探討美國的科倫拜校園事件與桑迪·胡克小學槍擊案事發期間遊戲報導的視角：自由派與保守派的觀點分析 - 政大學術集成

全文

(1)國立政治大學國際傳播英語碩士學位學程 International Master’s Program in International Communication Studies College of Communication National Chengchi University. 碩士論文 Master’s Thesis. 探討美國的科倫拜校園事件與桑迪·胡克小學槍擊案事發期間遊戲報導的視角：自由派與保守派的觀點分析 From Columbine to Sandy Hook: Video Games in US Liberal and Conservative Newspaper Reporting During School Shootings. Student: T.F Swinburne 孫柯 Advisor: J.H.T Lin, Ph.D 林日璇博士中華民國 107 年 6 月 June 2018. DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(2) 探討美國的科倫拜校園事件與桑迪·胡克小學槍擊案事發期間遊戲報導的視角：自由派與保守派的觀點分析 From Columbine to Sandy Hook: Video Games in US Liberal and Conservative Newspaper Reporting During School Shootings. 研究生：T.F Swinburne Student: 孫柯指導教授：林日璇博士 Advisor: J.H.T Lin, Ph.D. 國立政治大學國際傳播英語碩士學位學程碩士論文. A Thesis Submitted to International Master’s Program in International Communication Studies National Chengchi University In Partial Fulfillment of the Requirement For the degree of Master of Arts. 中華民國 107 年 6 月 June 2018 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(3) Abstract. This thesis aims to discover how liberal and conservative US newspapers report on video game during times of school shootings from partisan viewpoints in newspapers, specifically the liberal New York Times, Washington Post, and the conservative Chicago Tribune and New York Post, along with the Wall Street Journal to add a neutral viewpoint to offer balance. Understanding how newspapers report on video games through different moral weightings is essential for further understanding the place video games inhabit in both the media and society at large. It needs to be understand whether or not video games were reasons for school shootings and how this discourse has changed from the Columbine Shooting in 1999, to the Sandy Hook shooting in late 2012. Video games are no longer a fringe culture, they are mainstream and further research needs to be conducted to understand how the media disseminates their views on video games to readers who base their perception of video games on frames produced with partisan influence. Through the theoretical framework of Blame Theory’s Path Model to Blame, a qualitative content analysis was performed to analyse newspaper articles through thematic coding using template analysis. This thesis contributes to existing literature because this field has not been researched thoroughly and there is a lack of consensus on how the media report on video games in current literature, particularly from a qualitative point of view. Keywords: Liberal, Conservative, Newspapers, Video Games, Content Analysis, Template Analysis, Media bias, New York Times, Washington Post, Chicago Tribune, New York Post, Wall Street Journal, Columbine, Sandy Hook.. i DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(4) Acknowledgements I would like to thank everyone that has made it possible for me to conduct my research. Firstly, I want to thank the Taipei Representative Office in Ireland for awarding an MOE scholarship. I want to thank my advisor, Tammy Lin, for offering her guidance and expertise on the study of video games. Her input and enthusiasm aided me greatly. I also want to thank my committee members. They took the time and effort to read through my many iterations of this thesis, and each time, gave me fantastic insights and feedback that improved this thesis. Finally, I want to thank my partner, Amber Lai, for being supportive and helping me to stay confident in my ability to perform academic research.. ii DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(5) Table of Contents 1. Introduction 2. Literature Review 2.1 Morality, Blame, Panics, Media, and Video Games 2.2 Moral Foundation Theory and Blame Theory 2.3 Moral Panics and Video Games 2.4 Media Bias: Conservative & Liberal Moral Leanings and Judgements 2.5 Media Portrayal of Video Games 2.6 Changing Perception of Video Games Research Questions 3. Methodology 3.1 Qualitative Content Analysis 3.2 Theoretical Framework - Blame Theory 3.2.1 Blame Theory and the Path Model of Blame 3.3 Template Analysis 3.3.1 Epistemological Approach of Template Analysis for This Thesis 3.3.2 Defining Codes and Themes 3.3.3 Priori Codes 3.3.4 Processes of Template Analysis 3.3.4 Coding Audits 3.3.4 Template Audits 3.4 Newspapers to be Analyzed 3.4.1 Presidential Endorsement Benchmark 3.4.2 Choosing a Neutral Newspaper 3.5 Organizing The Data for Coding 3.6 Data Collection Through ProQuest and Search Terms 3.6.1 Search Terms Employed 3.7 Comparing Newspapers for Partisan Agreement 3.8 Sampling Method 3.9 Issues in Data Collection and Refinement 3.9.2 Data Collection: Second Phase of refinement 4. Findings and Results 4.1 Template 4.1.1 Columbine Primary Reasons 4.1.1 Sandy Hook Primary Reasons 4.2 Columbine Template 4.2.1 Columbine Video Game Secondary Reporting 4.2.1.1 Liberal Video Game Secondary Reporting 4.2.1.2 Conservative Video Game Secondary Reporting 4.2.1.3 Neutral Video Game and Guns Secondary Reporting 4.2.2 Columbine Defending Reasons 4.2.2.1 Liberal Defending Reasons of Video Games 4.2.2.2 Conservative Defending Reasons of Guns 4.2.3 Columbine Other Significant Secondary Reporting - Guns 4.2.3.1 Liberal Significant Secondary Reporting- Guns 4.2.4 Columbine Minor Reasons. 1 3 3 4 7 10 13 15 16 17 17 18 19 21 22 22 23 23 25 27 29 30 30 31 31 32 34 35 36 38 39 40 42 43 43 45 45 47 50 51 52 54 55 55 57. iii DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(6) 4.2.4.2 Conservative Minor Reasons - Greater Society and Local Communities and Guns 57 4.2.4 Comparing Liberal and Conservative Reporting During Columbine 59 4.3 Sandy Hook Template 60 4.3.1 Sandy Hook Video Game Secondary Reporting 61 4.3.1.1 Liberal Defending of Video Games Vs Significant Reason of Guns 62 4.3.1.2 Conservative Reason of Video Games 63 4.3.1.3 Neutral Secondary Reasons Guns and Video Games 65 4.3.2 Sandy Hook Defending Reasons 67 4.3.2.1 Defending Reasons of (CT) Video Games and (NYP) Guns. 67 4.3.3 Sandy Hook Significant Reasons 69 4.3.3.1 Liberal Focus on Guns 69 4.3.3 Sandy Hook Minor Reasons 71 4.3.3.1 Liberal Minor Reasons of Mental Health 71 4.3.4 Comparing Liberal and Conservative Reporting During Sandy Hook 73 4.4 Comparing Columbine and Sandy Hook 74 5. Discussion 75 5.1 Video Games as a Partisan Issue 76 5.2 Changes of Reporting Over Time 77 5.3 Changing Moral Panics 78 5.4 Contributions to MFT and Blame Theory and future research 79 5.5 Concerns of Media Bias 79 5.6 Limitations 81 6. Conclusion 81 7. Bibliography 83 8. Appendix 104 8.1 Coding and Template Audits 105 8.1.1 First Codebook Draft 105 8.1.2 Second Codebook 109 8.1.3 Third and Final Codebook 112 8.2 Template Audits 114 8.2.1 First Template 114 8.2.2 First Audit Outcome 116 8.2.3 Second Audit Outcome 118 8.2.4 Final Audit Outcome 119 8.3 Priori Codes & ProQuest Number of Reports for School Shootings 121. iv DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(7) Table of Figures Figure 1 - FBI Crime Data 1964-1994 (FBI, 2015). ................................................................ 13 Figure 2 - FBI Crime Data 1995-2014 (FBI, 2015) ................................................................. 14 Figure 3 - Malle, et al., (2014) Path Model to Blame flowchart.............................................. 20 Figure 4 - Newspaper reporting on Video Game Violence between 1990-2017 for news reports ............................................................................................................................... 32 Figure 5 - School shootings in the US from 1996-2016 .......................................................... 34 Figure 6 - Data collection: First Phase for Columbine 1999 ................................................... 37 Figure 7 - Data Collection: First Phase for Sandy Hook 2012/13 ........................................... 37 Figure 8 - Data collection: Second Phase for Columbine 1999 ............................................... 38 Figure 9 - Data collection: Second Phase for Sandy Hook 2012/13........................................ 39 Figure 10 - Distribution of newspaper reports between Columbine and Sandy Hook ............ 39 Figure 11 - Liberal Reporting of Columbine School Shooting in 1999 .................................. 44 Figure 12 - Conservative Reporting of Columbine School Shooting in 1999 ......................... 44 Figure 13 - Breakdown of the Columbine Reporting .............................................................. 59 Figure 14 - Liberal Reporting of Sandy Hook School Shooting 2012/2013 .......................... 60 Figure 15 - Conservative Reporting of Sandy Hook School Shooting 2012/2013 .................. 61 Figure 16 - Breakdown of Sandy Hook Reporting .................................................................. 73 Figure 17 - Liberal Comparison of Columbine and Sandy Hook Reporting ........................... 74 Figure 18 - Conservative Comparison of Columbine and Sandy Hook Reporting ................. 