![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
Terrorist attacks often dominate news coverage as reporters seek to provide the public with information. Yet, not all incidents receive equal attention. Why do some terrorist attacks receive more media coverage than others? We argue that perpetrator religion is the largest predictor of news coverage, while target type, being arrested, and fatalities will also impact coverage. We examined news coverage from LexisNexis Academic and CNN.com for all terrorist attacks in the United States between 2006 and 2015 (N = 136). Controlling for target type, fatalities, and being arrested, attacks by Muslim perpetrators received, on average, 357% more coverage than other attacks. Our results are robust against a number of counterarguments. The disparities in news coverage of attacks based on the perpetrator’s religion may explain why members of the public tend to fear the “Muslim terrorist” while ignoring other threats. More representative coverage could help to bring public perception in line with reality.
Keywords:
Acknowledgments
Earlier versions of this article appeared in the Washington Post’s Monkey Cage and in the Hidden Brain series on NPR. E.K. would like to thank the Global Studies Institute at Georgia State University for supporting her during this research. E.K., A.B., and A.L. would like to thank Lindsey DeWick, Val Durojaiye, David Revzin, Paris Stroud, Zenaida Torres, and Tamilore Toyin-Adelaja from the Honors College at Georgia State University for their invaluable research assistance. We would also like to thank Belen Lowery-Kinberg for thoughtful comments on framing our argument, Emil O.W. Kierkegaard for suggesting additional robustness checks for our models, and numerous others who asked questions or proposed additional counterarguments that we have tested for in the present version of the article.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
2 https://www.theatlantic.com/politics/archive/2017/02/trump-centcom-media-terror-cover-up/515823/; http://time.com/4489405/americans-fear-of-foreign-terrorists/
3 In the current study, the definitional criteria for what constitutes terrorism have been established in the development of the Global Terrorism Database (GTD) by the National Consortium for the Study of Terrorism and Responses to Terrorism (Citation2016). According to the GTD Codebook, terrorism is “the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation.” Additional details about the definition of terrorism used in the GTD are available at http://www.start.umd.edu/gtd/using-gtd/. See Schmid (Citation2013) for a more detailed discussion of the challenges related to defining terrorism, along with consideration of over 250 definitions that have been applied over time.
8 If this were binary, it would assume an hours-long foot search and a month-long hunt through the wilderness are the equivalent. If we count duration, then that implies the few days-long search for the Tsarnaev brothers that shut down Boston is less meaningful than the 48-day search for Eric Frein through the Pennsylvania wilderness. Given the diversity of what a manhunt can entail, we do not think it is advisable to control for this in a regression model.
9 Starting in 2006, an increasing percentage of Americans used the Internet as their main source of news. http://www.pewresearch.org/fact-tank/2013/10/16/12-trends-shaping-digital-news/ Since the news sources used for this study include both print and online newspaper articles, we started our analysis in 2006. In years prior to 2006, we may see fewer articles overall since print was more common and is subject to space constraints.
10 The Global Terrorism Database is a systematic and unbiased source that codes terrorism at the incident-level around the world from 1970 to 2017. The GTD is the most comprehensive and complete dataset available on terrorism. At the time of data collection, 2015 was the most recent year of data released by the GTD.
11 While we wanted to include searches from sources across the political spectrum, such as Fox News and Huffington Post, neither has a searchable archive going back to 2006 and email requests for archive access were not answered.
12 It is beyond our current scope to conduct a systematic study of television and radio coverage from both national and local stations across a decade span. Furthermore, broadcast media have a fixed amount of airtime so coverage disparities should be exacerbated. Including TV and radio coverage in our study would likely bias the results in favor of larger or more sensational events that dominate news coverage.
13 By the end of 2016, all known perpetrators had either pled guilty or gone to trial with the exception of Robert Lewis Dear. Dear is currently not competent to stand trial, so we expect occasional coverage of this going forward. Otherwise, we do not expect any ongoing coverage of the incidents, perpetrators, or victims listed in this dataset.
14 For example: Dylann Roof’s attack sparked debate about the Confederate Flag and gun control; Robert Lewis Dear’s attack led to discussion about gun control and abortion rights; the Boston Bombing increased discussions about immigration; and, the San Bernadino attack generated a discussion about immigration, gun control, and Apple refusing to unlock the perpetrator’s iPhone.
15 There are five major, national media outlets in our dataset: CNN.com, The New York Times, Wall Street Journal, The Washington Post, and USA Today.
16 All variables were double coded, inconsistencies in coding were discussed, and final codes were agreed upon for all variables in each incident. In a few instances where coding could be disputed, we estimated models both ways and the results were unchanged.
