Sci Eng Ethics (2014) 20:717–733 DOI 10.1007/s11948-013-9502-z ORIGINAL PAPER Social Media in Disaster Risk Reduction and Crisis Management David E. Alexander Received: 24 April 2013 / Accepted: 27 November 2013 / Published online: 4 December 2013  Springer Science+Business Media Dordrecht 2013 Abstract This paper reviews the actual and potential use of social media in emergency, disaster and crisis situations. This is a field that has generated intense interest. It is characterised by a burgeoning but small and very recent literature. In the emergencies field, social media (blogs, messaging, sites such as Facebook, wikis and so on) are used in seven different ways: listening to public debate, monitoring situations, extending emergency response and management, crowd-sourcing and collaborative development, creating social cohesion, furthering causes (including charitable donation) and enhancing research. Appreciation of the positive side of social media is balanced by their potential for negative developments, such as disseminating rumours, undermining authority and promoting terrorist acts. This leads to an examination of the ethics of social media usage in crisis situations. Despite some clearly identifiable risks, for example regarding the violation of privacy, it appears that public consensus on ethics will tend to override unscrupulous attempts to subvert the media. Moreover, social media are a robust means of exposing corruption and malpractice. In synthesis, the widespread adoption and use of social media by members of the public throughout the world heralds a new age in which it is imperative that emergency managers adapt their working practices to the challenge and potential of this development. At the same time, they must heed the ethical warnings and ensure that social media are not abused or misused when crises and emergencies occur. Keywords Social media
Disasters
Emergency management
Ethics
Twitter
Facebook D. E. Alexander (&) Institute for Risk and Disaster Reduction, University College London, Gower Street, London WC1E 6BT, UK e-mail: david.alexander@ucl.ac.uk 123 718 D. E. Alexander Introduction and Definitions At 09:02, local time, on 29th May 2012, a damaging earthquake struck EmiliaRomagna and Lombardy regions of northern Italy. This was the second major seismic event to affect the area in 10 days. It killed 17 people and caused extensive damage to 40 municipalities. Within 50 minutes a clear and relatively comprehensive picture of the earthquake and some of its most important impacts was available. It could be consulted via the Internet from almost anywhere in the world. The information presented was essentially accurate and the speed with which it became available was largely a result of the use of social media to communicate from the sites affected to places where data could be collected and presented to the public. The term ‘social media’ embraces blogs, micro-blogs, social book-marking, social networking, forums, collaborative creation of documents (via wikis1) and the sharing of audio, photographic and video files (Balana 2012). It is characterised by interactive communication, in which message content is exchanged between individuals, audiences, organisations and sectors of the general public. Social media usage is, to some extent, negatively correlated with age and positively with educational attainment. For example, people over the age of 55 tend to prefer conventional sources of news. The degree of adoption of social media varies from country to country but is generally dynamic in most environments and hence any summary statistics are liable to become outdated rapidly. Attempts to relate social media to personality factors have suggested that they are most attractive to people, of both sexes, who are relatively extrovert (Correa et al. 2010), but there is no indication of the extent to which any effort to develop profiles of users might be culturally conditioned. Information on gender differentiation is, at best, fragmentary (Armstrong and McAdams 2009). In the United States, the Internet is the most important source of information for people under the age of 30. For other Americans, it is second only to television (Krimsky 2007). Elsewhere, the use of ‘smart’ phones and social media resources is increasing so rapidly that they are now a force to be reckoned with throughout the world. Social media dispense with ‘‘information gatekeepers’’, which include doctors giving on-line medical advice and journalists relating a news story. These figures are replaced by apomediaries, in which network filtering or group moderation are the only processes by which the spontaneous feed of information is regulated—a matter of apomediation or disintermediation (Eysenbach 2008). This paper offers a review of the use of social media in disasters and major incidents. I consider both how citizens, emergency managers and first responders make use of social media in crisis and how researchers perceive and characterise the phenomenon. I examine seven ways in which social media are put to use for disaster response, recovery and risk reduction. As social media have both beneficial and potentially malign connotations, their advantages and drawbacks are discussed. Next, I consider the ethical implications of social media in disaster, including the 1 The term ‘wiki’ is defined by the OED as ‘‘A type of web page designed so that its content can be edited by anyone who accesses it, using a simplified markup language.’’ It was apparently first used in 1995. 123 Social Media in Disaster Risk Reduction 719 risks and dilemmas of unregulated communication and the degree of inclusiveness of new media. In order to end on a positive note, examples of successes with social media in disaster are briefly discussed. Finally, some conclusions are drawn, but these must necessarily be provisional, as the field is in the early stages of rapid evolution in relation to both technological development and social acceptance. The Research Literature The research literature on social networking and social media in disasters and crises is still quite limited. Moreover, it focuses on the short-term aspects of emergency response and rapid recovery. It is understandable that there are as yet no studies of the longer term, both because social media are a relatively new phenomenon and because the research is also new. Although ‘new media’, such as the Internet, have received attention from academics for a decade or more, very little of the research on social networking predates 2007. However, there is a trend towards a rapid increase in the number of papers that have been published. In this context, the literature on ‘social media’ needs to be differentiated from that on the social aspects of mass media, which is a much wider field that embraces more conventional and long-standing forms of dissemination of information, such as radio and television (Quarantelli 1989). Studies of social media in disasters have been conducted as part of a general tendency to examine the functioning of social interaction by means of the Internet and mobile devices (Krimsky 2007). Both sets of literature concentrate mainly on specific themes, which are: • • • • how social networks function and how they are used how to build and utilise algorithms either to enhance social networking or to monitor it the extent to which people use social networks, how they perceive them and what their communication preferences are the penetration of devices such as ‘smart’ mobile telephones and the extent to which these provide people with access to social media. In addition, students of risk, crisis and disaster have studied: • • • how social media are used in crises the views and opinions of emergency managers and journalists regarding social media and the extent to which the new media are integrated with more traditional means of communication how social media interact with the traditional sources of information. There is a broad distinction between studies of the technical and social aspects of new media. The creation of new platforms and algorithms characterises the former (Cheong and Lee 2010; White and Plotnik 2010), while studies of the kinds of usage and messages sent relate to the latter (Hughes and Palen 2009; Lindsay 2011). The technical side includes by studies of the rate and modality of diffusion of messages (Song and Yan 2012). 123 720 D. E. Alexander While researchers work to develop software for the efficient dissemination of messages via social networks during crisis situations (e.g. Plotnick et al. 2009), Reuter et al. (2012) advocated a more systematic approach to the use of social networking software in crisis situations, starting with classification of uses and potentials. Researchers are equivocal about the balance between the advantages and drawbacks of social media (see below), but they are united in identifying the uses to which the media can be put. Social media promote cross-platform accessibility and a constant flow of information. Situational updates can be complemented by geographical and locational data (Vieweg et al. 2010). Just-in-time information can be provided on how to cope with developing situations. Moreover, social media provide a framework for the work of journalists and for public discussion and debate. Social Media in Disaster and Crisis The following are some of the ways in which social media can be used in disaster risk reduction and crisis response. 1. A listening function. Social media are able to give a voice to people who do not normally have one. They also enable a remarkably democratic form of participation in public debate and facilitate the exchange of information and points of view. During an emergency, through their tendency to coalesce opinions (or stimulate monetary donations), social media are capable of revealing some aspects of the mental and emotional state of a nation. This may seem a rather exaggerated claim, but it should be noted that Quarantelli (1997) argued that the advent of modern information and communications technology involves changes that are as profound as those that occurred after the invention of printing. These changes do, or soon will, affect directly the majority of the population and the rest indirectly. The listening function involves constantly or periodically sampling the varied output of social media. This enables currents of popular opinion and public preference to be gauged. It may also indicate how the public is behaving and reacting to events. Crawford (2009) classified online listening into three categories: background listening, reciprocal listening and delegated listening by corporations. Crawford defined background listening as mere tuning in, a minimal form of engagement. Reciprocal listening involves two-way exchange of messages, with mutual sensitivity to their content and implications; and delegated listening is a form of ‘‘arm’s length engagement’’, in which messages are monitored and responded to, if at all, en masse. Hence, listening is a question of keeping track of opinions, giving advice or collecting information that is of interest to corporations. As Crawford (2009, p. 526) noted, ‘‘there has been a glorification of ‘voice’ as the prime form of participation online.’’ 2. Monitoring a situation. Whereas the listening function involves the passive collection of information, monitoring is conducted in order to improve reactions 123 Social Media in Disaster Risk Reduction 721 to events and better to manage the general public by learning what people are thinking and doing. Current research (Bird et al. 2012) suggests that harmful and inaccurate rumours are not particularly enhanced by the use of social media. One reason for this is that, with mass participation, the false rumours that do begin to circulate are easily corrected by knowledgeable people. Hence, in the aftermath of the Japanese earthquake and tsunami of March 2011, there was little indication that the massive use of social media by the Japanese public led to the successful propagation of rumour and wildly incorrect information (Hjorth and Kim 2011). Moreover, Stirratt (2011) found that, in the Japan disaster, 49 % of Twitter messages were either positive or somewhat positive in their attitude to emergency preparedness and only 7 % were negative. Floods in Queensland, Australia, led to extensive use of social media for public interaction and communication, but not for the mass propagation of false information. Bird et al. (2012, pp. 30–31) noted that: ‘‘While rumours were common at the height of the disaster, respondents reported that the moderators of the Facebook pages were prompt at confirming information and providing official sources when available.’’ Hence, despite the presence of a major crisis—the floods—the use of social media did not lead to a situation of general anarchy that was out of control. 3. Integration of social media into emergency planning and crisis management. In a questionnaire survey (Barr 2011), it was found that 80 % of US general public and 69 % of online users felt that it would be beneficial for national emergency response organisers to monitor social networking sites regularly. However, in most places this has not happened. Agencies are afraid that social networks will produce inaccurate information of dubious provenance (Goolsby 2010). Moreover, the full integration of social networks into disaster management would require many of them to change their working practices, as, in the words of Palen et al. (2007), ‘‘command-and-control models do not easily adapt to the expanding data-generating and data-seeking activities by the public.’’ Nonetheless, there is immense potential to make data dissemination a two-way process, in which information is both received from the public and fed to it (Crowe 2012; Jennex 2012a; Sykes and Travis 2012). The assertion that command and control may be at odds with social media deserves further elaboration. I have argued elsewhere (Alexander 2008) that there is a continuum between command-based and collaborative models of emergency management. The command end of the spectrum tends to be authoritarian, and to divide competencies by level of command into strategic, tactical and operational. The collaborative end of the spectrum tends to divide competencies by theme, such as communication, logistics, and shelter. As there is no clear hierarchical structure in social media, they fit much better into a collaborative model than a command one. Present experience suggests (Yates and Paquette 2011) that issuing orders to the general public is likely to generate an adverse reaction on social media, whereas issuing requests for collaboration may elicit a more positive response, based on involvement rather than alienation. Moreover, the thematic organisation of 123 722 D. E. Alexander collaborative models of emergency management favour collective information sharing on tasks, topics and sectors. Hughes and Palen (2012) observed that the strict bureaucratic nature of emergency management systems, such as the US NIMS, is at variance with the open system and free access character of the social media of which emergency managers are being exhorted to make use. However, the direct, person to person nature of social media is a boon to public information officers, as it helps them avoid the common pitfall of being misquoted by the official media. Moreover, citizens are widely recognised to be the real first responders after disaster (Helsloot and Ruitenberg 2004): they hold the key to the use of social media as an extension of emergency management. Rarely are emergency management organisations ready to utilise such developments. As Westbrook et al. (2012, p. 2) observed, ‘‘The community, volunteer organizations, and news organizations are currently embracing social media, but EM is slow to adopt and implement it on a full scale.’’ There are demonstrable benefits from doing so. For example, Vihalemm et al. (2012) found that social media can help citizens receive, understand and cope emotionally with warning messages. Yet there is an imperative to act: the public can now share information and disseminate critical news to the world and each other without going through government communication methods. This is revolutionizing the way in which people seek help and the way first responders and managers receive and exchange information. The very structure of communication and information sharing dynamics is changing for both for emergency managers and the public. As the Director of the US. Federal Emergency Management Agency, Craig Fugate, stated in a Senate Homeland Security Hearing in 2011, one of the social elements that is changing in the field of emergency management is the way the public can now be viewed ‘‘as a resource and not a liability.’’ For example, social media can be used to deliver warnings to users. In the most sophisticated cases, these may involve local information in the form of maps and data, as well as instructions on what to do during an impending crisis. 4. Crowd-sourcing and collaborative development. In most disasters, the first responders are the public. Moreover, social capital is involved in the form of the mobilisation of skills, leadership, networks, support systems, and so on (Dufty 2012). This involves the concept that social networks and interaction between people increase productivity and lend added value to outcomes. The social networks benefit from the particular skills of their members. One aspect of the formation of social capital through social media is crowd-sourcing. For example, Ushahidi is the name of a crowd-sourced crisis mapping platform (www.ushahidi.com—Gao et al. 2011a). Sahana, and its derivatives Eden, Vesuvius and Mayon, are open source disaster management systems. These initiatives rely on spontaneous contributions to make them work. This endows them with positive feedback, in that the more they are used, the more popular they become and the more they encourage users to contribute to them. Ushahidi and Sahana are examples of the use of social media to create and disseminate methods and good practices, and to form social capital. They are open-source, 123 Social Media in Disaster Risk Reduction 723 free-access platforms that can be used and modified by anyone. In this respect, they are particularly useful for places where disaster management and response are poorly developed and resources are scarce. In crowd-sourcing, it is suggested that 1 % of the crowd will create content, 10 % will validate it and 89 % will use it (Goolsby 2010). However, this is sufficient to maintain a constant flux of information and a high level of consultation of the sites built upon crowd-sourcing. The drawbacks are that crowd-sourcing lacks a common mechanism to facilitate coordination between organisations, it lacks security features, and it does not necessarily provide the information that is most needed or most accurate (Hammon and Hippner 2012). Nevertheless, crisis mapping is particularly suited to crowd-sourcing through the use of social networks, in that reports can be received from many users, and compiled into the resultant maps, which can be widely disseminated. Maps can depict survivors’ temporary settlement camps, resource distribution sources, accessible roads, impacted areas, and so on. In the words of one researcher (Goolsby 2009), crowd-sourcing creates a sort of ‘‘open intranet’’ in relation to the Internet, or in other words a community of users. 5. Creating social cohesion and promoting therapeutic initiatives. Social media can be used to make people feel part of particular initiatives. They can foster a sense of identification with local or on-line communities. Researchers (e.g. Taylor et al. 2012, p. 25) have noted that people caught up in disaster reported feeling more supported and more optimistic about the future when social media were extensively involved. Moreover, social media can be used to enhance voluntarism by increasing the profile and connectedness of voluntary organisations. In this way, they can have a positive impact on the esprit de corps of the members. An American Red Cross survey of social media usage was carried out in 2010 (Blanchard et al. 2010). It indicated that 24 % of the US population and 31 % of the online population would use the media to tell family and friends they are safe. This reflects both the utility of social media and a well-founded lack of confidence in means of communication such as direct telephone calls, which are subject to network saturation. 6. 7. The furtherance of causes. Social media such as Twitter can be used to launch an appeal for donations. With respect to the 2010 Haiti earthquake disaster, Lobb et al. (2012) found that television had a much greater impact in this respect, but nevertheless a Twitter appeal did elicit a considerable response from public donors. Gao et al. (2011b) found that In 48 hours the American Red Cross received $8 million in donations merely from text messages. Lobb et al. (2012) observed a rapid rise in donations straight after the disaster, when news coverage was maintained at a high level, and then a gradual, persistent decline as coverage dwindled and disappeared. Research. The understanding of social reactions to stress, risk and disaster can be enhanced by the use of social media. This represents a challenge to researchers, who are struggling to create what one of them has called a ‘‘digital ethnography’’ (Murthy 2011a). Some authors (e.g. Castillo et al. 2011) have 123 724 D. E. Alexander chronicled the move towards automatic credibility analysis. Others have compared activity on social media sites with the timeline of events in the field (Chung 2011). The Negative Side of Social Media Reported above are seven ways in which social media are useful and through which they show promise for development in the fields of disaster response and resiliency. However, they do have a darker side (Chung 2011; CSS-ETH 2013). Rumour propagation is not to be ruled out, nor is the dissemination of false or misleading information, whether this is done inadvertently or deliberately. Anyone who doubts the power of Internet-based information to disseminate false information should type the words ‘‘earthquake prediction’’ into a search engine. The resulting sites are a mixture of those that purvey dispassionate scientific information and those that are based on highly debatable, perhaps utterly unscientific, premises and methodologies. Yet the sites run by charlatans often have the same visual impact, and thus superficial legitimacy, as the authoritative scientific sites (Flanagin and Metzger 2007). Worse still, potentially social media can be used to orchestrate crime, depending on the strength or weakness of any function that arises, probably quite spontaneously, to ‘‘self-police’’ their output (Goolsby 2010). Castillo et al. (2011, p. 675) observed that ‘‘immediately after the 2010 earthquake in Chile, when information from official sources was scarce, several rumours posted and re-posted on Twitter contributed to increase the sense of chaos and insecurity in the local population.’’ This contradicts the rumour-quelling propensity of social media recorded in Tokyo after the 2011 earthquake (Bird et al. 2012). Castillo et al. (2011) also found that people had more faith in traditional media, and headlines on Twitter were regarded as less credible. Twitter was seen as a tool for political and commercial propaganda. Moreover, these authors found that outbursts of public sentiment correlated with unreliability in Twitter messages. They attributed some of the drawbacks of Twitter to the disconnection between online messages and the reality on the ground, which the message writers had difficulty in assessing properly. One event in mid-2013 illustrates the negative effect that social media can have during a major incident. In the night of 4 May 2013 a freight train derailed at Wetteren near the Belgian city of Ghent, releasing acrylonitrile gas in an explosion and toxic cloud. Concentrations of the gas reached 600 parts per million, which is 6–8 times the lethal dose. One person died and 93 were injured by the fumes. As the area that was affected is densely urbanised, many aspects of this major incident were shared on social media. However, according to the commander of the Ghent Fire Brigade (GFB), Mr Christian van de Voorde [personal communication], much of the information being put about by the general public was wildly inaccurate. Unfortunately, the situation on the ground remained unstable for more than a week and, during this time, GFB was not able to gather accurate information with which to counteract rumours and exaggerations. In this respect, the impact of social media 123 Social Media in Disaster Risk Reduction 725 on the public image of the incident was decidedly negative because it remained distorted, inaccurate and alarmist. When a situation is simpler and more easily interpretable, one may be able to rely on self-policing of social media by users who are concerned to get the facts right, but when the true facts are, on the instant, unobtainable or in dispute, that cannot be the case. A further example is that of the massive surge in social media activity in the aftermath of Hurricane Sandy, which impacted the Atlantic coast of the United States at the end of October 2012. There is no doubt that social media did provide the opportunity for greatly enhanced exchange of information between the authorities and the public and between members of the public, but not without substantial costs (CSS-ETH 2013, p. 4). Photoshop-style image manipulation was widely used by people who shared photographs of the storm. Exaggerated and false news items, for example, about which places in New York City were flooded, were shared and reposted by so many social media users that they were picked up by mainstream media and thus began to assume the status of true stories until they could be discounted by field checking. Murthy (2011b, p. 11) noted that the very people who are most in need of support may be those who have least access to and understanding of the technology that they would need in order to participate in the social media revolution. But apart from the basic digital divide between the ‘haves’ and ‘have nots’, researchers have noted the persistence of demarcations on the basis of race and class among users of the technology and services. Perhaps the greatest challenge in using social media is the sheer volume of information involved. One researcher (Goolsby 2009, p. 3) commented that ‘‘finding useful ‘tweets’ during a major event… is a little like panning for gold in a raging river.’’ According to Cheong and Lee (2011), ‘‘Twitter has been identified as both a potential facilitator and also a powerful deterrent to terrorism.’’ Hence, there is currently considerable ambiguity about whether social media exert a benign or a malign influence on public safety and security. For example, although China is known for its attempts to repress free usage of the Internet, there is an alternative story. Denis-Remis et al. (2013) described how patriotic Chinese used social media to orchestrate disruption of the activities of a French hypermarket company in China because of anger at French Government policies towards Tibet. Once it had started, the researchers saw the reaction as something that was impossible to stop, an indication of the strength of positive feedback in the effects of messaging (DenisRemis et al. 2013, p. 53). This underlines the positive feedback inherent in social media usage in a crisis, a phenomenon that can generate unstoppable, uncontrollable developments with little regard to whether posterity will consider them positive or negative. Finally, when reviewing the potential drawbacks of social media, one should also note the physical weaknesses. Writing about social media usage during a major interruption of electricity supply to the southern Californian city of San Diego, Jennex (2012b) noted that: ‘‘Ultimately the Great Southwest Blackout can be considered a massive, unplanned, backup battery test.’’ Neither the users nor the providers of sites and cellular communication to reach them were ready in any way 123 726 D. E. Alexander for such an exigency. Tellingly, Jennex (2012b, pp. 4–5) concluded that ‘‘while the functionality of social media is useful, the maturity of social media availability is not sufficient to warrant including social media as operational crisis response systems.’’ In other words, the potential is there, but much needs to be done in order to realise it. This will require preparedness on both the technical and the social fronts, from both suppliers and users of services. One interesting question that deserves to be considered is whether the advantages and disadvantages of social media with respect to disaster mitigation and response are significantly different to those of traditional media. If one considers the case of earthquake prediction, there is little indication that television, radio and newspapers have behaved rationally, responsibly and within the bounds of scientific fact (see, for example, Dearing and Kazmierczak 1993). Would social media have closed or increased the gap between reporting and reality? It is too early to tell. However, the idea of a message that ‘‘goes viral’’ and is thus picked up by millions of users, is worrying. For example, the 30-min documentary film Kony 2012, about Joseph Rao Kony, the Ugandan founder of the Lord’s Resistance Army (LRA) was viewed almost 100 million times on YouTube, through which it reached more than half of young adult Americans. Most of the people who saw it know no more about Kony and the LRA than they found out in the film. However, it has been widely criticised as being significantly inaccurate and misleading (Cavanagh 2012). This could easily happen in disasters, and, in so doing, potentially widen the audience beyond those who would have gained information from the traditional media. On the other hand, the worldwide misinterpretation of the trial of seven officials after the Italian L’Aquila earthquake of 2009 shows that the conventional media are just as capable as social media of picking up a misleading story and vigorously propagating it (Alexander in press b). In synthesis, besides problems such as waning battery power, the use of social media brings forth issues of trust and privacy. As Johnson et al. (2011) observed, trust is asymmetric, personalised, dependent on context and potentially short-lived. Moreover, privacy is an increasing concern that is shared by many users of social media. Ethics of Social Media Usage in Disaster Students of social media have been quick to take note of potential ethical dilemmas. The first to investigate this aspect were legal and medical researchers. For example, the associations that represent medical personnel in the United States have counselled against uninhibited use of social networking, as it can lead to conflicts of interest, for example, when a doctor uses social media to befriend a patient. Decamp (2013) noted that problems such as this are a legal minefield, and one that lawyers have increasingly come to monitor in search of opportunities to litigate. Emergency managers normally have to walk a tightrope between actions that may be deemed excessive and any failure to respond adequately that could be considered negligence (Alexander in press a). As McKee (2013, p. 3) observed, ‘‘The changing nature of technology, as well as the relatively recent use of social media for research, means 123 Social Media in Disaster Risk Reduction 727 that ethical considerations will have to be reviewed regularly.’’ The same will be true of protection against unwarranted official intrusions into personal privacy (Nissenbaum 2004). According to Nissenbaum (2004), privacy is subject to a set of norms about what can be divulged and how that can happen, but the norms vary from one social context to another. This is the theory of ‘contextual integrity’. Violation of the norms amounts to a loss of privacy. Grodzinsky and Tavani (2010) noted that norms established or respected by one participant, for example the author of a personal blog, can easily be violated by the republication of material in a different context, for example, a public blog. As noted above, risks are associated with a largely unregulated Internet-based system of public mass communication. In summary, the use of social media for nefarious or malignant purposes could potentially include attempts to persecute people or damage their reputations (Boggs and Edwards 2010), attempts to spread malicious rumour, efforts to foment violent protest, and attempts to organise terrorist activities. Lesser degrees of harm could involve invasion of privacy and unauthorised dissemination of personal information. Moreover, any system of disaster response or risk reduction that depends on social media for access to its services risks excluding those people who lack access to the requisite means. No amount of self-congratulation about the high levels of penetration of ‘smart’ mobile telephony can obscure the fact that there are citizens who for reasons of poverty, disability or choice do not possess the instruments in question and do not know how to use them. Moreover, it is clear that the wealthier, younger, fitter and more aware a person is, the more he or she is likely to be fully aware of the services available via social media and its potential under different circumstances. ‘‘Computer illiteracy’’ is a form of disadvantage in a world that has become dependent on digital communication for many services. It is only partially compensated for by the fact that, by relaying information by word of mouth, other people will be able to help a disadvantaged individual cope. However, total equality cannot be a sine qua non of social media, as there will always be some level of social exclusion to hinder their development. If these problems become serious, critics will begin to argue that the democratic function of social media is a mere illusion. Instead, we see the media as a means of propagating and sustaining direct democracy, and of fostering participatory governance in disaster risk reduction. In this respect, it should be borne in mind that there is a difference between disseminating information and giving people executive power on the basis of what is disseminated. Nevertheless, even the mere propagation of information is empowering, to a certain extent. The risk of concerted abuse of social media begs the question of whether they should be more regulated, and if so, how could that be achieved? Regulation would require an assessment of the potential seriousness of the problem, estimation of whether measures to control social media are feasible, and a decision on whether the measures are necessary and would help abate the problem. Social media and the Internet constitute a truly open system, and one that has no centre. As Brafman and Beckstrom (2006) noted in their book The Starfish and the Spider, when such a system is under attack it tends to mutate into something that is 123 728 D. E. Alexander even less centralised and even harder to control. Thus, attempts in the People’s Republic of China and the Islamic Republic of Iran to dictate what people can access on the Internet are only partially successful: they slow down rather than stop the dissemination of information. In the main, the question of how one polices a leaderless organisation is largely unsolvable. Individuals can be tracked down and arrested, sites can be taken down, services can be blocked or wound up, but others will replace them. The converse view is that social media are a good means of circumventing corruption, mendacity, unwarranted control and excessive monitoring of people’s activities. Moreover, we can view social media as a source of charity, solidarity and hope. While we should beware of treating such a system as a nirvana of democracy, while there is a good consensus among users about what is right and good, and what is acceptable, self-policing will be active. Bad posts and unethical messages will be censured by the users. In this respect, we can return to Charles Fritz’s (1961) conception of disasters as the home of the ‘‘therapeutic community’’—cf. Alan Barton’s ‘‘altruistic community’’ (Barton 1970) and Taylor’s ‘‘utopian feelings’’ (Taylor et al. 1970). In the immediate aftermath of a sudden-impact disaster, there tends to be a greater consensus in the community on what is right, good and ethical to think, say and do. Differences are put aside for the duration of the emergency, as the community faces a large external threat, for example the destruction caused by an earthquake, hurricane or flood. This is the period in which social media come into their own as a source of communication and support. Hence, they can be said to benefit from and reinforce the ‘‘therapeutic community’’, in a form of social symbiosis. One other aspect of the ethics of social media usage deserves to be examined here. I have written elsewhere (Alexander in press b) of the ‘‘death of discretion’’ in modern culture. In fact, wireless communication is part of a broad trend towards the gradual abandonment of personal discretion and increasing tendency to share intimate details. For example, people grieve in public far more than they did in the early days of mass media. Social media facilitate mass participation in this process. Those who go against the grain and pose grossly offensive messages in relation to people’s grief at losses are likely to be pursued vigorously by the authorities, as the level of outcry forces action to be taken. Social media can thus produce a very robust consensus, but not an absolute one. There is an ethical question about the misrepresentation of disaster, whether this is deliberate, inadvertent or a combination of the two (Singer and Endreny 1994). The first ‘‘modern disaster’’ was probably the famine and starvation that occurred in the Biafran war of secession from Nigeria in 1967–1970. This was the first occasion on which film footage of death and suffering in an African famine were shared with the Western public in near real-time. However, the famine was not the result of natural disaster or inability to produce food, as the television viewers believed, but was a tactic of war, the result of blockades and military incompetency (Mayer 1969). Decades later, the Live Aid concerts and ensuing donation bonanzas propagated the misconception that Ethiopians were incapable of responding to catastrophic drought, when in reality they were the victims of forced migration, which was practically sustained by the donation spree (Müller 2013). Theoretically, 123 Social Media in Disaster Risk Reduction 729 social media are able to correct the imbalances of perception, because realistic reports are likely to arrive from the field. However, much depends on who takes note of the reports and how they are mediated by social perception. For research, this is largely uncharted territory. One is prompted to ask whether there are any solutions to the ethical problems that at least potentially, and in some cases demonstrably, beset social media usage in disasters. In more general terms, high-profile cases of electronic media-based ‘‘stalking’’ and persecution, propagation of racism, endangerment of minors, threats of violence and incitements to criminal activity have prompted the authorities of various countries to discuss or adopt measures to take the offenders off line, track them down and, where appropriate, prosecute them. With respect to deliberate misuse of information in disasters, rarely have there been punitive measures with regard to the traditional media, let alone social media. However, the trial of the ‘‘L’Aquila seven’’ (Alexander in press b) is a landmark case in which the misuse of information given to the public was at the heart of the legal case. In that the prosecution successfully demonstrated that the results of this were fatal for some tens of recipients, the way has been opened for legal action on information that is harmful through negligence or design. Some Recent Successes in the Use of Social Media in Disasters Despite the potential for negative uses of social media, there have already been some success stories in their employment for disaster response and risk reduction. Social networks have been used instantly and spontaneously to report health issues, such as cholera in Haiti and dengue in Thailand and Indonesia (Resch et al. 2011). In Japan after the March 2011 earthquake and tsunami, social media facilitated public alerts, helped locate missing people and enabled mapping of different aspects of the emergency (Hjorth and Kim 2011). Flash-flooding in January 2011 in Queensland and Victoria, Australia, led to a six-fold increase in public accession to emergency service Facebook sites, representing a huge increase in interest and support (Bird et al. 2012). Researchers have developed an on-line monitoring tool to detect sharp increases (‘bursts’) in the frequency of key words that appear on Twitter (Cheong and Lee 2010). Furthermore, rumours and misuses of Twitter are being monitored by researchers at the University of Indiana, via a website called Truthy (truthy.indiana.edu). Lastly, the proliferation of crisis camps and their aggregations, crisis commons, has started to promote the more systematic organisation of social media for emergency response, on occasion using wikis (Blanchard et al. 2010). Conclusion The current situation regarding social media in disasters and risk reduction has been summarised very well by Sutton et al. (2008, p. 7): Our data suggest that social media support the influence of the existing publicside information production and distribution. As a consequence of the growing 123 730 D. E. Alexander utility of social media and the ubiquity of peer-to-peer communications, we believe that a change in disaster management models will come about in spite of any failure to formally recognize these widely distributed and often strikingly well-organized information activities. However, we argue that simply letting these inevitable changes take place would nevertheless result in needless delay, conflict and missteps. Instead, we call for efforts by public officials to actively consider how to align with peer-to-peer information exchange and to develop new conceptualizations of the information production and dissemination functions for disaster response. In other words, the incorporation of social media into pre-existing emergency management systems is inevitable, owing to the sheer weight of public usage of such facilities. Moreover, as social networks can be two-way means of communication, they can mix popular and official information. In this optimistic view, the citizenry is viewed as a powerful, self-organising and collectively intelligent force (Gao et al. 2011b). Lastly, it is notable that, after only a very few years of research, there are still many aspects of social media usage in crisis that are not adequately understood. One of these is the influence of gender on perception, attitudes and behaviour regarding usage of the new media (Armstrong and McAdams 2009). If social media are to be optimised as a means of communication during emergencies, such aspects will have to be understood thoroughly. In synthesis, social media offer immense potential for interaction with the public and monitoring of the public’s concerns. They have greatly increased the scope, volume and speed of information exchange. This has not occurred without risks, mostly associated with the propagation of false or inaccurate information, and the potential consequences if this takes place. However, mass participation tends to rectify some of the inadequacies associated with the free and unregulated flow of information. The future will probably see a rationalisation of the use of social media and new methodologies for judging the public mood and the utility of information supplied by the public. This will be a challenge that emergency planners and managers must necessarily face. One final point concerns the role of social media in the long periods dominated either by protracted recovery from major disasters or by mitigation (disaster risk reduction). Patient attempts will need to be made to incorporate social media into these processes and during them the technological, cultural and social realities will inevitably change. There is a strong need for institutions such as civil protection services and emergency warning systems to be adapt to the changing reality of social media, and also to ensure that they have robust plans to tackle any ethical dilemmas that social media usage may produce in the future. References Alexander, D. E. (2008). Emergency command systems and major earthquake disasters. Journal of Seismology and Earthquake Engineering, 10(3), 109–118. 123 Social Media in Disaster Risk Reduction 731 Alexander, D. E. (in press [a]). Celebrity culture, entertainment values … and disaster. In G. Bankoff et al. (eds) Cultures and disasters: Understanding cultural frameworks in disaster risk reduction (forthcoming). Alexander, D. E. (in press [b]). Communicating earthquake risk to the public: The trial of the ‘‘L’Aquila Seven’’. Natural Hazards. Armstrong, C. L., & McAdams, M. J. (2009). Blogs of information: How gender cues and individual motivations influence perceptions of credibility. Journal of Computer-Mediated Communication, 14, 435–456. Balana, C. D. (2012). Social media: Major tool in disaster response. Inquirer Technology, 15 June 2012, 5 pp. Barr, P. (2011). Staying connected. Social media put to work when disaster strikes. Modern Healthcare, 41(36), 33. Barton, A. H. (1970). Communities in disaster: A sociological analysis of collective stress situations. New York: Doubleday. Bird, D., Ling, M., & Haynes, K. (2012). Flooding Facebook: The use of social media during the Queensland and Victorian floods. Australian Journal of Emergency Management, 27(1), 27–33. Blanchard, H., Carvin, A., Whittaker, M. E., Fitzgerald, M., Harman, W., & Humphrey, B. (2010). The case for integrating crisis response with social media. White Paper. Washington, DC: American Red Cross. Boggs, B. C., & Edwards, M. L. (2010). Does what happens on Facebook stay on Facebook? Discovery, admissibility, ethics, and social media. Illinois Bar Journal, 98(7), 1–4. Brafman, O., & Beckstrom, R. A. (2006). The starfish and the spider: The unstoppable power of leaderless organizations. Harmondsworth: Penguin. Castillo, C., Mendoza, M., & Poblete, B. (2011). Information credibility on Twitter. In Proceedings of the 20th International Conference on World Wide Web, 28 March-1 April 2011. Hydrerabad, India, International World Wide Web Conference Committee, pp 675–684. Cavanagh, C. J. (2012). Kony 2012 and the political economy of conflict representation. Nordic Africa Institute, University of Uppsala, Uppsala, 7 pp. http://www.nai.uu.se/news/articles/2012/03/09/ 145947/Kony-2012_LongVersion_ConnorCavanagh.pdf. Accessed August 26, 2013. Cheong, M., & Lee, V. C. S. (2010). Twitmographics: Learning the emergent properties of the Twitter community. In N. Memon & R. Alhajj (Eds.), From sociology to computing in social networks: Theory, foundations and applications (pp. 323–342). Berlin: Springer. Cheong, M., & Lee, V. C. S. (2011). A microblogging-based approach to terrorism informatics: Exploration and chronicling civilian sentiment and response to terrorism events via Twitter. Information Systems Frontiers, 13(1), 45–59. Chung, I. J. (2011). Social amplification of risk in the Internet environment. Risk Analysis, 31(12), 1883–1896. Correa, T., Hinsley, A. W., & Gil de Zúñiga, H. (2010). Who interacts on the Web? The intersection of users’ personality and social media use. Comupters in Human Behaviour, 26(2), 247–253. Crawford, K. (2009). Following you: Disciplines of listening in social media. Continuum: Journal of Media and Cultural Studies, 23(4), 525–535. Crowe, A. (2012). Disasters 2.0: The application of social media systems for modern emergency management. Boca Raton, FL: CRC Press. CSS-ETH. (2013). Using ICT and social media in disasters: Opportunities & risks for government— Background document. 3RD Special Report. Zürich: Centre for Security Studies, ETH. Dearing, J. W., & Kazmierczak, J. (1993). Making iconoclasts credible: The Iben browning earthquake prediction. International Journal of Mass Emergencies and Disasters, 11(3), 391–403. Decamp, M. (2013). Physicians, social media, and conflict of interest. Journal of General Internal Medicine, 28(2), 299–303. Denis-Remis, C., Lebraty, J.