75. v DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(8) 1. Introduction. On January 29, 1979, Brenda Spencer entered San Diego’s Cleveland Elementary School and shot dead 2 adults, and injured 9 children. Whilst in the middle of a stand-off with the police, she explained to a reporter that she committed America’s ‘first high-profile school shooting’ with the phrase “I just don’t like Mondays” (Los Angeles Times, 2001, Apr 18). Spencer’s statement did not make any sense of her violence or actions, and only confused society further. Since 1979, there have been numerous high-profile school shootings across the United States (Follman, Aronsen & Pan, 2017) and newspaper reporting attempted to make sense of the senseless violence. After the Columbine High School Massacre in 1999, the media blamed video games such ‘Doom’ and ‘Quake’ for influencing senselessness behind the school shooters (Brown, 1999, April 24). After the 2011 Norway Attacks in which Anders Behring Breivik killed 77 people, reports indicated that he used the first-person shooter ‘Call of Duty Modern Warfare 2’ and the massively multiplayer online role-playing game ‘World of Warcraft’ to train for the massacre in Oslo (Sutter, 2012, April 20). Because the mass media frames what ‘views’ and ’behaviors’ are acceptable and which ones are outside the mainstream (Graber & Dunaway, 2014), to understand how the multibillion dollar industry of video games in the US is being portrayed in newspaper reporting is essential. Video games have been subject to widespread controversy through episodic moral panics. Past research has shown that video games are implicated as a reason for school shootings, regardless as to whether or not school shooters played them (Ferguson & Colwell, 2017). It is important to research how the media blames video games as a reason for school shootings because the media has - particularly the elite media - an influence over their audience. People with different moral leanings selectively choose news and research that supports their moral framework (Rothmund, et al., 2015) and partisan news can have 1 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(9) consequences on how people view an issue (Shoemaker & Reese, 2013). Newspapers can disseminate news that discusses the potential adverse effects of video game violence, thus influencing its audience to have a negative perception of video games and video game culture (Ferguson & Colwell, 2017). Currently, there has not been sufficient qualitative or in-depth research into how video games are reported on in the media and this thesis’s findings showed how conservative, liberal and neutral newspapers reported on video games during two school shootings; Sandy Hook and Columbine. Through a template analysis, 145 newspaper articles from all partisan leanings and from 1999 to 2012/13 were coded and analysed to find how video game reporting from liberal, conservative and neutral newspapers viewed video games across during the Columbine and Sandy Hook school shootings, and how the reporting changed over the 13-year difference between the two massacres. A template analysis was used, the theoretical framework of Blame Theory’s ‘Path Model to Blame’ was used to guide it. Through this thesis, new avenues of research can be opened up to analyze through preexisting frameworks and methods, how video games are reported on. Understanding the $30.4 billion industry of video games in 2017 is important because 67% of all households own a dedicated gaming console, and 96% of parents are ‘very confident’ the ERSB (Entertainment Rating Software Board) rating system for video games is accurate (Entertainment Software Association, 2017), yet newspaper reporting has still been critical of video games. Understanding how newspapers from partisan opposites report on video games will give insights into how society views video games, and how that view has changed over time.. 2 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(10) 2. Literature Review For the purposes of this thesis, it is first important to define school shootings. Previous research has neglected to do so and leaves the interpretation of what a school shooting is quite open. According to Muschert (2007), school shootings can vary. For example, the school shooting of Columbine was a ‘Rampage Shooting’, while other mass shootings fall into ‘Mass Murders’, ‘Terrorist Attacks’, ‘Targeted Shootings’ and ‘Government Shootings’. Unlike the rest, rampage shootings garner the most media attention. For the purposes of this research, school shootings will be considered rampage shootings. Current literature is fragmented on how to research video games, and there have only been a handful of attempts to understand how video games are reported in the media. What is clear from literature, is that research needs to focus on the moral judgements people make about video games, through moral foundation theory, blame theory and other contributions such as moral panics through the lens of moral panic tsunamis. The following literature was formed to create a cohesive argument from fragmented theories and research to make sense of something as senseless as a school shooting and where video games belong in newspaper reporting.. 2.1 Morality, Blame, Panics, Media, and Video Games The concept of morality has been debated and elaborated upon since the earliest philosophers, but for the purposes of this study, morality is defined as being a system of essential contentions our society lives by (Donagan, 1979). Where morality fits into the research of video games in the media is how the readership react to it in conjunction to school shootings. Rothmund, et al. (2015) found that when there is a threat to ‘nonviolence’ as a ‘moral value’, people (pacifists) will support ‘political actions’ against ‘violence’ (or in this. 3 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(11) case, video game violence). They will ‘selectively’ search for research that supports their moral leanings and call for political action against video games. To better understand the morality behind these moral judgements and how they lead to blame, Moral Foundation Theory and Blame Theory will be discussed, along with how these two theories explain why newspapers and readers have different moral weights and judgements towards school shootings and their ideological leanings. This will explain why newspapers of different partisan views (conservative and liberal) morally judge and blame video games differently during school shootings.. 2.2 Moral Foundation Theory and Blame Theory Moral foundation theory has been used to explain how and why people make moral judgements. These moral judgements are defined by Haidt (2001) as being ‘the way we evaluate whether actions or the character of something are good or bad depending on the virtues held by individuals or cultures.’ In MFT, these moral judgments are based upon the difference in moral weighting of five sets of moral intuitions as outlined by Graham, Haidt & Nosek (2009, p. 1029): 1. Harm/care, 2. Fairness/reciprocity 3. Ingroup/loyalty 4. Authority/respect 5. Purity/sanctity. These five intuitions can be weighted differently, hence why people have different moral values and judgements on certain situations and actions; such as conservatives and liberals (Haidt & Graham, 2007, Graham, Haidt & Nosek 2009). Conservatives and liberals have. 4 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(12) divergent sets of moral foundations, and Inbar, Pizarro & Bloom (2009) found that conservatives tend to be more easily ‘disgusted’ by immoral issues than liberals. Day, et al., (2014) had empirically shown that moral foundation frames could affect people’s political attitudes. Hence moral foundation theory is important in the consumption of news for liberals and conservatives because conservative and liberals hold different moral weights on the five sets of moral intuitions. Backing past research (Pizarro & Bloom, 2009, Graham, et al., 2009), Fulgoni, et al., (2016) empirically found that news articles from partisan sources in the US had framed news stories differently based on their political bias. Haidt & Graham (2007) used their 5 sets of moral intuitions to show how moral judgements were made based upon the weights placed on the intuitions for liberals and conservatives. In terms of school shootings, further into the literature review, it will be shown that newspapers and the media do make partisan judgements on video games based upon their ideological leanings in conservatism and liberalism when setting the agenda for news reporting on school shootings. Moral judgements based on Haidt & Graham’s, 2007 and Graham, Haidt & Nosek’s (2009) research can also help explain another aspect of moral judgements; blame. Blame Theory is a comprehensive theory that explains what blame is, what it is not and how blame judgements can be made through the Path Model of Blame (Malle, et al., 2014). In relation to MFT, blame theory does distinguish the different moral judgments made as a result of the different moral weights people have, and are explained by Malle, et al. (2014) as:. 1. Setting and affirming norms, such as declaring a prohibition, expressing an imperative, or avowing one norm as overriding another. 2. Evaluating events (outcomes, behaviors) in light of those norms, such as by judging an event as bad, good, wrong, or (im)permissible.. 