17 In terrorism, the perpetrator is often unknown so treating these as missing data and dropping the incidents is not appropriate. We recognize that for incidents where the perpetrator is unknown, it is possible that some were committed by Muslims but there is no way to know this. Essentially, there are three categories of perpetrator: Perpetrator Known & Muslim; Perpetrator Known & Not Muslim; and Perpetrator Unknown. Even when the individual perpetrator is unknown, we often know the group responsible so “perpetrator unknown” is not a theoretically sound category on its own, though we account for these incidents in robustness checks. In the models reported, we collapsed Perpetrator Unknown and Perpetrator Known & Not Muslim into a single category (0) and compared to Perpetrator Known & Muslim (1). To ensure that our results are not an artifact of whether or not the perpetrator is known, we also estimated all models where Perpetrator Unknown or Perpetrator Known & Muslim are collapsed into a single category (0) and compared to Perpetrator Known & Not Muslim (1). Across all models reported in the main text and the appendix, attacks where the perpetrator is known and not Muslim do not receive a significantly different amount of news coverage. In contrast, incidents where the perpetrator is known and Muslim receive significantly more coverage in all models. These findings give us additional confidence in our conclusions.
18 This is common for terrorism: approximately 13% of incidents globally are claimed (Kearns, Conlon, & Young, Citation2014) and 40% are attributed to a particular group (GTD, 2016). Even when the individual perpetrator is unknown, we often know the group or movement responsible. For example, attacks claimed by the Animal Liberation Front still send a clear message even in the absence of an arrest or identification of the individual(s) responsible. Thus, simply considering attacks where the perpetrator is unknown is not appropriate in terrorism studies. Instead, we control for unknown responsibility in two ways. First, we created a dummy variable for incidents where neither the perpetrator nor group are known. Second, we created a dummy variable for incidents where the perpetrator, group, and motive are all unknown.
19 We include the 2012 Sikh temple shooting in Oak Creek, Wisconsin and the 2015 attack on the Sikh bus driver in Los Angeles in this calculation. Evidence suggests that these attacks were Islamophobia-inspired and the perpetrators were unaware of the difference between Sikhs and Muslims.
20 The number of people wounded may also impact the amount of coverage that an attack receives. The vast majority (96.3%) of attacks wounded fewer than 10 people. Five attacks had more than 10 wounded: the Austin IRS attack, San Bernadino, Fort Hood, the West Texas Explosion, and the Boston Bombing. While casualties likely impact coverage, injuries are not of the same magnitude as fatalities. If we were to include the counts of both, this would assume that fatalities and casualties have the same impact on media coverage and that the relationship is linear. Rather, to account for the non-linear relationship between casualties and coverage, we logged the number wounded. The correlation between the number of fatalities and the log of number wounded is 0.63 so including both variables in a model introduces concerns of multicollinearity. We created an additive variable (number killed plus log of number wounded) to bluntly account for the impact of casualties on coverage, though this measure is difficult to substantively interpret. As shown in the appendix (Models A1-A20), results are substantively and statistically similar across all models.
21 All models are estimated with bootstrapped standard errors to minimize the impact of outliers with the small number of observations. Akaike Information Criterion (AIC) and Baysian Information Criterion (BIC) are presented to compare model fit where lower values suggest greater congruence with the true model. The extent to which one model is preferred to another depends on the magnitude of difference between model fit statistics (Raftery, Citation1995). Models discussed in text have either a weak or positive difference between alternatives.
22 A high proportion (N = 36, 26.5%) of the attacks in these data did not receive any news coverage. Thus zero-inflated negative binomial regression models were also estimated. Vuong tests of the zero-inflated negative binomial versus a standard negative binomial indicate that the negative binomial models are preferred.
23 Descriptive statistics for each variable are relatively unchanged, as shown in the appendix.
24 The next most covered attack, Faisal Shahzad’s attempted bomb in Times Square, received less than half the coverage of these. By the statistical definition, 17% of the cases are outliers due to the skewed distribution of coverage. Yet, there is not a sound argument for dropping all of these observations from the dataset since this is the reality of media coverage for these attacks.
Additional information
Notes on contributors
Erin M. Kearns
Erin M. Kearns is an Assistant Professor at the University of Alabama. Her work has been funded through a number of sources, including the National Consortium for the Study of and Responses to Terrorism (START), and has been featured on CNN, NPR, the Washington Post, and Vox. E.K. contributed to the conceptualization of the project and its empirical strategy, oversaw the data collection, cleaning and coding efforts, conducted the analyses, and wrote the methodology, results, discussion, and conclusions.
Allison E. Betus
Allison E. Betus is a presidential fellow in the Transcultural Conflict and Violence Initiative and a PhD student at Georgia State University. Her primary research focuses on domestic terrorism, white supremacy, and how media influences and reinforces public perceptions of what is and is not terrorism. A.B. wrote the literature review, contributed to the theory, and assisted with data collection, cleaning, and coding.
Anthony F. Lemieux
Anthony F. Lemieux, PhD, is Director of the Global Studies Institute, and Professor of Global Studies and Communication at Georgia State University. As a social psychologist working across disciplines, he has extensively studied motivations for terrorism and violent extremism, with research funding from the US DOD, DHS, and NSF. A.L. contributed to the conceptualization of this research, including methodological elements of data sourcing, inclusion criteria, and variables.