Research in Social Sciences and Technology (RESSAT)
E-ISSN: 2468-6891
30
Fatal Encounters: A Content Analysis of Newspaper Depictions of the Deaths of Unarmed
People of Color at the Hands of Law Enforcement or Security Personnel
Deanna Jacobsen Koepke 1, David A. Thomas2, & Alexis Manning3
Abstract
Research has been conducted for several decades on the framing of stories in the media. This study
furthers that field by analyzing the way newspaper articles report on the deaths of unarmed people
of color at the hands of law enforcement and security personnel between 1999 and 2017 to
determine if local and national print media frame these stories using similar terminology and
concepts. Tabular and graphical analysis from the Leximancer program, as well as statistical
analysis (Chi-square, p 0.0001), all demonstrate that local and national newspapers do not use
similar terminology and concepts when presenting these stories. Therefore, the null hypothesis is
rejected.
Key words: Print media, framing, law enforcement, deaths, Leximancer
Introduction
Trayvon Martin, Philando Castile, Alton Sterling, and Freddie Gray were all unarmed people of
color who lost their lives after an interaction with law enforcement or security personnel. They
are also names most of the public learned due to the extensive media coverage their situations
received. One might consider these occurrences to be the kind of thing about which only a local
audience would care. However, as Vakili (2016) writes, national media cover local events when
they believe it is either interesting or important to a national audience. Because many people in
the US have formed opinions on these and similar cases based in part on the way that the media
have portrayed them, the authors of this study wanted to know what that media coverage looked
like. They also wondered if there were differences in the ways these stories were presented to the
public based on the proximity of the newspaper to the story. The authors agree with Vakili (2016)
when he says that while newspapers are no longer the primary vehicle for news delivery that they
1 Assistant Professor of Sociology, University of Providence, deanna.koepke@uprovidence.edu
2 Professor of Mathematics, University of Providence, dave.thomas@uprovidence.edu
3Undergraduate Student, University of Providence, alexis.retterath@uprovidence.edu
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
once were, they continue to play an important role in the distribution of information. Not only do
they distribute this information, but they also put a frame around it, potentially influencing
audience reaction. Formally our research question is, “Do local and national newspaper articles
frame reports of deaths of unarmed people of color at the hands of law enforcement or security
personnel using similar terminology and concepts?”
Literature Review
Framing is commonly understood as the ways that material can be presented to suggest how the
audience should understand it (Jacobsen Koepke, 2009). Entman (2005) tells us that some aspects
of a story are highlighted while others are downplayed to drive a certain interpretation of the story.
Meyers (2004) reminds us that facts do not have inherent meaning. Meaning is acquired as the
facts are organized into stories. This organization can be done via the media, and it shapes society
and how those within the society understand events (Lee & Solomon, 1990). Further, framing has
been demonstrated to influence public opinion (Iyengar & Simon, 1993).
Many people in the U.S. have traditionally trusted and believed that which is reported in
newspapers (Gitlin, 2003). Even though one regularly hears the accusation of “fake news”, this is
usually directed at social media and online sources. For example, in Hunt’s (2016) article, she
addresses many aspects of “fake news”, all of which are attributed to Reddit, 4chan, Twitter, online
message boards, Facebook and “the social web”. In fact, she clearly contrasts these outlets with
such traditional newspapers as the New York Times and the Washington Post. Thus, the trust that
many in the US have for mass media allows us to continue to believe they are socially responsible
and report the facts objectively (Ehrlich, 2005). We regularly assume they have more insight into
the matters being reported than our experiences give us (Lee & Solomon, 1990), and this leaves
us open to the interpretations of those who write and present the stories.
Most people have an inherent understanding that not all the news can be reported. Thus, they
further trust that the topics presented by the media are appropriately important and the facts were
chosen with care. Newspapers have been known to criticize various sides of issues to maintain
the impression that the reporting is fair and objective (Jacobsen Koepke, 2009). Yet the very
telling of a story by a person means this is not possible (Vakili, 2016). Humans bring bias with
Koepke, Thomas, & Manning
them, and this bias can be seen in the decisions made about which details will be included in a
story and which will be excluded in an effort to keep the audience interested, suggesting how the
information should be interpreted (Jacobsen Koepke, 2009). Stories are hung on a “peg” (Gitlin,
2003), or placed in a context that readers will understand. Thus, very complex information may
be simplified (Gitlin, 2003), leading to additional shaping of opinions due to a lack of exposure to
the full depth and breadth of an event. Pipkins (2017) reports that the media guide many audience
members in their understanding of social problems by the very way that they define the problems
themselves. As these messages are reinforced over time, the public begins to consider their view
to be “common sense” (Gitlin, 1979). What is a concern to the present study is the “common
sense” view applied to interpretations of the deaths of people of color at the hands of law
enforcement and security personnel.