-F., & Philippe, H. (2013). The 2008 anti-French demonstrations in China: Learning from a social media crisis. Journal of Contingencies and Crisis Management, 21(1), 45–55. Dufty, N. (2012). Using social media to build community disaster resilience. Australian Journal of Emergency Management, 27(1), 40–45. Eysenbach, G. (2008). Medicine 2.0: Social networking, collaboration, participation, apomediation, and openness. Journal of Medical Internet Research, 10(3), e22, 1–15. Flanagin, A. J., & Metzger, M. J. (2007). The role of site features, user attributes, and information verification behaviours on the perceived credibility of web-based information. New Media and Society, 9(2), 319–342. 123 732 D. E. Alexander Fritz, C. E. (1961). Disaster. In R. K. Merton & R. A. Nisbet (Eds.), Contemporary social problems (pp. 651–694). New York: Harcourt, Brace & Jovanovich. Gao, H., Barbier, G., & Goolsby, R. (2011a). Harnessing the crowdsourcing power of social media for disaster relief. Intelligent Systems, IEEE, 26(3), 10–14. Gao, H., Wang, X., Barbier, G., & Liu, H. (2011b). Promoting coordination for disaster relief: From crowdsourcing to coordination. In J. Salerno, S. J. Yang, D. Nau & S.-K. Chai (Eds.), Social computing, behavioural-cultural modelling, and prediction. Lecture Notes in Computer Science 6589 (pp. 197–204). Heidelberg: Springer. Goolsby, R. (2009). Lifting elephants: Twitter and blogging in global perspective. In J. Salerno, S. J. Yang, D. Nau & S.-K. Chai (Eds.), Social computing, behavioural-cultural modelling, and prediction. Lecture Notes in Computer Science 6589 (pp. 2–7). Heidelberg: Springer. Goolsby, R. (2010). Social media as crisis platform: The future of community maps/crisis maps. Association for Computing Machinery Transactions on Intelligent Systems and Technology, 1(1), Article No. 7, 1–11. Grodzinsky, F. S., & Tavani, H. T. (2010). Applying the ‘‘contextual integrity’’ model of privacy to personal blogs in the blogosphere. International Journal of Internet Research Ethics, 3, 38–47. Hammon, L., & Hippner, H. (2012). Crowdsourcing. Business and Information Systems Engineering, 4(3), 163–166. Helsloot, I., & Ruitenberg, A. (2004). Citizen response to disasters: A survey of literature and some practical implications. Journal of Contingencies and Crisis Management, 12(3), 98–111. Hjorth, L., & Kim, K.-H. Y. (2011). The mourning after: A case study of social media in the 3.11 earthquake disaster in Japan. Television and New Media, 12(6), 552–559. Hughes, A. L., & Palen, L. (2009). Twitter adoption and use in mass convergence and emergency events. International Journal of Emergency Management, 6(3–4), 248–260. Hughes, A. L., & Palen, L. (2012). The evolving role of the public information officer: An examination of social media in emergency management. Journal of Homeland Security and Emergency Management, 9(1), Article 22. Jennex, M. (2012a). Managing crises and disasters with emerging technologies: Advancements. Hershey, PA: IGI Global. Jennex, M. E. (2012b). Social media-viable for crisis response? Experience from the great San Diego/ Southwest blackout. International Journal of Information Systems for Crisis Response and Management, 4(2), 53–67. Johnson, H., Lavesson, N., Zhao, H., & Wu, S. F. (2011). On the concept of trust in online social networks. In L. Salgarelli, G. Bianchi, & N. Blefari-Melazzi (Eds.), Trustworthy internet (pp. 143–157). Berlin: Springer. Krimsky, S. (2007). Risk communication in the internet age: The rise of disorganized skepticism. Environmental Hazards, 7(2), 157–164. Lindsay, B. R. (2011). Social media and disasters: Current uses, future options, and policy considerations. CRS Report for Congress. Washington, DC: Congressional Research Service. Lobb, A., Mock, N., & Hutchinson, P. L. (2012). Traditional and social media coverage and charitable giving following the 2010 earthquake in Haiti. Prehospital and Disaster Medicine, 27(4), 319–324. Mayer, J. (1969). Famine in Biafra. Post-Graduate Medicine, 45, 236–240. McKee, R. (2013). Ethical issues in using social media for health and health care research. Health Policy (early view). Müller, T. R. (2013). ‘The Ethiopian famine’ revisited: Band Aid and the antipolitics of celebrity humanitarian action. Disasters, 37(1), 61–79. Murthy, D. (2011a). New media and natural disasters: Blogs and the 2004 Indian Ocean tsunami. Information, Communication and Society, 1, 1–17. Murthy, D. (2011b). Twitter: Microphone for the masses? Media, Culture and Society, 33(5), 779–789. Nissenbaum, H. (2004). Privacy as contextual integrity. Washington Law Review, 79(1), 119–157. Palen, L., & Liu, S. B. (2007). Citizen communications in crisis: Anticipating a future of ICTsupported public participation. In ACM Conference on Human Factors in Computing Systems CHI 2007 Proceedings: Emergency Action 28 April-3 May 2007, San Jose, California, pp. 728–736. Plotnick, L., White, C., & Plummer, M. (2009). The design of an online social network for emergency management: A one stop shop. In Proceedings of the 15th Americas Conference on Information Systems, San Francisco, California 6–9 August 2009, pp. 1–9. 123 Social Media in Disaster Risk Reduction 733 Quarantelli, E. L. (1989). The social science study of disasters and mass communication. In L. Walters, L. Wilkins, & T. Walters (Eds.), Bad tidings: Communication and catastrophe (pp. 1–19). Hillsdale, NJ: Lawrence Erlbaum Associates. Quarantelli, E. L. (1997). Problematical aspects of the information/communication revolution for disaster planning and research: Ten non-technical issues and questions. Disaster Prevention and Management, 6(2), 94–106. Resch, B., Crowley, D. N., Breslin, J. G., Sohn, G., Burtner, R., Pike, W. A., et al. (2011). Crowdsourcing, citizen sensing and sensor web technologies for public and environmental health surveillance and crisis management: Trends, OGC standards and application examples. International Journal of Health Geographics, 10(1), 1–29. Reuter, C., Marx, A., & Pipek, V. (2012). Crisis Management 2.0: Towards a systematization of social software use in crisis situations. International Journal of Information Systems for Crisis Response and Management, 4(1), 1–16. Singer, E., & Endreny, P. M. (1994). Reporting on risk: How the mass media portray accidents, diseases, disasters and other hazards. Risk: Health, Safety and Environment, 5(3), 261–270. Song, X., & Yan, X. (2012). Influencing factors of emergency information spreading in online social networks: A simulation approach. Journal of Homeland Security and Emergency Management, 9(1), Article 30. Stirratt, A. A. (2011). Social media use in March 2011 Japanese crisis: Impact on emergency preparedness advocacy. Purdue University, West Lafayette, Indiana: Master of Public Health Dissertation. Sutton, J., Palen, L., & Shklovski, I. (2008). Backchannels on the front lines: Emergent uses of social media in the 2007 southern California wildfires. In F. Fiedrich & B. Van de Walle (Eds.), Proceedings of the Fifth International ISCRAM Conference, Washington, DC, May 2008, pp. 1–9. Sykes, T. F., & Travis, E. A. (2012). Social media and disasters: Uses, options, considerations. Hauppauge, NY: Nova Science Publishers. Taylor, M., Wells, G., Howell, G., & Raphael, B. (2012). A Facebook study from ‘Cyclone Yasi Update’: The role of social media as psychological first aid as a support to community resilience building. Australian Journal of Emergency Management, 27(1), 20–26. Taylor, J. B., Zurcher, L. A., & Key, W. H. (1970). Tornado: A community responds to disaster. Seattle, Washington: University of Washington Press. Vieweg, S., Hughes, A. L., Starbird, K., & Palen, L. (2010). Microblogging during two natural hazards events: What Twitter may contribute to situational awareness. In ACM Conference on Human Factors in Computing Systems, CHI 2010 Prodeedings: Crisis Informatics. 10–15 April 2010, Atlanta, Georgia, pp. 1079–1088. Vihalemm, T., Kiisel, M., & Harro-Loit, H. (2012). Citizens’ response patterns to warning messages. Journal of Contingencies and Crisis Management, 20(1), 13–25. Westbrook, R., Karlgaard, T., White, C., & Knapic, J. (2012). A holistic approach to evaluating social media’s successful implementation into emergency management operations: Applied research in an action research study. International Journal of Information Systems for Crisis Response and Management, 4(3), 1–13. White, C., & Plotnik, L. (2010). A framework to identify best practices: Social media and Web 2.0 technologies in the emergency domain. International Journal of Information Systems for Crisis Response and Management, 2(1), 37–48. Yates, D., & Paquette, S. (2011). Emergency knowledge management and social media technologies: A case study of the 2010 Haitian earthquake. International Journal of Information Management, 31(1), 6–13. 123 Psychology of Violence 2017, Vol. 7, No. 2, 316 –327 © 2016 American Psychological Association 2152-0828/17/$12.00 http://dx.doi.org/10.1037/vio0000056 Psychological Outcomes in Reaction to Media Exposure to Disasters and Large-Scale Violence: A Meta-Analysis Tanya L. Hopwood and Nicola S. Schutte This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. University of New England Objective: A quantitative meta-analysis set out to consolidate the effect of experimental studies of media exposure to disasters and large-scale violence on negative psychological outcomes. Method: The meta-analysis included 18 experimental studies with an overall sample size of 1,634 to obtain an overall effect size and information regarding moderators of the effect size. Results: An overall significant and large effect size of Hedges’ g of 1.61 showed that, across studies, media exposure to disasters and large-scale violence was followed by negative psychological outcomes. Outcome type was a significant moderator, with anxiety reactions showing an especially strong effect. Community sensitization was a significant moderator, with studies conducted in a region that had recently been exposed to the type of disaster or violence portrayed in the media showing especially large effect sizes. Conclusion: The results indicate that media exposure to disasters and large-scale violence can cause negative psychological outcomes, at least transiently. Limitations included a lack of statistical power in some moderator analyses and the inability to draw inferences about the duration of effects. There is a need for further research aimed at identifying the possible cumulative effects of media exposure and identification of groups at greatest risk for harmful outcomes. Keywords: media exposure, disaster, violence, meta-analysis antecedent of posttraumatic symptoms (Houston, 2009; Pfefferbaum et al., 2014), and Murray (2008) claimed that the extant research has supported three types of effects of televised violence—increased aggression, desensitization, and fear. Further, Wilson (2008) concluded that children’s extensive use of screen media (including violent material) may affect their socialization, though these effects may be mediated by factors such as age, gender, the extent to which they identify with the characters depicted, and how real they believe the media content to be. Although some studies have found either no effect or beneficial effects of disaster-related media exposure (Linley, Joseph, Cooper, Harris, & Meyer, 2003; Williams & Khan, 2011), and other researchers have asserted that media exposure may only exacerbate preexisting symptoms of trauma (Ahern et al., 2002), the majority of studies suggest that media consumption of disaster and large-scale violence-related material may evoke psychological reactions similar to those experienced by direct victims of trauma (Houston, 2009; Pfefferbaum et al., 2014; Slone & Shoshani, 2010). Longitudinal and survey studies found that people distally located from the September 11 terrorist attacks of 2001 and exposed to the events primarily via mass media experienced significant personal