precautionary thinking and other psychological variables (e.g., trust, initial risk perceptions, anticipated regret as a contributor to risk decisions during a public health crisis) can interact over time with characteristics of a biological hazard (e.g., capable of human-to-human transmission, infection rate, incubation period, ratio of deaths to number of infected individuals), media coverage (e.g., characterization and framing of risk event, coverage of conflicting opinions) and risk communication features and strategies (e.g., false positive errors, addressing uncertainty, emphasis on expertise of risk managers, frequency of updates). One main finding from the model and other empirical research was that unexpected failures in risk communication strategies during an act of bioterrorism can occur when the timing and content of communications neglect the values and preferences that are salient due to the adoption of precautionary thinking by the media (a frequent occurrence for news coverage of extreme events) or 14

15 subgroups in the broader population. This and other communication shortcomings can influence the adoption of recommended difficult health-protective actions, thus limiting the effectiveness of mitigation strategies. Moreover, the undesirable effects of a mismatch between cultural perspectives on uncertain risk circumstances and implemented crisis communication strategies can persist or even become amplified over time as an incident progresses. The model and empirical studies associated with this project suggest future research directions, possible strategies to improve the outcomes of risk/crisis communication and how to maximize the adaptability of communication approaches during an unfolding extreme event. 2. Working Papers Intended for Publication Vaughan, E., & Brothers, A. A precautionary orientation toward risk and uncertainty: Modeling the impact of certain cultural perspectives on the effectiveness of risk and crisis communication during a dynamic and extreme risk event. Vaughan, E. The multidimensional nature of precautionary thinking and associations with risk perceptions, trust and willingness to adopt difficult public health recommendations during a significant risk crisis. Vaughan, E. Risk perceptions, trust and use of a precautionary principle to evaluate government responses to a bioterrorism threat in ethnically diverse urban populations. C. Timothy Sellnow (University of Kentucky, NCFPD): Risk Communication in Case of Intentional Food Contamination 1. Summary. This project used the Crisis Communication Convergence model (CCC) in both formative and experimental research. The CCC model seeks to explain the process by which stakeholders sort the content of risk messages. Specifically, the model identifies the way in which stakeholders accept and process information from distinct sources to make decisions about self-protection and crisis recovery. The formative research involved collecting data related to message preferences from individuals recovering from the Hurricane Katrina devastation on the Gulf Coast. The initial study revealed a profound preference for local media sources and government agency representatives. National media sources were seen as having ulterior motives that created an emphasis on disaster myths. Elements of this study have been presented at a variety of conferences. The study is currently under review by the refereed journal entitled, Argumentation and Advocacy. Based on the Hurricane Katrina study, we developed a simulation representing a potential terrorist attack on the school lunch program. We created a series of newscasts that were viewed by a number of focus groups. Our objective was to expand the CCC model to better understand how source credibility functioned in decision-making particularly when terrorism was a possibility. We conducted focus groups in Lexington, Kentucky; Washington, D.C.; Little Rock, Arkansas; and Detroit, Michigan. Focus group participants were chosen to reflect under-represented populations such as African-Americans and Arab-Americans. Initial findings of the study were presented at the International Dairy Expo. Additional 15