5 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(13) 3. Evaluating agents for their involvement in such norm-relevant events, such as by judging someone as morally responsible, blameworthy, or praiseworthy. (p. 148). As explained by Malle, et al., (2012), “blame is a moral judgment that has a cognitive and a social nature” (p. 313) and according to Malle, et al., (2014) it contains ‘four fundamental properties of what blame is’: namely 1. It is cognitive and social 2. It is social regulation 3. It is a concept that relies on social cognition 4. It is a concept that requires warrant And further explains what blame is not: 1. It is no only anger 2. It is not just event evaluation 3. It is not only a wrongness judgement. Blaming an agent requires people to know their own ‘behavior-guiding norms’ and those that are blaming others must know an act is a norm-violating behavior (Malle, et al., 2012). Those behaviour-guiding norms are different, and as seen in MFT, they can be weighted different. As such, blame is a well-researched concept that differentiates itself with ‘wrongfulness’. Blame and wrongfulness are different, for example, someone might think homosexuality is wrong, but they don’t blame the person who is a homosexual. They might feel an activity is wrong, but they are not warranted to feel blame. In other words, blame. 6 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(14) judgements target an agent through warrant, while wrongfulness targets solely behavior, and is thus only one aspect of blame and not the same concept (Malle, et al., 2014). However, an important aspect of blame is the reasons for an action that requires a blame judgement to be enacted. Reasons are explained in Malle, et al’s., (2014) Path to Blame model as being the reasons for acting and in this case, the reasons why schoolers shooters intentionally. Reasons are essential for ‘moral perception because they are the reasons that determine the meaning behind the action of an agent.’ Malle (2004, 2014) states that ‘people will always find it painful to not know the reasons behind the action that causes a blame judgment.’ So while blame theory can help explain the how blame is assigned to school shooters, it also takes the added step of reasons, a way to look into the moral perception of the agent issuing blame. The reasons of blame will be an important part of this study. Both MFT and blame theory can also explain the phenomenon of moral panics, which ensue often after school shootings, and video games have often been the focus of media reporting as a result.. 2.3 Moral Panics and Video Games Moral judgements from the five intuitions, that lead to blame all tie into the concept of ‘moral panics’ which ensues after events that cause panic and crises, such as school shootings. The concept of ‘moral panic’ is defined as being a period when individuals, groups, and even concepts are differentiated from society as being a ‘threat to societal values and interests’ and the backlash from those who feel morally violated is one of a stereotypical fashion which can induce the mass media to create moral barricades. These barricades could be manned by ‘editors,’ ‘bishops’, ‘politicians’ and left and right-thinking individuals (Cohen, 2002).. 7 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(15) Sørensen (2013) explained that moral ‘panics’ ensues from pacifists - found in Rothmund, et al’s., (2015) research - whose norms and values have been violated by normbreaking events and agents. An important aspect of moral panics is mass media, because the media frames what ‘views,’ and ’behaviors’ are acceptable in mainstream society in their news coverage (Cohen, 2002; Graber & Dunaway, 2014). Panic is the result of there being no easy explanation behind the violence. When violence is meaningless, those who are most morally appalled by the actions will find an explanation to make sense of a senseless act, such as a school shooting. Moral panics are defined into different classifications, based on the moral panic’s “intensity, duration, and social impact.” ‘Teddy Boy,’ ‘Mod’ and ‘Rocker’ culture in the 1950s – much like recent goth culture – created a moral panic against youth culture. Other moral panics were widespread and ‘transformed social landscapes’ such as the witch-craze in 16th-17th century Europe. Other moral panics include ‘drug panics’ and ‘child abuse outcries’. These individuals and groups are essentially ‘folk devils’ or a démon du jour (Cohen, 2002; Garland, 2008, p 6; Griffiths, 2010) and garner the blame for events and actions that break the social norms. Moral panic induced by school shootings about video games was researched by Karlsen (2015) and they found that from video games such as ‘Deathrace’ in 1976 to Mortal Kombat in 1993, there have been media and moral panics on the issue of video games for several decades. Ferguson (2008) refers to the moral panic behind video game violence as a ‘straw man’ argument. In his research, he mentions the causal relationship between video game violence and aggression as being nothing more than a manifestation of ‘moral panic’ by those with moral frameworks that are offended by violence. Interestingly, Sørensen (2013) states media reporting on video games invariably spikes when a school shooting occurs. He took the example of the tragic Emsdetten school shooting. 8 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(16) in Germany. His conclusion stated that video game violence was ‘seasoning’ to the bigger picture of a school shooting. While the vast majority of readers are encumbered by the moral and media panic of a school shooting, media outlets will supplement their coverage with ‘filler,’ namely, violent media effects on gamers. Past research backs up Ferguson & Colwell’s (2017) theory of the ‘moral panic tsunami.’ They used the example of the Sandy Hook Elementary School Shooting, in which Adam Lanza killed 27 people. For 11 months, the media speculated that he played violent video games. When the investigation was over, investigators uncovered he played nonviolent video games. Thus, people look for the presence of ‘research that affirms their fear,’ and they are ‘critical of research that critiques their fear’. This feeds into the media who promote this fear affirming research in popular reports and subsequently through vertical and horizontal communication between politicians, media, and the masses (Entman, 2003; Brants; Voltmer, 2011). The media then picks up on the importance of the political interest and reaffirms the need for further research into the harmful effects of the issue that is causing moral panic which in turn causes more moral panic which re-initiates the process again in which individuals’ reject critiques and opt for research that reaffirms their own bias. (Ferguson & Colwell, 2017) As stated by Chyi & McCombs (2004) and McGinty, et al. (2014), most coverage of school shootings - and subsequent reporting on SMI and video game violence takes place proficiently within a 1-2-week period. Thus, the moral panics that ensue are shortterm moral panics, which aim to change the behavior of those within society (Rohloff & Wright, 2010). Essentially, moral panics are aroused from issues relating to the violation of social norms from the point of view of people who have different weighted moral leanings, such as conservatives and liberals.. 9 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(17) 2.4 Media Bias: Conservative & Liberal Moral Leanings and Judgements It is almost cliché to mention Entman’s (1993) statement that the media cannot tell us ‘what to think,’ but rather ‘how to think,’ but it is still a powerful message today as it was in 1993. The significance of this statement is backed up with research that shows news and views in the media are framed to distinguish what behaviors and views are acceptable in mainstream society (Graber & Dunaway, 2014). A more updated version of Entman’s statement is by Lutz (2007, p. 3), who states that “It’s not what you say, it’s how you say it.” Framing can be understood to be the method in which the media processes information to make sense of the world for an audience who possibly cannot understand it without a frame for it to fit into, which in other words is an ‘interpretive schema.’ Such a scheme aids in the classification and interpretation of information (McCombs & Shaw; 1972; Goffman, 1974; Scheufele & Tewksbury, 2007). However, despite the power of the media, even the best message is still at the whimsical of the audience, who view it through the lenses of their ‘emotions, preconceptions, prejudices and pre-existing belief ‘(Lutz, 2007). This is only made more complex by the ideological leanings that news organizations adopt. In the US media, there is a dichotomy between conservative and liberal news and views, which is best described by Shoemaker & Reese (2013): “Conservatives accuse the media of concentrating on negative news and expressing a liberal bias. Liberals accuse the media of kowtowing to conservative presidents” (p. 59). Such a dichotomy is still relevant today. Media bias is the selectiveness of details that are featured in the media, which ultimately distort the same facts that other news organizations have to create a different news story (Gentzkow & Shapiro, 2006). This media 10 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(18) bias often follows partisan lines, conservative and liberal. This was found by tracing newspapers political affiliations through who they supported in presidential elections, and this is indicative of what media bias a newspaper may have (Puglisi & Snyder, 2015). Graham, Haidt & Nosek (2009) found that -- through experiments with a basis in moral foundation theory -- that liberals tended to be far more concerned with issues relating to ‘harm’ and ‘fairness’, while conservatives were most concerned about issues relating to ‘ingroups’, such as ‘authority’ and ‘purity.’ In other words, liberals tend to see the optimistic side of the human condition, while conservatives see the pessimistic aspects more pronounced (Graham, Haidt & Nosek, 2009). This backed up Haidt & Graham’s (2007) study that showed liberals morality, in their five sets of intuition, relies heavily on fairness/reciprocity and in harm/care, while conservatives rely heavily on all five sets of intuition. Thus, conservative and liberals weight their moral judgements differently, and their blame too. The choice between conservative and liberal information is, as Iyengar & Hahn (2009) aptly put it, a choice between ‘red and blue.’ Their research found that Fox News receives far more conservative viewers, while MSNBC News garners more liberal ones. They explain this is due to competition; news agencies will create a niche to distinguish themselves and this is done on a partisan basis. Stroud’s (2010) research also found that partisan selectivity in exposure to media has contributed vastly to political polarization. Scheufele’s (1999) ‘process model of framing’ best explains how audiences, journalists, institutions, and frames mingle together to form the news we see. Audiences influence the news with their input and views, which affects journalists as audiences whose work is then influenced by their media organization, which then produces the frames to match both their political leanings and the attributes of their audience.. 11 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(19) As audiences are polarized, the outcome has been the phenomenon of the echo chamber effect (Boutyline & Willer, 2017). This echo chamber leads to ‘homophily’, which is the tendency of individuals to want to be involved with ‘homogeneous groups’ and be associated with people that share their political leanings. This was empirically shown by Colleoni, Rozza & Arvidsson (2014) by their research which found conservatives and liberals on Twitter had high homophily rates. The news’ agenda and frame production is heavily influenced by the audience’s tendency to be political, either in a conservative or liberal sense. Althaus & Tewksbury’s (2002) study into the agenda-setting of the New York Times among readers of paper and online versions found that even after five days of exposure to the New York Times their participant’s issue agendas had changed to fit the New York Times. Entman (2007) had also shown that in the media, political and ideological influences lead to a ‘distortion bias’ in which news distorts reality, or a ‘content bias’, when a news organization favours only one side of an issue over another. Finally, a ‘decision-making bias’ wherein a journalists’ bias seeps into the content. Thus, liberal and conservative leanings are entrenched in our media, and the information we see is simply a frame of reality that could be distorted for an audience who want to see reality in a certain way to affirm their fears and satisfy their moral judgments and blame judgements. Added to this is the Rashomon. The effect is taken from a film by the same name which explains a situation wherein four people have four different opinions on the same crime they witnessed. The media, especially around the time of school shooting, suffer from the Rashomon effect, as they construct their frames to fit the partisan nature of their audience and to also fit into their moral leanings (Muschert, 2007). Thus, even with the same facts, the perception of media organizations audiences will distort the same information to match the underlying moral weights existing within themselves.. 12 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(20) 2.5 Media Portrayal of Video Games Current research on how video games are reported on in the media is somewhat controversial. Researchers (Ferguson & Colwell, 2017; Martins, et al., 2013; Bushman & Anderson, 2001) all argue divergent, but equally important viewpoints. Firstly, Bushman & Anderson argue (2001) that the media has gone from weakly supporting the idea that violent media causes aggression to arguing that it simply does not. They blame networks because, as they claim, it is their agenda to prevent people from realizing the harmful effects of watching the violent media their broadcast. They tie their argument to the FBI’s uniform crime reports data from 1951 to 1999. Fig 1, presents the data for the uniform crime reports from 1964 to 1994.. Figure 1 - FBI Crime Data 1964-1994 (FBI, 2015).. They argue that crime rates rose as television became more prevalent. In their content analysis, they found that since 1985, the media failed to incorporate ‘media violence effects’ research findings in their reporting. They also stated that there was a strong ‘negative correlation’ between the ‘average effect size’ and the ‘average news report rating.’ They blame the media’s irresponsibility for the rise in violent media and its deterrence to reporting on violent media as a bias that is spreading misinformation.. 13 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(21) This research is flawed, however. The flaw lies in presenting the crime rates from 1951 to 1999 and suggesting that violent media was a factor in the rise. Tonry (2014) presented a multitude of research into why crime rates rose and, ultimately why they fell in the Western world. He also criticized the notion that violent media affected people to conduct more crime because as he explains, even cops watch the same media and do not conduct the same crime as a straight line through the origin as violent media become more prevalent. If Bushman & Anderson’s (2001) insights were correct, then the crime rate should continue to grow, not fall as shown in figure 2.. Figure 2 - FBI Crime Data 1995-2014 (FBI, 2015). The main thesis of Bushman & Anderson’s (2001) research conforms to Martins’, et al’s, (2013) findings that found in recent times; the media tends to take a less firm stance or at least a neutral tone on violence in video games causing aggression. The media does not follow the same ‘trajectory’ of media violence research, in the sense that at first there was an ‘early period of uncertainty,’ a ‘middle period of consensus’ and a ‘third period of renewed critiquing’ (Martins, et al., 2013). While researchers were uncertain of the relationship between video games and aggression, the media, as a moral paragon and conduit for the opinions of those most morally obliged, took the stand that video games were dangerous. News media can either take a neutral tone on its portrayal of video games or outright criticise 14 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(22) the medium of entertainment according to Ferguson & Colwell (2017). Ferguson (2011, December 07) stated that video game violence is a non-issue and the findings by researchers in the past were biased to the point that most findings are irrelevant. Many researchers have stated the media coverage is bias against stating the media effects of video game violence, Ferguson & Colwell (2017) states quite the opposite. In their research, they state that the media is pressured into publishing news on the dangers of video game violence. Ferguson & Colwell (2017) use moral panic theory and the Goldilocks effect to illustrate how the media, along with the general public and politicians can be morally cued to publish reports video game violence. What can be deduced from current literature, is that there is no consensus on how the media portrays video games. Literature does, however, point the way for further studies – such as this – to ask pertinent research questions to act as springboards for other aspects of research that past research has failed to address, which will be discussed next.. 2.6 Changing Perception of Video Games Muschert (2007) found that since the 1990’s, the occurrence of school shootings has remained steady over time and that the media tended to concentrate on school shootings from the ‘late 1990s until the early 2000’s. Muschert (2009) had found that news media will change frames overtime to maintain the ‘salience of a story’. This confirms Chyi & McCombs (2004) who found that during the framing process, frames changed over time. While frames change to maintain salience of school shootings in the media, how school shootings framing over time has not been researched extensively. Muschert & Carr’s (2006) study on the media salience and change framing across events from nine school shootings between 1997-2001 utilized Chyi & McCombs’ (2004) analytical scheme that they used to analyze news reporting on the Columbine Shooting. They used this analytical scheme and. 15 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(23) found that between 1997-2001, between nine different school shootings, frame-changing occurred over time, particularly when they plotted the ‘core frames across events’ and found that frame-changing occurred in the later school shootings. Whereas in earlier school