Van Gorp (2007) tells us that frames limit the potential explanations an audience will use to
evaluate an event. Thus, it is important to understand what frames have been used to understand
race and law enforcement to date. Mills (2017) reported that as far back as 1962, Malcolm X
suggested that the media was used by the police to depict Black people as criminals. We also tend
to see law enforcement and security personnel as warriors or “…soldiers engaged in battle with
the criminal element” (Sinyangwe, 2015, p. 7). The media has generally been believed to reflect
the values of white upper-class society (Alexander J. C., 2003). It makes sense that they have long
supported the “law and order” narrative in which law enforcement and security personnel are held
in high regard as protectors and those who keep their communities safe from crime. News reports
that reinforce this “party line” are highly likely to persuade a large portion of their audience
(Iyengar & Simon, 1993). These two “common sense” views intersect in cases when the officers
report believing that a person was armed when that was not true, which happens more often when
the person was Black.
Cowart, Saunders and Blackstone (2016) performed a content analysis of the Twitter feeds of nine
major media outlets for a month following the shooting of Michael Brown, hypothesizing that
more images would fall under the “law and order” classification than any of the other categories
they had identified (violence, civil liberties, talking heads, or grieving/coming together). They
were unable to support that hypothesis, finding that the “violence” category accounted for more of
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
the images while “law and order” was a close second in the ranking. Interestingly, they also found
that most images separated police and protesters rather than depicting them both in the same image,
suggesting a narrative of opposing forces and reinforcing our earlier discussion of the depiction of
law enforcement personnel ready for conflict, as soldiers or warriors in battle.
Several researchers have investigated differences in national media coverage and local media
coverage of events that can be said to have a racial component. Dill and Wu (2009) examined
front-page newspaper content for the two weeks following Hurricane Katrina searching for the
assignment of blame for suffering and damage. The local papers printed more content about death
and injury, with rescue/relief/evacuation matters coming in second. National newspapers
primarily printed content dealing with the distress suffered by the displaced residents, crime in the
city, and the failure of the government.
Klahm, Papp and Rubino (2016) conducted research similar to the present study, performing a
content analysis of newspaper articles covering police shootings between January 2014 and April
2015. Rather than explicit references to race, they found that the stories were more likely to
mention officer misconduct, less likely to mention criminal history, and less likely to mention the
possession of a weapon when the suspect was Black than when the suspect was White.
Holody, Park and Zhang (2013) conducted a study of the framing of race in coverage of the
Virginia Tech shootings. They found that the national newspapers included in the study were more
likely than the local newspapers to mention the shooter’s race and to do so in a more prominent
way. Similarly, Vakili (2016) performed a content analysis of The New York Times (national) and
The Baltimore Sun (local) and found that although both national and local coverage of the Freddie
Gray situation framed race through negative stereotypes, national articles were more likely to use
race as a frame than local articles were. The national articles suggested that his race was the reason
for his death while local articles attributed his death to his injuries and, on occasion, his own
behavior.
To date, research has looked either at a shortened time frame or individual events to compare
media coverage of these cases. This study expands both the time frame and number of cases to
Koepke, Thomas, & Manning
present a more comprehensive picture of this field. Additionally, this study analyzes the data and
reports on the findings using multiple methods to add robustness to the results.
Method
Research Design
The storehouse of human knowledge and experience is vast, complex, messy, and growing
exponentially. To cope with the information explosion, scholars in many knowledge domains rely
on sophisticated information technologies to search for and retrieve records and publications
pertinent to their research interests. But what is a scholar to do when a search identifies hundreds
of documents, any of which might be vital or irrelevant to his/her work? More and more, scholars
are turning to automated textual (i.e., content) analysis technologies to achieve what they do not
have time to do themselves (Thomas, 2014). This study used Leximancer, an automated textual
analysis technology, to identify and characterize relationships between concepts (e.g., police,
shooting, gun) in hundreds of U.S. newspaper articles, a body of work large and complex enough
to challenge the most systematic, reliable, and unbiased reader.
There are several reasons why a scholar would want an automated system for content analysis of
documents (Smith & Humphreys, 2006). Researchers are subject to influences that they are unable
to report which may lead to subjectivity in data analysis and the interpretation of findings (Nisbett
& Wilson, 1977). Limiting researcher subjectivity often involves extensive investments of time
and money to address interrater reliability and other sources of bias. One goal of automated content
analysis is to reduce this cost and to allow more rapid and frequent analysis and reanalysis of text.