threat and posttraumatic stress reactions (Callahan, Hilsenroth, Yonai, & Waehler, 2005; Dougall, Hayward, & Baum, 2005; Schuster et al., 2001; Silver, Holman, McIntosh, Poulin, & Gil-Rivas, 2002). Some studies found a positive association between hours of September 11 TV coverage consumed and stress reactions (Blanchard et al., 2004; Schlenger et al., 2002), and other researchers identified a link between disaster-focused distress and perceived similarity to the victims as depicted by media (Wayment, 2004). Although substantial research has been conducted in the area of media exposure to disasters and large-scale violence, most of this Advances in technology are affording people unprecedented second-hand exposure to disasters and large-scale violence (Kaplan, 2008; Slone & Shoshani, 2010). Given the vast number of people consuming modern media and the growing propensity of news outlets to employ techniques such as rolling coverage of disasters and large-scale violence as they occur (Jain, 2010; Kaplan, 2008), it is increasingly important for research to explore the potential impact of threat-related content. The purpose of the present study was to consolidate the effect of experimental studies of media exposure to disasters and large-scale violence on negative psychological outcomes. There has been a long-standing academic debate as to whether or not media with violent content can constitute exposure to violence. Some scholars claim that much of the research linking violent media to aggressive behaviors has drawn unfounded inferences of causation from largely correlational research (Grimes & Bergen, 2008). In addition, the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM–5; American Psychiatric Association, 2013) has amended the previous version (4th ed., text. rev.; DSM–IV–TR; American Psychiatric Association, 2000) to explicitly state that media consumption cannot constitute exposure to trauma for a diagnosis of posttraumatic stress disorder (PTSD). Bolstering the other side of the debate, numerous studies have found evidence to suggest that media exposure may act as an This article was published Online First May 5, 2016. Tanya L. Hopwood and Nicola S. Schutte, School of Behavioural, Cognitive, and Social Sciences, University of New England. Correspondence concerning this article should be addressed to Nicola S. Schutte, Psychology, University of New England, Psychology Lane, Armidale, Australia. E-mail: nschutte@une.edu.au 316 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. MEDIA EXPOSURE TO LARGE-SCALE VIOLENCE research has employed self-report survey methods. These studies provide valuable details and are rich in ecological validity, collecting information on how individuals have chosen to consume media and retrospectively recording subjective reactions. Researchers have examined media forms such as TV, newspaper, radio, and the Internet, and have looked at coverage of both terrorist events and natural disasters. Many such studies have found strong links between disaster-related media consumption and negative psychological outcomes, including increased anxiety (Schuster et al., 2001), fear and depression (Lachlan, Spence, & Seeger, 2009), a heightened sense of threat (Maeseele, Verleye, Stevens, & Speckhard, 2008), aggression (Argyrides & Downey, 2004), and posttraumatic stress symptomology (Pfefferbaum, Pfefferbaum, North, & Neas, 2002; Pfefferbaum et al., 2000; Schlenger et al., 2002). A recent descriptive research synthesis by Pfefferbaum et al. (2014), which examined correlational studies of the relationship between disaster-related media consumption and psychological outcomes, found evidence of an association between TV viewing of disaster news and negative outcomes such as posttraumatic stress (PTS) symptoms, stress reactions, depression, and fear. Similarly, a meta-analysis by Houston (2009) found a significant overall effect size (r ⫽ .162) for the relationship between terrorism-related media consumption and PTS. However, as with the individual studies, the correlational nature of the data comprising these meta-analyses does not allow for inferences regarding causality; people who watch traumatic events on the news may consequently experience fear. Alternatively, people experiencing fear may watch traumatic news, perhaps for information-seeking, surveillance purposes, or reassurance. Longitudinal studies that measured psychological dimensions pre- and postdisasters (Cohen et al., 2006; Kennedy, Charlesworth, & Chen, 2004; Otto et al., 2007; van Zelst, de Beurs, & Smit, 2003) have helped support the theory of media exposure to disasters and large-scale violence effecting negative psychological outcomes. These studies, although again high in ecological validity, cannot control for a range of potentially confounding exposure variables, and so are unable to isolate the effects of media. Although many researchers have used experimental methods to explore individual psychological reactions to media exposure to disasters and large-scale threats, no meta-analysis of these studies exists. By conducting a meta-analysis of experimental studies of media exposure to disasters or large-scale violence, we aimed to identify an overall effect size for psychological outcomes and also consolidate information regarding the main types of reactions to such media exposure. A review of the literature indicated that anxiety (or stress) and anger are commonly measured outcomes (Lerner, Gonzalez, Small, & Fischhoff, 2003; Pfefferbaum et al., 2014). As pointed out by Slone and Shoshani (2010), the experience of these emotions is predicted by the theory of protection motivation (Rogers, 1983). This theory asserts that when a person interprets a situation as threatening, anxiety will often result. This anxiety may promote a need to defend the self and others, which may in turn lead to anger. We believe that this theory may help provide a useful scaffold for understanding how media exposure to disasters and large-scale violence may communicate a sense of personal or community threat, which may in turn provoke reactions such as anxiety, anger, and other forms of negative affect. 317 Another theory that may add to the conceptual framework for understanding people’s reactions to this type of media content is the conservation of resources theory (Hobfoll & Lilly, 1993). The conservation of resources theory claims that a person’s ability to cope with challenges depends on his or her perceived inventory of practical, social, and emotional resources. Maguen, Papa, and Litz (2008) posited that large-scale threats (such as terrorism) intensify perceptions of resource loss in areas such as self-esteem, selfefficacy, and internal locus of control. The loss of these safeguarding resources may increase levels of negative affect and diminish adaptive coping (Moos & Holahan, 2003). Brewin, Andrews, and Valentine (2000) showed that many studies have supported a cascading effect of perceived resource loss, with multiple stressors leading to increased vulnerability to further stress. If it can be demonstrated that one-time media exposure to disasters or large-scale violence plays a causal role in negative psychological outcomes— even transiently—this may suggest the presence of maladaptive and more enduring effects in some individuals as the result of cumulative long-term exposure. In the current meta-analysis, we predicted that across studies, there would be a significant effect size for the impact of media exposure to disasters or large-scale violence on negative psychological outcomes. There are a number of variables that we believed might moderate outcomes across studies: intentionality of the event portrayed, media format, whether the study was conducted before or after September 11, community sensitization to trauma in relation to the sample, and differences in participant sample gender and age. The background for selection of these moderator variables follows. Intentionality of Event Research has indicated that a disaster designed with human intent (e.g., a terrorist attack) may be associated with a higher risk of subsequent psychopathology than a disaster of accidental or natural origin (DiMaggio & Galea, 2006). In addition, research results have suggested that different forms of emotions may be more common in the wake of accidental versus intentional trauma (e.g., anger for intentional events, fear for random events; Rosoff, John, & Prager, 2012). We examined whether portrayals of intentionally created disasters would result in stronger negative psychological outcomes. Media Format Communications research has provided evidence for the efficacy of video footage, compared with more traditional forms of media such as newsprint or radio, in creating a more emotionally arousing experience—a sense of realism that has been referred to as presence (Lombard & Ditton, 1997). Graphic footage of disasters may evoke a sense of immediacy and engagement, and perhaps even a potent communication of threat (Callahan et al., 2005; Cho et al., 2003). Meta-analytic studies and research syntheses of correlational research (Houston, 2009; Pfefferbaum et al., 2014) indicated that consumption of disaster or large-scale threat news via TV is significantly associated with numerous negative psychological outcomes, including PTSD, PTS, depression, anxiety, and anger. We examined whether video portrayal of disasters would be associated with larger effect sizes. HOPWOOD AND SCHUTTE 318 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. The Impact of September 11 Much research on media consumption of disasters and largescale violence occurred in the wake of the terrorist attacks in the United States on September 11, 2001 (Neria & Sullivan, 2011). The rolling coverage of these events in the media continued for days, and then sporadically during the weeks and months that followed, giving millions of people across the globe access to details of the disaster (Pfefferbaum et al., 2002). The extensive coverage and large scale of the mass casualties that occurred may have evoked changes in belief systems related to safety and security for many individuals (Linley et al., 2003). We examined whether post-September 11 studies would show stronger effect sizes. Community Sensitization to Trauma Although some theorists contend that previous exposure to trauma may serve as a form of inoculation against adversity (Eysenck, 1983, as cited in Shrira, Palgi, Hamama-Raz, Goodwin, & Ben-Ezra, 2014), many studies have indicated that a history of previous trauma increases susceptibility to adverse psychological outcomes, such as PTSD (Brewin et al., 2000; Chatard et al., 2012; Shrira et al., 2014). Some global locations have had more recent direct experience of disasters (e.g., war, terrorism, disease, and natural disasters) than others. In the current meta-analysis, the global region in which each experiment was conducted was coded as a possible moderator variable. Also, an additional variable called community sensitization was created to identify whether or not the region in question had recently (within the previous 5 years) experienced a disaster of the type portrayed in the stimulus material. We examined whether studies conducted in areas that have recently experienced large-scale disaster or violence would show larger effect sizes. contended that children may be a particularly vulnerable population in terms of adverse PTS reactions (Comer & Kendall, 2007; Dirkzwager, Kerssens, & Yzermans, 2006; Pfefferbaum, 1997; Saylor, Cowart, Lipovsky, Jackson, & Finch, 2003). We examined whether studies of adults with older participants would show stronger effect sizes and whether studies with child participants would show stronger effect sizes. Current Study The current study aimed to consolidate findings of experimental research providing information regarding the causal impact of disaster-related media exposure on psychological reactions through a meta-analysis. Our predictions are as follows. Hypothesis 1: Across studies, there is a significant effect size for the impact of media exposure to disasters or large-scale violence on negative