shooting coverage, ‘community-present frame’ was most salient, later school shootings saw ‘societal-present frame’ as being most salient. By the end of the school shootings, both frames were equally salient (Muschert & Carr, 2006). However, these researchers only found the trends, and did not take a qualitative perspective on the articles. Some studies (Bushman & Anderson, 2001; Martins, et al., 2013; Ferguson & Colwell, 2017) discuss how in recent times, new media takes a less firm stance on video games affecting aggression. Martins, et al., (2013) discussed how the change in media reporting on video game violence has been at odds with what scientific research states. Thus, it can be seen that the framing of video games has changed over time. Frames on both school shootings and video games have changed over time. This indicates that news media reporting on these two is not the same across time. While some studies have factored this into their research, as previously mentioned above, there has not been any recent research to contrast the late 1990s to the 2010s new media coverage of school shootings, or even their relationship with video game. Therefore, the following research question is proposed.. Research Questions To understand how video game reporting during school shootings is different between conservative and liberal newspapers and between the late 1990s and early 2010s, the following research questions will be answered:. 16 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(24) 1.. How did Conservative and Liberal newspapers report on video games during. the Columbine school shootings? 1.1.. Was there a significant difference between conservative and liberal. newspaper reporting on the issue of school shootings in terms of video games during the Columbine Shooting? 2.. How did Conservative and Liberal newspapers report on video games during. the Sandy Hook school shooting? 2.1.. Was there a significant difference between conservative and Liberal. newspapers reporting on the issue of shootings in terms of video games during the Sandy Hook school shooting? 3.. Did US Liberal and Conservative newspaper reporting on video games. change between the Columbine school shooting in 1999 and the Sandy Hook school shooting in 2012/3. 3. Methodology This thesis examined, through the theoretical framework of Blame Theory’s ‘Path Model of Blame” and through the coding of a template analysis, how five newspapers reported on video games as reasons for school shootings. The newspapers were divided into three groups: firstly, two liberal newspapers, The New York Times (NYT) and Washington Post (WP), secondly, two conservative newspapers, The Chicago Tribune (CT) and The New York Post (NYP), and thirdly, the neutral Wall Street Journal (WSJ. 3.1 Qualitative Content Analysis While past research on the issue of school shootings has been focused on quantitative approaches, this content analysis was performed qualitatively. Past research emphasized. 17 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(25) ‘quantitative factors’ such as size and frequency, which do not always correlate to impact. According to Shoemaker & Reese (2013) it is not correct to assume ‘quantitative factors’ are the ‘main determinants’ of ‘media impact’ and although past research did yield important findings, it did not “capture the context within which a media text becomes meaningful” (Newbold, et al., 2002, p. 84). One such study on the framing of the Columbine School Shooting by Chyi & McCombs (2004) showed that content analysis can be an effective way to analyze how newspapers report and frame the issue of school shootings in the US media using the three variables; date of publication, space of the frame and time of the frame (Chyi & McCombs, 2004). While they presented interesting findings, their content analysis did not gauge the sentiment, or the message behind the text. Instead, they focused on far more quantitative aspects. A qualitative content analysis examines how the reporting and audience have a relationship through the polysemic meaning of the text (text has multiple meanings to different readers). Qualitative content analysis tries to determine meanings behind the text that the audience is likely to read (Macnamara, 2005). Understanding the polysemic meaning of texts behind partisan perception of video games aided this research. This content analysis was deductive, instead of inductive. Deductive coding in this case, drew from existing theory to analyze the texts (Joffe & Yardley, 2004). At first, an inductive approach through the framework of Grounded Theory was considered for this study. However, despite the fragmented nature of this research area, there was still enough focus among researchers to find a suitable theoretical framework and a method of analysis.. 3.2 Theoretical Framework - Blame Theory Forming a theoretical framework was difficult. This area of research has been mainly focused on quantitative theoretical frameworks and analytical approaches. Chyi & McCombs. 18 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(26) (2004) analyzed school shootings news frames across time. Their findings showed that frames of school shootings change over time, but they did not comprehensively analyze the content of their frames and instead coded them into general categories for the purposes of a having a statistical outcome. That is not to say their research was lacking. It was a good benchmark to begin to understand school shootings in media reporting. Their work has been replicated by Muschert (2009), and his research brought up, briefly, the issue of video games in the reporting; but not in a comprehensive or in-depth way. Conducting qualitative research required a different approach. Based on the section of the literature review on morality, moral judgements, and blame, this literature highlighted the need to find a theoretical framework elsewhere in the social sciences. Blame Theory stood out due to the Path Model of Blame (Malle, et al., 2014) having a strong conceptual structure to answer the research questions, and guide the content analysis.. 3.2.1 Blame Theory and the Path Model of Blame Previous studies used MFT to analyze partisan judgements and leanings, or used frame theory to quantitatively analyze newspapers. Both approaches serve to find trends within bigdata and neglect the polysemic meanings inside texts. To compensate for this, Blame Theory was used as the theoretical framework for this study. This model offers a conceptual structure to allow researchers to process information that generate judgements based on a seven step process that determines whether blame is warranted, and to what degree depending on the responses - as outlined by Malle, et al., (2012), and Malle, et al., (2014) to the following seven factors: 1. Event Detection - The norm violating/ negative event (newspaper reporting of school shooting) 2. Agent causality - Did this agent cause the event? (School shooters) 19 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(27) 3. Intentionality - Yes/no (Did the agent intentionally cause or intend for the event to occur?) 4. Obligation - Yes/No (Did the agent have an obligation to prevent or stop the event? if no, low to no blame) 5. Capacity - Yes or no (Did the agent have the capacity to prevent or stop the event? If no, low to no blame) 6. Reasons - If there was intentionality, then there have to be reasons to justify the intention. For this study, reasons for justifying the intention through reasons will be important. 7. Degrees of blame - Depending on what reasons were given, what type of obligation and capacity the agent had, there will be different varying degrees of blame.. The focus of this study was on ‘reasons’, while also utilizing the procedure to help guide the coding process of the data sets by following steps 1-7 as shown in figure 3.. Figure 3 - Malle, et al., (2014) Path Model to Blame flowchart.. Malle, et al., (2014, p. 151). 20 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(28) The events being detected were school shootings, the agents were the school shooters, and they were found to have intention, and this led to understanding the reasons for the school shooting. This model allowed for a structured approach to answering the research questions. However, in order to make sense of the data set through this theoretical framework, a template analysis was chosen to code and make sense of the newspapers articles.. 3.3 Template Analysis A Grounded Theory and Sentiment Analysis approach were both considered, but neither of these approaches were appropriate for this particular research. Instead, a Template Analysis as outlined by King (King, 2004, Symon & Cassell, 2012, Brooks, et al., 2015) was chosen to analyze the units of analysis. Template analysis is a type of thematic analysis and is explained by Brooks, et al., (2015) as a form of analysis that:. “.... emphasises the use of hierarchical coding but balances a relatively high degree of structure in the process of analysing textual data with the flexibility to adapt it to the needs of a particular study.” (p. 203). Consideration was taken to find a method of analysis that had a high degree of structure during the analyses of text. King (2004) argues that although template analysis makes use of codes and coding of data, it is ‘not as prescriptive as Grounded Theory and does not follow its realist methodology’. The usefulness of template analysis is that it can be ‘used within a range of epistemological positions’ (King, 2004).. 