A related goal is to facilitate the analysis of massive document sets and to do so unfettered by a
priori assumptions or theoretical frameworks used by the researcher, consciously or unconsciously,
as a scaffold for the identification of concepts and themes in the data (Zimitat, 2006). Since textual
analysis technologies operate directly on words (as well as other symbols), a rationale for inducing
relationships between words is needed. Beeferman, Berger, & Lafferty (1997) observed that words
tend to correlate with other words over a certain range within the text stream. Indeed, a word may
be defined by its context in usage (Leydesdorff & Hellsten, 2006). For instance, few North
Americans would have trouble completing the sentence, “A breakfast food of lightly fried batter
disks served with butter and syrup is called a … (pancake).”
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
This study asked the research question, “Do local and national newspaper articles frame reports of
deaths of unarmed people of color at the hands of law enforcement or security personnel using
similar terminology and concepts?” In addressing this question, a research design was developed
that went beyond counting concept occurrences as isolated events to characterizing the strength of
co-occurrence relationships between concepts and their respective media sources. Data analysis
was conducted using three approaches (tabular, graphical, and statistical), each of which provides
a different perspective of the relationships between the discovered concepts in the local and
national print media.
In tabular format, co-occurring concepts are presented as word lists in table columns (see Table
1). Each column lists the most frequently occurring concepts appearing in articles from the
indicated sources. In an informal manner, the lists may be compared and concepts common to both
media sources marked. The scope of this analysis is limited to an intuitive awareness of the extent
to which the lists resemble one another.
Table 1
Concepts Associated with Different Sources
Source 1 Source 2
concept1 concept21
concept2 concept22
concept3 concept3
concept4 concept23
concept5 concept24
concept6
…
concept1
…
In graphical format, co-occurring concepts are represented in network spanning trees, or concept
maps (see Figure 1). In these maps, concepts, individual files, and file folders appear as circular
nodes. Concepts positioned closely to one another co-occur frequently. Concepts more widely
separated co-occur less frequently. Segments joining nodes identify the most likely co-occurrences
associated with those concepts. The same principles apply to the positioning of the file source
folders, FOLDR1_localword and FOLDER1_natlword. The scope of this analysis, while still
intuitive in interpretation, employs both Bayesian and machine-learning algorithms to depict
associations between concepts and sources. In this study, graphical representations like Figure 1
Koepke, Thomas, & Manning
were used to identify concepts close to the source folders and, therefore, characteristic of the
articles in those sources and those which were “in the middle” and, therefore, less indicative or
diagnostic, so to speak.
Figure 1 Network Spanning Tree / Concept Map
The Chi-square test of association compared the relative frequencies of occurrence of several
concept clusters. According to Lowry (2018),
…when observed items are sorted according to two or more separate dimensions of
classification concurrently, they are said to be cross-categorized. The most efficient way
to represent cross-categorized frequency data is with what is known as a contingency table.
When chi-square procedures are applied to a contingency table of this sort, it is typically
with the aim of determining whether the two categorical variables are associated; hence the
name of this version, the chi-square test of association.
In this study, the Chi-square test of association was used to determine the probability that both
the local and national data were drawn from the same population of articles.
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
Sample
The authors compiled a list of people of color who lost their lives during the period 1999–2017
following interactions with law enforcement or security personnel. This was a difficult task
because there is no comprehensive list of such victims in the US. The authors relied on information
found in a variety of sources. These included the online databases Fatal Encounters and Operation
Ghetto Storm, the Hip Hop and Politics website, The Root website, victims listed by McCarthy
(2015a) and Alexander (2016), and personal knowledge gained through traditional media reports
and social media discussions. Lists of the victims, local and national newspapers, and articles
referenced in this study may be found at https://mathedcom.wordpress.com/shootings/ in the
Appendices. Appendix A identifies the victims.
Given the list of victims, the authors selected newspapers recognized as local media. Selected
national newspapers were each associated with major metropolitan areas. To achieve a national
sample, the authors partitioned the US into eight distinct geographic areas and selected the most
recognizable newspapers from each. Consequently, the northwestern and southwestern coastal
US, the northeastern and southeastern coastal US, two north-central US regions, and two south-
central US regions are represented. The list of newspapers accessed may be found in Appendix B
on the website referenced above.