psychological outcomes. Hypothesis 2: Media portrayals involving intentionality result in stronger negative psychological outcomes. Hypothesis 3: Video portrayal results in larger effect sizes. Hypothesis 4: Post-September 11 studies show larger effect sizes. Hypothesis 5: Studies conducted in areas recently experiencing large-scale disaster or violence show larger effect sizes. Hypothesis 6: Studies with a higher percentage of females show larger effect sizes. Hypothesis 7: Studies with older participants or child participants show stronger effect sizes. Method Individual Differences In the correlational research of media consumption of disasters and large-scale violence and psychological outcomes, the most widely replicated moderating variable is gender, with females consistently demonstrating greater susceptibility to negative outcomes than males (Baum, Rahav, & Sharon, 2014). It has been suggested that this gender effect may be partially accounted for by women’s higher levels of fear for others (altruistic fear; Nellis & Savage, 2012), heightened empathy (Nellis, 2009), increased levels of perceived vulnerability (Baum et al., 2014; Nellis, 2009), or greater propensity to acknowledge distress (Lachlan, Spence, & Nelson, 2010). We examined whether studies with a higher percentage of females would show stronger effect sizes. Eysenck’s (1983) inoculation theory (as cited in Shrira et al., 2014) suggests that older people have experienced more challenges across their lifetimes and have thus developed greater resilience—a quality that may help protect them from the adverse effects of further trauma. Some correlational studies have found evidence to support these inoculation and maturation effects (e.g., Schlenger et al., 2002; Shrira et al., 2014). However, other studies have found that elderly people may be more susceptible than younger people to PTSD and other detrimental outcomes (e.g., Kun, Han, Chen, & Yao, 2009). Further, many researchers have Inclusion and Exclusion Criteria Studies were deemed eligible if they used an experimental methodology in a controlled environment to measure individuals’ psychological outcomes in response to media coverage of disasters or large-scale violence. The key terms were operationalized as follows: media coverage (the independent variable [IV]) included factual reports from TV, Internet, radio, newspaper articles, or realistic simulations of any of these. Disasters or large-scale violence included major accidents (e.g., plane crashes or multiple road accidents), natural disasters (e.g., hurricanes), acts of terrorism, war or combat, climate change, economic crises, and crime with the potential for casualties. Psychological outcomes (the dependent variable [DV]) included state anxiety, negative affect, fear, or perceived threat from the type of trauma in the exposure condition. The DV could also include other negative outcomes— either emotional or cognitive—such as anger or blame, or positive outcomes such as trust and empathy. One might expect outcomes such as trust to decrease after some media exposure, such as coverage of a terrorist attack, and other outcomes, such as empathy, to increase after other media exposure, such as coverage of a natural disaster. Eligible studies needed to either assign groups across the IV (e.g., media exposure to disaster or large-scale threat content vs. MEDIA EXPOSURE TO LARGE-SCALE VIOLENCE neutral content) or use repeated measures, with pre- and postexposure measures on the DV. Given the meta-analysis aimed to assess effect size variability across ages, genders, and locations, eligible populations included the general public in any geographical region. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Search Strategies and Data Extraction In June and July 2015, the PsycARTICLES, ProQuest, and Summon bibliographic databases were used to carry out a systematic search for experimental studies measuring psychological outcomes in the context of media exposure to disasters and large-scale violence. Keywords included “media coverage,” “media exposure,” “television news,” “disaster media coverage,” “newspaper,” “internet news,” “terrorism,” “war,” “natural disaster,” “virus,” “pandemic,” “crime,” “anxiety,” “negative affect,” “fear,” “threat,” “anger,” “positive affect,” and “experiment.” A series of search expressions were created for each disaster type to cover all relevant possibilities (for example, terrorism AND [“media coverage” OR “television news”] AND [anxiety OR “negative affect” OR fear OR threat OR anger OR “positive affect”] AND experiment). To ensure that the search was methodical and met recommended guidelines for meta-analysis research, the PRISMA search protocol (Moher, Liberati, Tetzlaff, Altman, & PRISMA Group, 2009) was used to record all articles identified, screened, and assessed for eligibility. The reference list of each eligible article was reviewed and assessed for additional relevant studies, and Google Scholar was used to search for relevant research by identified experts in the field. Separate eligible reports by the same research teams were examined to ensure the use of independent data sets. Finally, each study was assessed to determine whether adequate information was provided to ensure suitability of the stimulus material (exposure group media content), and whether the methodology was appropriate for the purpose of determining a causal relationship between the IV and DV. Studies were excluded if there were insufficient data to allow for the calculation of an effect size of at least one relevant DV. In total, 4,364 reports were identified through the database searches, with an additional 137 reports obtained through other sources, such as a search of literature cited in pertinent articles. After removal of duplicate items, 2,565 records remained. Screening via title and abstract (and full text when necessary) resulted in the exclusion of 2,523 reports that either did not employ an experimental methodology, did not manipulate an appropriate form of media exposure, or did not measure a relevant outcome. Of the remaining 42 eligible reports, three were excluded because of insufficient information regarding the stimulus media material, 20 were excluded because of methodologies that were incompatible with the research question (e.g., correlational studies), and four were excluded because of insufficient reported data to allow for calculation of an effect size of at least one relevant DV. Ultimately, the search resulted in 15 reports containing 18 studies suitable for inclusion in the meta-analysis (see Table 1). Some groups that were not appropriate for testing our hypotheses (such as those using a treatment condition prior to media exposure) were excluded; the experimental groups excluded from each study and the rationales for these decisions are outlined in Table 2. The overall sample size was 1,634, with 959 females (58.7%) and 675 males (41.3%). 319 Publication Status of Studies Several eligible unpublished reports were located during the search phase but were excluded because of incompatible methodologies or insufficient information. Thus, all studies included in the meta-analysis were published. Coding Procedures Following the recommendations of Cooper (2010), all studies were coded separately by two raters. The coding was consistent across raters (97% agreement) and disparities were resolved through discussion. Outcome types. The studies included in the meta-analysis measured a variety of psychological outcomes, most considered negative (e.g., state anxiety, anger, perceived threat), but some with positive valence (e.g., trust and empathy). To allow us to investigate whether type of outcome impacted effect size, each outcome in each study was coded in one of six broad categories— state anxiety, negative affect, fear, perceived threat, other negative outcome (e.g., anger), and positive outcome. Moderator variables. Based on the findings of correlational research and the theories espoused by previous researchers (as mentioned in the introduction), several study variables were coded with the aim of assessing their influence on the effect size of psychological outcomes. Moderator variables included intentionality (intentional, unintentional, mixture); media format (video, audio, print, static images, mixed), date of study (pre- or postSeptember 11, 2001), global location (U.S., Middle East, Europe/ GB, Asia, Australia, other), community sensitization to disaster type (exposure or no exposure to given disaster type within the previous 5 years), gender breakdown (percentage of females), and mean age of participants. Study quality. Given the potentiality for study quality to impact results, study design and the reliability of psychometric measures (for DVs) were coded as moderator variables. Study design was coded with four levels of empirical design: (a) pre–post measures with no control group; (b) experimental and control groups with no random assignment; (c) random assignment to experimental and control group—posttest only; and (d) random assignment to experimental and control group—pre- and postmeasures. The reliability of psychometric measures for the outcomes (DVs) was coded as a continuous variable, with a mean Cronbach’s alpha calculated for outcomes in each study (five of the 18 studies had missing data for this variable). Data Analysis All data analyses were performed using Comprehensive MetaAnalysis Version 3 (CMA; Biostat, Inc., 2013). As most studies reported results in the form of means (and standard deviations) for exposure and control groups, and based on the recommendation of Borenstein, Hedges, Higgins, and Rothstein (2009), Hedges’ g was chosen for the effect size metric. The meta-analysis provided a weighted mean summary effect size for negative psychological outcomes. Positive outcomes (e.g., trust and empathy) were included but were entered to reflect an opposite-effect direction when appropriate. For the main analysis, if outcomes of different types were reported for the same group of participants in a study, 101 58 58 120 120 60 67 67 80 22 40 40 40 40 40 40 116 248 100 100 100 237 43 43 84 84 100 100 40 78 78 78 Barlett and Anderson (2014, Study 2) Boyle (1984) Boyle (1984) Comer et al. (2008) Comer et al. (2008) Fischer et al. (2007, Study 2) Fischer et al. (2007, Study 4) Fischer et al. (2007, Study 4) Fischer et al. (2010, Study 1) Fischer et al. (2010, Study 3) Fischer et al. (2011, Study 1) Fischer et al. (2011, Study 1) Fischer et al. (2011, Study 1) Fischer et al. (2011, Study 2) Fischer et al. (2011, Study 2) Fischer et al. (2011, Study 2) Lightstone et al. (2005–2006) Ortiz et al. (2011) Shoshani and Slone (2008) Shoshani and Slone (2008) Shoshani and Slone (2008) Slone (2000) Slone and Shoshani (2006) Slone and Shoshani (2006) Slone and Shoshani (2008) Slone and Shoshani (2008) Slone and Shoshani (2010) Slone and Shoshani (2010) Williams and Khan (2011) Zeidner et al. (2011) Zeidner et al. (2011) Zeidner et al. (2011) 19.24 22.60 22.60 10.80g 10.80g 30.37 24.68 24.68 27.43 41.47 25.67 25.67 25.67 26.85 26.85 26.85 NR 9.24 22.65 22.65 22.65 34.70 22.65 22.65 22.65 22.65 23.00 23.00 NR 24.44 24.44 24.44 Mean age 73 83 83 61h 61h 52 81 81 63 73 50 50 50 50 50 50 78 54 50 50 50 50 50 50 50 50 50 50 50 59 59 59 %Femalea Other (Econ.) Acc./Nat. Acc./Nat. Terrorism Terrorism Terrorism Mixture Mixture Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Acc./Nat. Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Terrorism Disaster type U U U I I I M M I I I I I I I I I U I I I I I I I I I I I I I I Intent.b Video Video Video Video Video Print Images Images Mixed Images Images Images Images Print Print Print Print Video Video Video Video Video Video Video Video Video Video Video Video Video Video Video Media 3 1 1 1 1 3 3 3 3 3 3 3 3 3 3 3 4 3 4 4 4 4 4 4 4 4 4 4 3 3 3 3 Study designc U.S. AUS AUS U.S. U.S. EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB EUR/GB U.S. U.S. ME ME ME ME ME ME ME ME ME ME U.S. ME ME ME Location No Yes Yes No No No No No No Yes No No No No No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Sensitizationd Neg. SA Neg. SA PT PT Neg. Pos. PT PT NA PT Pos. NA PT Pos. SA SA SA Neg. Pos. SA SA Neg. SA Neg. SA Neg. SA NA PT Pos. Outcome .91 NR NR .88 .84 NR .86 .87 .84 NR .86 .79 .88 .86 NR NR .89 .85 .95 .94 .83 .94 .92 .91 .95 .94 .95 .94 .89 .90 .90 .94 Reliability (␣)e 1.028 .759 .532 .832 .341 .520 .401 .683i .561 1.950 1.021 .598 .183i .780 .456 .466i .857 .427 11.034 8.241 6.182i 1.662 5.122 4.925 6.588 .617 10.823 7.009 ⫺1.429 .768 .768 .447i ESf 1.028 .607 .607 .504 .504 .520 .542 .542 .561 1.950 .496 .496 .496 .542 .542 .542 .857 .427 7.910 7.910 7.910 1.662 5.023 5.023 3.603 3.603 8.916 8.916 ⫺1.429 .661 .661 .661 Mean ES Note. NR ⫽ not reported; U.S. ⫽ United States of America; AUS ⫽ Australia; EUR/GB ⫽ Europe or Great Britain; ME ⫽ Middle East; Acc./Nat. ⫽ large-scale accident or natural disaster; Econ. ⫽ economic crisis; SA ⫽ state anxiety; NA ⫽ negative affect; PT ⫽ perceived threat; Neg. ⫽ other negative psychological outcomes; Pos. ⫽ positive psychological outcomes. a The percentage of females in the sample to the nearest whole percent. Studies reporting a fairly even gender breakdown were coded as 50%. b Intent. ⫽ intentionality; I ⫽ human agency intentional event; U ⫽ unintentional event; M ⫽ mixture of intentional and unintentional events. c 1 ⫽ pre–post with no comparison group; 2 ⫽ experimental and control groups with no random assignment; 3 ⫽ random assignment experimental and control group—posttest only; 4 ⫽ random assignment experimental and control group—pre and post. d Sensitization indicates whether the location had a history of the disaster type portrayed in the media within 5 years prior to study publication. e Cronbach’s alpha values for reported reliability were averaged when more than one measure for that outcome type was recorded for the study. f Raw effect size for the outcome type, assuming independence of the data for each outcome. Raw effect sizes are shown averaged when more than one measure for the outcome type (e.g., for type other negative effect) was recorded for the study. g The mean age for child participants in the study (n ⫽ 90); a subgroup of mothers was also included (n ⫽ 30), but no age data were reported. h The percentage of females across both child and adult participants in the study (48% for child sample; 100% for adult sample). i For positive outcomes, a high score indicates a less positive outcome. n Authors (year, study no.) Table 1 Characteristics of Included Studies and Effect Sizes This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 320 HOPWOOD AND SCHUTTE MEDIA EXPOSURE TO LARGE-SCALE VIOLENCE 321 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Table 2 Sample Numbers Used and Excluded From Each Study and Exclusion Reason Authors (year, study no.) n used n excluded Barlett and Anderson (2014, Study 2) Boyle (1984) Comer et al. (2008) Fischer et al. (2007, Study 2) Fischer et al. (2007, Study 4) 101 58 120 (90 children, 30 mothers) 60 45 Fischer et al. (2010, Study 1) Fischer et al. (2010, Study 3) Fischer et al. (2011, Study 1) 80 22 parents 40 Fischer et al. (2011, Study 2) Lightstone et al. (2005) Ortiz et al. (2011) Shoshani and Slone (2008) 40 116 248 100 20 0 0 200 Slone (2000) Slone and Shoshani (2006) 237 43 0 77 Slone and Shoshani (2008) 84 84 Slone and Shoshani (2010) 100 200 Williams and Khan (2011) Zeidner et al. (2011) 40 78 0 0 Exclusion reason 0 0 60 mothers Excluded mothers who were given experimental training premedia exposure. 0 22 for each comparison test 0 22 children 20 the outcomes were averaged. Raw effect sizes (assuming independence of outcomes) for each coded outcome type were also calculated and are reported in Table 1). Subgroup analyses (for the categorical moderator variables) and metaregressions (for the continuous variables) were based on averaged outcomes for those studies with multiple outcomes. A subgroup moderator analysis was conducted to find weighted effect sizes for each type of psychological outcome represented (e.g., state anxiety, negative affect, positive outcomes). In order to make full use of available information regarding different types of outcomes, in this analysis, each outcome from studies reporting information for multiple outcomes was used. According to Borenstein et al. (2009), a random effects model is the appropriate computational model for most meta-analyses in the social sciences, in which it usually cannot be assumed that the primary studies will yield the same effect size. In contrast, a fixed effects model is appropriate when it can be assumed that all studies are measuring the same effect in the same population. In the case of the current meta-analysis, given the variations in study design, participant groups, locations, and outcomes measured, the effect sizes were expected to vary and a random effects model was used. Both experimental groups compared individually against control (neutral media) group. Children did not view media. Meaning group excluded, as this could be interpreted as an amelioration intervention that would affect outcome. As above. Excluded participants in experimental amelioration groups (intervention prior to media exposure). Excluded participants in experimental amelioration groups (intervention prior to media exposure). Excluded participants in experimental amelioration groups (intervention was post media exposure but Time 2 outcomes recorded after intervention). Excluded participants in both pre- and postexperimental amelioration groups. indicates that approximately 96% of the variance across studies is the result of difference in the true effect sizes, rather than sampling error. The high heterogeneity found here also supports the decision to explore potential moderator variables. Results Overall Mean Effect Size In order to test the major hypothesis that media exposure to disasters or large-scale violence results in negative psychological outcomes, a mean effect size was calculated for all studies included in the meta-analysis (k ⫽ 18). All effect sizes are presented as standard mean differences in the metric of Hedges’ g (see Table 1). All except one study (Williams & Khan, 2011, with g ⫽ ⫺1.43) showed a mean effect size indicating an increase in strength in negative psychological outcomes. The overall mean weighted effect size for negative psychological outcomes was large, g ⫽ 1.61 (standard error [SE] ⫽ 0.27, 95% confidence interval [CI] [1.07, 2.14], p ⬍ .001), indicating that across studies media exposure to disasters or large-scale violence had a significant effect on negative psychological outcomes. Heterogeneity Analyses Inspection of heterogeneity statistics revealed a significant Q statistic (Q ⫽ 378.53, p ⬍ .001) and a high I2 index (p ⬍ .001, I2 ⬎ 75%). These results indicate high heterogeneity, with the effect sizes across studies varying significantly (Cochran, 1954; Higgins & Thompson, 2002). This variation supports the decision to use a random effects model, as the I2 index (I2 ⫽ 95.51%) Moderator Analyses Method of moments metaregression examined the association between percentage of females in samples, reliability of measures, and mean age of samples with effect size. The percentage of females was significantly associated with the effect size (slope ⫽ ⫺0.059, SE ⫽ 0.025, 95% CI [⫺0.108, ⫺0.009], Z ⫽ ⫺2.31, p ⬍ .05). These HOPWOOD AND SCHUTTE This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 322 results indicate that the higher the percentage of female participants, the weaker the impact of media exposure on negative psychological outcomes. However, the regression slope indicated only a very small association. The metaregression for reliability of measures also found a significant association (slope ⫽ ⫺0.059, SE ⫽ 0.025, 95% CI [⫺0.108, ⫺0.009], Z ⫽ ⫺2.31, p ⬍ .05). These results indicate that the higher the mean reliability level of measures in each study, the stronger the effect of media exposure on negative psychological outcomes. Using a linear metaregression, the mean age of samples was not significantly associated with the effect size (Z ⫽ 0.30, p ⫽ .76), indicating that a change in mean age of participants did not correspond with a linear change in psychological outcomes. Because we predicted a nonlinear relationship (with effects for both children and older people predicted to be greater than young or middle aged people), a curvilinear metaregression using a quadratic model was used to test this hypothesis. The curvilinear model, although not significant (Z ⫽ ⫺1.60, p ⫽ .11), approached a better fit for the data than did the linear model. Results of the categorical moderator analyses are shown in Table 3. Intentionality of the event portrayed did not moderate the effect size, Q(1) ⫽ 2.73, p ⫽ .098. The effect size difference between the pre- and post-September 11 groups also was not significant Q(1) ⫽ 0.52, p ⫽ .47. However, this test lacked power, with only two studies being conducted prior to September 11, 2001. Grouping by media format did not yield a significant difference across effect sizes, Q(3) ⫽ 5.99, p ⫽ .112. As one subcategory of this variable was a mixture of the other formats, a repeat analysis excluded the study that used a mixed media format and yielded a result trending toward significance, Q(2) ⫽ 5.15, p ⫽ .076. Given that some subgroups were underrepresented (no studies used audio, three used print, and three used images), these results should be interpreted with caution. Global location significantly moderated the effect size, Q(3) ⫽ 36.17, p ⬍ .001, as did sensitization to disaster type, Q(1) ⫽ 18.54, p ⬍ .001. The results showed that studies conducted in the Middle East showed the highest effect sizes. The results also indicated that studies conducted in communities with a recent history (within the previous 5 years) of the type of disaster portrayed showed larger effect sizes than did studies conducted in communities without this type of threat salience. Given that the studies in the meta-analysis measured different types of psychological outcomes, we examined whether the type of outcome moderated the effect size. As the mean effect across outcomes for studies with multiple outcomes could not be used here, we ran this analysis with the assumption of independence of outcomes within each study (see Table 4). This assumption of a zero correlation between outcomes represents a conservative approach—it inflates the p value, lowering the statistical power to detect heterogeneity (Borenstein et al., 2009). The result of this moderator analysis was significant, Q(4) ⫽ 24.65, p ⬍ .001, indicating that the strength of the effect varied across the types of outcomes measured. State anxiety, the most commonly measured outcome type (k ⫽ 12), was associated with the strongest effect, g ⫽ 3.11, SE ⫽ 0.38, 95% CI [2.368, 3.849], Z ⫽ 8.23, p ⬍ .001. Publication Bias A potential source of bias in meta-analyses is publication bias, the potential for studies with significant results to be more abundant in the published literature and more easily located by re- Table 3 Results of Categorical Moderator Analyses Using Mean Outcomes for Participants Category Intentionality, Q(1) ⫽ 2.73, p ⫽ .098 Intentional Unintentional Pre- or post-September 11, Q(1) ⫽ .41, p ⫽ .523 Pre-September 11 Post-September 11 Media format, Q(2) ⫽ 5.15, p ⫽ .076 Video Print Static images Global location, Q(3) ⫽ 36.17, p ⫽ .000 U.S. Middle East Europe/Great Britain Australia Sensitization, Q(1) ⫽ 18.54, p ⫽ .000 Yes No Study design, Q(2) ⫽ 40.03, p ⫽ .000 Pre–post, no control group Random assignment, post only Random assignment, pre–post g 95% CI Z p k 1.939 .686 [1.289, 2.588] [.650, 2.022] 5.85ⴱ 1.01 .000 .315 14 3 1.134 1.699 [⫺.490, 2.758] [1.097, 2.301] 1.37 5.53ⴱ .171 .000 2 16 2.229 .642 .965 [1.478, 2.979] [⫺.748, 2.032] [⫺.460, 2.390] 5.82ⴱ .91 1.33 .000 .366 .184 11 3 3 .299 4.087 .744 .607 [⫺.676, 1.275] [3.115, 5.060] [⫺.167, 1.656] [⫺1.549, 2.764] .60 8.24ⴱ 1.60 .55 .548 .000 .109 .591 5 6 6 1 2.702 .356 [1.972, 3.432] [⫺.424, 1.136] 7.26ⴱ .89 .000 .371 10 8 .556 .519 4.066 [⫺.875, 1.987] [⫺.140, 1.178] [3.147, 4.985] .76 1.54 8.67ⴱ .446 .123 .000 2 10 6 Note. g ⫽...
Purchase answer to see full attachment