21 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(29) The main issue finding a suitable method of analysis was that it needed to also be flexible to accommodate this thesis, considering previous studies were largely quantitative in nature and previous approaches in frame theory and MFT theory were not suitable. Template analysis was a suitable approach for analysing newspaper textual content because it focuses on a mixture of having both descriptive and interpretive aspects to the coding of the data (King, 2004). In other words, the findings of this research were not solely describing listed codes -- but rather -- how the codes were interpreted, and how they answered the research questions.. 3.3.1 Epistemological Approach of Template Analysis for This Thesis As outlined by Brooks, et al., (2015) template analysis can be taken from a ‘contextual constructivist position’, meaning, researchers assume there are ‘multiple and polysemic meanings to be made from textual data.’ The same approach will be used in this thesis. However, to maintain the template and results’ quality, audits were conducted as the data was coded and when the template was refined. These audits were included in the appendix in section 8.1 and 8.2.. 3.3.2 Defining Codes and Themes A code is a piece of text that is indexed due to its relatedness to a theme which a researcher interprets as important and it is also a process to transform textual data into a standardised form (King, 2004, Babbie, 2015). In thematic analysis, codes are representative of themes. Themes are defined by Braun & Clarke (2006) as ‘representing a patterned meaning in data’ and codes capture important information that is relevant to the data and research questions. Symon & Cassell (2012) define themes further as being ‘repetitive and not simply isolated phenomenon and also being distinct enough from each other with some overlap, but not too much.’ 22 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(30) 3.3.3 Priori Codes Many codes and themes were found throughout the coding phase, but priori codes were also included. As a deductive template analysis, priori codes were included because these codes were deduced from previous literature. Some priori codes -- which can all be found in the appendix (section 9.3) -- included: 1. Partisan Newspapers 1.1. Liberal 1.2. Conservative 1.3. Neutral 2. Reasons School Shooting 2.1. Video Games 2.2. Guns (NRA) 2.3. Mental Health 2.4. Society 2.5. Parents Template analysis is a mixture of bottom up (IPA & Grounded Theory through exploratory means to analyze data) and top bottom (Matrix analysis & framework analysis using prior theories to guide research) because it allowed this research to adopt priori codes and themes -- or themes and codes that are derived from previous research that can aid in current research -- while also allowing this research to discover new codes and themes (Symon, & Cassell, 2012). While these priori themes and codes were useful, they were also not fixed. Symon, & Cassell (2012) explain the flexibility of Template analysis allows for the ‘deletion of priori codes and initial codes and the scope can be changed if a code is too narrowly or broadly defined and it can be it can be re-defined.’. 3.3.4 Processes of Template Analysis The process of performing a template analysis differs from thematic analysis. Fereday & Muir-Cochrane’s (2006) process outlined six stages for their thematic analysis that share and differentiate certain steps with template analysis. Rather than focusing on developing a code 23 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(31) book to rely on throughout the entire process of coding, template analysis instead relies on creating initial templates that evolve as more data sets are coded. This will be explained below by Brooks, et al (2015) in six steps that this study used to code the data.. 1. Become familiar with the unit of analysis; which in this study, was sentences, paragraphs, and entire reports. This leads into the second step. 2. Perform a preliminary coding of the newspapers. This is done to highlight parts of the text that will contribute to understanding how video games are blamed as reasons in the data. Priori codes can be included. 3. With emerging themes, organize them. The organization of themes were placed in a hierarchy. This hierarchy allowed this study to analyze texts in different levels of specificity by organization themes in a lateral manner to showcase links between themes and also allowed for the distinction between main themes, sub-themes. This hierarchy of themes was also used to present the data in the findings, results, and discussion (Fereday & Muir-Cochrane, 2006, Symon & Cassel, 2012). 4. Once the initial coding was completed, the template was defined. Deciding when a template can be defined depends on the data, and how long it will take to find a subset of themes that best capture the issue of the study in the data takes time. 5. The template was modified through applying the initial template to further data. New data was examined from the data-set and pinpointing where themes fit and do not fit resulted in the modification of the template. The modifications were not made until at least several units of analysis had been examined to justify reorganization of the template. Overall, these modifications served to enrich the template to accommodate new insights and was an ongoing process of analysis that differed from other thematic analysis techniques (Fereday & Muir-Cochrane, 2006) that prescribed a fully formed. 24 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(32) codebook before the entire data set could be analysed. While it is important to form an initial template to further analyze the data, the template is formed throughout the process of coding all of the data. 6. Once the template became comprehensive, it was applied it to the full data set to further refine the template and to ensure all themes were relevant. Once the template reflected the full data set and explained the meaning of the data to answer research questions, the template was then analyzed further for results and discussion. It is important to note that King (2004), Symon & Cassell,2012, and Brooks, et al., (2015) all state that judging whether a template is the final version is difficult and researchers will usually come to the decision that their template is finished because it is sufficient to answer their research questions and when there is no longer any ‘substantial sections of data are relevant, but cannot be coded’. The data was eventually exhausted and two comprehensive templates were formed, and the findings and results were then formulized and written.. 3.3.4 Coding Audits. An essential part of coding is auditing. The data has to be exhausted before it can be analysed. Part of coding and forming templates was analysing and coding at the same and refining codes to both be representative of the data and answer the research questions. The coding process took one month of reading and the data was eventually exhausted for relevant information for further codes that were related to this study The first codebook was an amalgamation of codes found through exploration of the datasets. Rather than limit the focus of the study through selective coding, a wider spectrum was chosen to allow a more holistic understanding to prevail. A total of 66 codes were found across liberal, conservative and neutral newspaper reporting. This codebook best represents 25 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(33) the first attempt at making a template for both the school shootings. In later codebooks, partial and full reasons were combined. The reasoning for this will be explained below. As a side note, the codebooks found in section 8.1 used denotations such as: (C) denotes Columbine, while a (SH) denotes Sandy Hook. The first audit of the codebook resulted in the first initial codebook, as seen in section 8.1.1. The second codebook audit focused the study further. After coding the entire dataset again and attempting to form themes in a template, it was found that certain codes were not relevant to 1) the research questions and 2) and were not relevant because they were not referenced enough. As such, the following edits to the codebook were made (mainly reducing codes, and putting codes together to create greater society and local community) in section 8.1.2. This codebook best represents the second attempt at creating a template in section 8.2.3 A third codebook audit was finalized after a final audit. In this finalized version, partial and focused reasons were removed in place of a general reason. This was because the differences between partial and focused reasons were not significant enough and did not add anything to answering the research question or gave a better understanding of the data. As such, partial and focused reasons were combined and explained to a degree in the findings.. Of the original 66 codes found, only 25% of them were used in the final codebook and template (17). The codebook and templates were formed simultaneously and were crafted over time. This had the benefit of creating a template that best emulated the codebook while it was being crafted. By coding and forming a template at the same time, they were interlinked at each audit.. 