Articles were then selected for analysis. While we are aware of the function of repetition in the
success of making meaning and shaping opinions, we believe that the initial presentation of a story
sets the stage for how it will be interpreted. Thus, we chose to focus solely on the first reporting
of each case in each newspaper. Each victim’s name was entered into the search engine of the
website for each of the eight national newspapers and the local newspaper associated with the case.
The article with the earliest date of publication was chosen for inclusion in this study. A coding
system was developed to identify the victim and the newspaper, and each article was saved as an
Adobe Portable Document Format (.pdf) file using the appropriate code. With a list of 97 victims,
we felt that selecting a smaller sample would not necessarily provide representative results and
decided to include all the pertinent cases for analysis. Thus, there was a potential population size
of 834 articles. However, not all newspapers reported on all cases as listed. Therefore, the number
Koepke, Thomas, & Manning
of articles reviewed was 350, with 85 local articles and 265 national articles. The list of articles
included in the data analysis can be found in Appendix C on the website referenced previously.
To compare the content of national articles with local articles, each corresponding file was saved
to a comprehensive group folder. In other words, all 85 individual local files were saved to a
single “local” folder and all 265 individual national files were saved to a single “national” folder,
as the researchers saw no need to identify individual articles for this project. These folders were
then uploaded to Leximancer for analysis. To further strengthen the qualitative content analysis,
the quantitative Chi-square test of association was also performed on the data.
Data Analysis
Using Bayesian statistical methods, Leximancer automatically extracts a dictionary of terms from
source documents, discovers concepts, and constructs a thesaurus of terms associated with each
concept using Boolean algorithms. Concepts identified in this manner are unbiased, robust
statistical artifacts and are depicted graphically in Leximancer as concept spanning trees. In these
trees, concepts appear as circular nodes, frequent co-occurrences appear as segments, and concept
nodes positioned near to one another co-occur more frequently than more widely separated
concepts. Document files and/or folders are positioned in the trees using similar principles to
facilitate identification and interpretation of relationships between concepts and documents.
Leximancer treats words as collections of symbols and analyzes the occurrence and co-occurrence
of words across the entire set of documents submitted for analysis. As a metaphor of this process,
imagine Leximancer concatenating a set of documents and treating their content as a single string,
while maintaining the original structure of each document in case it is needed for further analysis.
Leximancer begins by eliminating parts of speech (e.g., a, an, it, the, those, is, was, were, etc.) that,
while necessary for sentence structure and meaning, are unrelated to the co-occurrence of other
words. It counts the number of occurrences of each word and builds a thesaurus from the “concept
seeds”, or the words that occur most often. It then looks for words that co-occur, either in the same
sentence or adjacent sentences. When a threshold of frequency is found, the program creates a
concept. Metaphorically, concepts in Leximancer are collections of words that generally travel
together throughout the text. Presented with a list of discovered concepts, the researcher may add
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
(e.g., acronyms), delete (e.g., Figure, Table, pp., etc.), or merge concepts (e.g., teacher/teachers,
instructor/instructors), thereby introducing a certain level of informed bias.
Thomas (2014) cites several studies in which Leximancer provided the analysis of textual data.
These studies looked at the progress of design science research in academic conference
proceedings papers, the evolution of international business paper content and the current state of
research on data quality. Thomas has conducted his own research using Leximancer. In one study,
he analyzed papers from the annual meetings of the Society for Information Technology in
Education to determine the emergence of themes and concepts (Thomas, 2014). In another, he
analyzed journal articles and proceedings papers on the use of technology in math and science
distance learning courses to determine if those fields are using the same language, so to speak
(Thomas, 2018). Thomas, Handy and Stols (2017) also used Leximancer to analyze scholarly
articles, reports, and publications dealing with mathematics education in South Africa to determine
the various voices and communities of thought discussing it. It is this type of work that provides
a methodological foundation for the present study.
Findings
Tabular Findings
Leximancer reports identify discovered concepts using both tabular and graphical representations.
Table 2 presents a complete list of the 63 discovered word-like (as opposed to name-like) concepts
in an ordered (by count) list. In this table,
concepts are collections of words that travel together in the text;
count denotes the total number of content blocks containing the given concept; and
relevance denotes the percentage frequency of text segments which are coded with that
concept, relative to the frequency of the most frequent concept in the list.
The relatively frequent use of the concepts police, officers, shooting, shot, and man suggests that
newspaper reports emphasize actors and actions more than complex concepts like community and
justice.