26 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(34) 3.3.4 Template Audits Keeping in line with King (2004, 2015), template audits were conducted. These audits were conducted to ensure that the template best exemplified the coded data, while also being useful to answer research questions. As the codebook became more refined through rereading datasets and putting codes into themes, and deleting and adding codes, so too did the template change. There were three major stages of template audits and they are found below. The first template that was formed had four themes, namely; Video Game Secondary Reporting, Video Game Primary Reporting, Other Reasons for the School Shooting, Defending Reasons. Two templates were formed, not one. This was because a single template with the information from five newspapers across two time periods was too complex and would be difficult to understand. Instead, the template was broken into two parts; 1) The Columbine reporting and 2) The Sandy Hook reporting. This allowed Columbine and Sandy Hook datasets to be researched separately to answer the research questions and for them also to be more easily compared and contrasted to answer research questions. This is shown in section 8.2.1 of the appendix. This first initial template was found to be too confusing. The codes under the themes were very general, and were not descriptive of what information for liberal and conservative newspapers was most relevant and salient. To help showcase in the template the differences across liberal and conservative reporting, the template was refined further by the importance of codes through the process of picking codes with the highest amount of references and relevancy to answering the research questions, while codes with the lowest references and relevancy were removed. The second audit led to the template being more refined and mirrored the coded data in terms of taking the number of coding references and their context into the process of forming 27 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(35) the template. The themes were broken into four main groups and were given more in-depth summaries. 1. Video Game Secondary Reporting - This theme was placed as the most important part of the template because it is the most relevant information needed for this study to answer research questions. The reporting was broken into partial and full reasons. Partial reasons were relevant because video games were not seen as the sole reason for the school shootings, but simply a factor among many. Being placed as the full reason for school shootings was uncommon and was only present in mostly conservative newspaper reporting. 2. Significant Secondary Reporting - Understanding how newspapers reported on video games was not enough to answer the research questions. Considering video games were seen as one factor among many as to why school shootings may occur, it was essential to explain what other significant reasons were reported to have been a reason for the school shooting. 3. Minor Reasons - Outside of video games, and other significant reasons for school shootings, there were other minor reasons listed as perhaps leading to school shootings, and it was important to understand them. 4. Defending reasons - Some reasons such as video games and guns were defended from being reasons for school shootings. Defending reasons were found to be important in the template hierarchy because they offered an insight into which reasons both partisan views deemed defendable in their reporting. While showcasing how partisan newspapers report on video games is important, it is also essential to show what it is they defend as well. Significant secondary reporting and minor reasons were not as relevant as defending reasons and this was taken into account for the second audit outcome for the templates.. 28 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(36) The second audit in section 8.2.3 of the appendix shows the template in a more holistic sense. As stated before, the importance of Defended reasons was taken into account. Also taken into account was that the number of references of particular sections of the template were not substantial enough to be mentioned, and because they were not necessary for the research questions, they were removed. After reviewing the second audit and refining the codebook, a final audit was performed. The final audit, found in section 8.2.4 of the appendix shows the finished template. The Columbine and Sandy Hook templates were both refined further by removing sections that, although relevant for context building and understanding the general news reporting behind video games, were not necessary for answering the research question and were not referenced enough to become relevant. As such, the final audit gave the finalized template. The audits were a time-consuming part of the thesis and this summarized section only accounts for the basic outcomes of the audits. Performing audits takes time, concentration, and an in-depth understanding of the documents being researched. Audits cannot be rushed and take time to form.. 3.4 Newspapers to be Analyzed Originally, the NYT was the sole newspaper to be analyzed. This was because as Lee (2002) states, the NYT is considered part of the ‘fifth estate’ and it is part of the elite media that have a prestige that can influence their readers and consequently their perceptions of reality. Weaver & Bimber (2008, p. 19) used the NYT in their research because they stated it was “intrinsically important, or is treated as a proxy for the agenda of the entire news system”. The NYT along with the WP were both considered to be the ‘elite media’ and carry far more influence on their audience (Shoemaker & Reese, 2013). Added to this, they are also two of the largest liberal newspapers in the US according Cision (2016, May 11) with the 29 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(37) NYT having a daily circulation of 2,101,611 newspapers, and the WP having 356,768 newspapers in 2016. The CT and NYP were chosen because they are right-leaning and are both counted among the top 10 newspapers regarding circulation in the United States in 2016 (Cision, 2016, May 11), with the NYP circulating 424,721 newspapers daily, and the CT circulating 384,962 newspapers daily.. 3.4.1 Presidential Endorsement Benchmark The benchmark to decide on which newspapers could be considered liberal or conservative was based on which political candidate newspapers endorsed for US presidential elections. Chiang & Knight (2011) found that newspapers endorsed liberal and conservative presidents, which in turns influences their readers. Thus, the benchmark for partisan leanings for many newspapers was based on their endorsements of presidential candidates. While the NYT and WP overwhelmingly supported Democrat/liberal candidates, the NYP overwhelmingly endorsed Republican/conservative. The CT presented some interesting endorsements, namely, until 2008, they only endorsed Republicans, but this changed in 2008 when they to supported Barack Obama’s 2008 and 2012 campaign. However, they did not support Democrat or Republican candidates in 2016 (Veltman, 2016). For most of the CTs history, it has largely favored Republican/conservative candidates and can be considered a slightly conservative newspaper, or at least having conservative tendencies.. 3.4.2 Choosing a Neutral Newspaper To best gauge where the median point between conservative and liberal reporting is, a neutral newspaper had to be chosen. Unlike section 3.4.1 wherein presidents could be used to denote the partisan standpoint of a newspaper, the same could not be said for neutral newspapers. Instead, Mitchell, et al., (2014) reported that the Wall Street Journal is one of the 30 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(38) few newspapers and media outlets that is trusted by both liberals and conservatives, and is also considered a neutral newspaper through their reporting. Following this logic, the Wall Street Journal was chosen as the neutral newspaper for this study.. 3.5 Organizing The Data for Coding Five newspapers were analyzed in the years 1999 and 2012/2013. Ten data-sets were gathered and this presented an issue on how best to approach the coding process. For this study and practicality, the data set was split into two parts, the 1999 Columbine Shootings -with five data sets -- and the 2012/2013 Sandy Hook School Shootings -- also with five datasets from five newspapers -- with the intention to analyze the two different time periods of newspaper reporting separately and then comparing the two large data-sets once they were coded and analyzed. This allowed for the comparison of liberal newspapers with one another, and the same with conservative newspapers across both time frames, as well as allowing the comparison of both time frames.. 3.6 Data Collection Through ProQuest and Search Terms To collect enough articles to conduct a content analysis, the ProQuest newspaper database was used. The ProQuest platform, according to Popik (2004), is still the ‘gold standard for historical database searching’ and it is very ‘successful with big city newspapers.’ ProQuest, according to Bingham (2010, p. 227), is “the leading digitizer of American newspapers.” However, a study by Ridou, et al. (2012) found that while databases such as ProQuest and LexisNexis all contained the same newspaper articles, search terms are what ultimately cause differences among databases due to these databases employing different criteria for searches. As much as ProQuest is a reliable database, it is so essential to be aware of its limitations. 31 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(39) 3.6.1 Search Terms Employed Search words and phrases are like strainers. They are used to identify relevant units of analysis for research purposes. Precision was essential in this study as search words yield high or low precision based on the manner they are inputted into the search bar. For example, if 35 articles were found and only 25 were relevant to the study, that would mean there is a 71% precision rate (Stryker, et al., 2006). The closer this precision to 100%, the better. After an open search term phase and refining search terms through reading newspaper articles, the search phrase ‘video game’ and the school shooting incident were chosen. For example, ‘Columbine video game’ would be used in the ProQuest Database to search for documents about how newspapers view video games in relation to school shootings. The Initial search term used was ‘video game school shooting’. A clear trend emerged from 39,353 reports that showed spikes occurring in 1999 (3,568) and 2013 (3,030).. Figure 4 - Newspaper reporting on Video Game Violence between 1990-2017 for news reports. 32 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(40) The 1999 spike can be explained by the Columbine school shooting in April of that year and the aftermath of newspaper reporting of the tragic event. The 2012/2013 spike also, like 1999, was due to a School Shooting, which occurred in Sandy Hook that garnered a media following. Through this process, it was found that school shootings caused spikes in video game reporting, which conforms to Sørensen’s (2013) findings. Further research into the 1999 and 2012/2013 school shootings revealed they had been among the worst in the US mass shooting history according to the “US Mass Shootings, 1982-2017: Data From Mother Jones' Investigation” (Follman, Aronsen & Pan, 2017). The report identified the dates, number of victims, the mental health of the shooter and also the type of mass shooting, whether it was a ‘spree’ or a ‘mass shooting.’ Below is a summary of all mass shootings that took place in schools between 1996-2016.. Case. Location. Date. Year. Fatalities Injured. Total victims. Venue. Umpqua Community College shooting Roseburg, Oregon. 01/10/2015. 2015. 9. 9. 18. School. Marysville-Pilchuck High School shooting. Marysville, Washington. 24/10/2014. 2014. 5. 1. 6. School. Isla Vista mass murder. Santa Barbara, California 23/05/2014. 2014. 6. 13. 19. School. Newtown school shooting. Newtown, Connecticut. 14/12/2012. 2012. 28. 2. 30. School. Oikos University killings. Oakland, California. 02/04/2012. 2012. 7. 3. 10. School. Oikos University killings. Oakland, California. 02/04/2012. 2012. 7. 3. 10. School. Northern Illinois University shooting. DeKalb, Illinois. 14/02/2008. 2008. 6. 21. 27. School. Virginia Tech massacre. Blacksburg, Virginia. 16/04/2007. 2007. 32. 23. 55. School. Amish school shooting. Lancaster County, Pennsylvania. 02/10/2006. 2006. 6. 5. 11. School. Red Lake massacre. Red Lake, Minnesota. 21/03/2005. 2005. 10. 5. 15. School. Columbine High School massacre. Littleton, Colorado. 20/04/1999. 1999. 13. 24. 37. School. 33 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(41) Thurston High School shooting. Springfield, Oregon. 21/05/1998. 1998. 4. 25. 29. School. Westside Middle School killings. Jonesboro, Arkansas. 24/ 03/1998. 1998. 5. 10. 15. School. Figure 5 - School shootings in the US from 1996-2016. Follman, Aronsen & Pan’s (2017) research show that both the Columbine High School Massacre and the Newtown School Shooting (Sandy Hook) had the highest number of victims for school shootings. Although the Virginia Tech Massacre had more total victims, this massacre took place on a university, and not a high school and was excluded from this study. The search terms agreed upon 1) “Columbine video games” and 2) “Sandy Hook video games”. Newtown and Sandy Hook were interchangeable for the name of the Newtown Connecticut school shooting, however, “Sandy Hook” yielded a higher volume of news reports. These search terms yielded the best newspaper articles across all five news agencies regarding content and quantity.. 3.7 Comparing Newspapers for Partisan Agreement While both liberal and conservative newspapers will be compared in how they blamed video games as reasons for school shootings in their reporting, each liberal and conservative newspaper was compared with its partnering liberal or conservative newspaper. For example, the NYT and WP was compared to determine if there was an agreement between both newspapers on their ideological stance. This was also conducted between the CT and NYP. This will prevent an unreliable reading of whether ideological leanings interfere with coverage. If corresponding partisan newspapers (liberal NYT & WP, conservative NYP & CT) are in agreement about how they blame and portray video games in their reporting, then their. 34 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(42) partisan views can be said to be similar. If not, then other factors such as internal workings of the newspapers might explain the difference in reporting.. 3.8 Sampling Method There were multiple units of analysis that were derived from individual newspaper articles from a sample consisting of the NYT, WP, CT, NYP and the WSJ. the units of analysis consisted of sentences, paragraphs, entire articles, and titles of newspaper articles. While random sampling is a good option for analyzing the research questions, for the use of traditional media, it was first decided to take the first three months of reporting from when school shootings happen. Originally, the news was to be split into segments of weeks of news (Hester & Dougall, 2007). However, just as Chyi & McCombs (2004) found and likewise Muschert (2009), initial news reports dealt with reporting events, and later coverage, as supplemental coverage, dealt with other issues; which are more salient to this study. (Sørensen,2013) found that newspaper reporting on video games was supplemental too, and thus, initial newspaper reporting split into weeks would not be the best method to gather newspaper articles on the issue of video games during school shootings. When collecting the datasets, it was found that single weeks of newspaper reporting only contained at most, a small handful of reports, and this was only more pronounced during the Sandy Hook school shooting. Instead, the first three months of reporting were taken, as it was seen that the first two months of reporting yielded the highest reporting, while the third month saw a decline in reporting, before the eventual irrelevancy of the news story took over. The rationale behind this sample method was taken from the ‘The Issue-Related Cycle’ which explains the five stages of a news story and its salience namely: 1) Pre-problem stage 2) Alarmed discovery stage 3) Realizing cost of progress 4) Decline of public interest and 5) Post-problem stage (Chyi & McCombs, 2004). In short, samplings will be taken from specific. 35 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.

(43) the first three months of reporting with the search criteria of ‘(school shooting name) + video game’ search term.. 3.9 Issues in Data Collection and Refinement The data collection was successful and all datasets were collected. However, there was concern that ProQuest may have created duplicates, or that the search terms had not been focused enough to collect meaningful newspaper reports for this study. However, it was found that the majority of articles that needed to be removed were mostly duplicates from the WP dataset on Sandy Hook. Other irrelevant articles were found throughout the initial reading of the data sets and were subsequently removed.. 3.9.1 Data Collection: First Phase of Refinement At first, it was postulated that the search strategy might have been lacking, but after reviewing the searching process, and verifying that the search term strategy was in fact not at fault, it was concluded that the ProQuest platform was at fault for duplicating newspaper articles. When all news was collected for the Columbine Shooting in 1999, there were 156 articles that included several different types of reporting. To keep the analyzed data uniform, ‘news’ articles were chosen. With the exception of news, all other forms of newspaper reporting were interlinked and multiple articles were under ‘commentary’, ‘feature’ and ‘review.’ To prevent overlap, news articles were chosen to reduce possible overlap. After refining the Columbine dataset, the initial 156 articles were refined to 100 news articles, collected from the five newspapers.. Columbine. New York Times. 40. News. 26. 36 DOI:10.6814/THE.NCCU.IMICS.007.2018.F05.