Koepke, Thomas, & Manning
Table 2
Ranked Concepts: All Newspapers
Word-Like Count Relevance
Word-Like Count Relevance
Police 3028 33%
Officers 2850 31%
Shooting 1453 16%
Shot 1174 13%
Man 860 09%
Fired 698 08%
Charges 667 07%
investigation 665 07%
Killed 662 07%
Gun 659 07%
Death 658 07%
Black 653 07%
Family 581 06%
Video 573 06%
Time 474 05%
Report 472 05%
Case 467 05%
Car 447 05%
People 419 05%
Fatal 367 04%
Told 364 04%
Released 357 04%
Force 355 04%
Called 348 04%
Unarmed 342 04%
City 338 04%
Died 298 03%
Involved 273 03%
Tried 273 03%
White 272 03%
Jury 270 03%
Day 268 03%
Attorney 268 03%
community 262 03%
department 257 03%
Law 255 03%
Down 238 03%
Incident 236 03%
Officials 231 03%
News 231 03%
Men 224 02%
Murder 219 02%
Statement 215 02%
Outside 212 02%
Happened 212 02%
Scene 211 02%
Civil 206 02%
Night 193 02%
Including 186 02%
Home 186 02%
Members 184 02%
Woman 174 02%
Son 174 02%
Several 171 02%
Public 164 02%
Heard 158 02%
Decision 149 02%
Mother 148 02%
Justice 142 02%
Head 140 02%
Saying 136 02%
Life 134 01%
Body 133 01%
Tables 3 and 4 offer insights into the differential uses of concepts in local and national reports. In
Table 3, the 20 most likely concepts to occur in national and local text blocks are presented. Note
that only 8 (noted in red) of the 20 concepts appear in both media formats.
Table 3
Twenty Concepts likely to occur in National and Local Reports: 8 duplicates (Red)
National Local
black Fatal
justice Scene
video Involved
white Unarmed
civil investigation
unarmed Murder
jury Video
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
city Black
life Day
decision White
community Outside
department Man
law Charges
including Released
gun Death
force Night
told Police
men City
day Officials
death Community
Table 4 lists the 20 concepts most likely to co-occur with the concept police. Note that there are
no duplicates between media formats. This strongly suggests that reports framed by national and
local print media employ different concepts to describe the deaths of unarmed people of color by
police.
Table 4
Twenty concepts most likely to co-occur with the concept police
National Local
black Scene
justice Woman
video Murder
white Night
civil Home
unarmed Several
jury Heard
city Incident
life Head
decision Officials
community Members
department investigation
law Involved
including Called
gun Time
force Mother
told Tried
men Car
day Died
death Shooting
Graphical Findings
The concept map seen in Figure 2 collapses the 350 files used in the study into national and local
file folders to avoid burying the 63 discovered concepts in a presentation of the article titles. Each
folder icon represents the aggregated content of its sample documents. The most striking feature
Koepke, Thomas, & Manning
of Figure 2 is the diametrical positioning of the national and local file folders. In this
representation, the proximity of a given concept (e.g., scene) relative to a given folder (e.g., local)
indicates frequent occurrence of the concept in the documents of the local folder. Conversely, the
separation of a given concept (e.g., scene) and a given folder (e.g., national) indicates infrequent
occurrence of the concept scene in the documents of the national folder. An inescapable
impression is that local and national reports differ systematically with respect to the concepts they
use to frame their stories.
Figure 2. Concept Map
In Figure 3, the concept map is spanned by a set of colored circles denoting sets of co-occurring
concepts called themes. Each theme is named according to its most prominent concept: police,
report, family, case, son. Unlike concepts, which are robust statistical artifacts, themes are
convenient organizational representations generated by Leximancer with some input from the
researcher. Like “a view from 10,000 feet”, themes provide a setting for intuitive generalization.
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
Figure 3. Concept Map with Themes
Statistical Findings
To achieve a higher-level perspective, frequency data related to the occurrence of the concepts
police, black, white, shooting, shot, fired, gun, death, killed, fatal, murder, justice, officials, jury,
and attorney were aggregated into five clusters as seen in Table 5: police, race, shooting, death,
and justice.
Koepke, Thomas, & Manning
Table 5
Frequencies of Occurrence of Key Concept Clusters
Police Race Shooting Death Justice
Police Black Shooting Death Justice
White Shot Killed Officials
Fired Fatal Jury
Gun Murder Attorney
National 3768 827 3007 1427 720
Local 1019 233 355 527 147
The null and alternative hypotheses used in the chi-square test of association were
H0: Local and national print media employ similar terminology and concepts to describe and
characterize incidents involving the deaths of unarmed people of color and subsequent legal
proceedings; and
HA: Local and national print media employ different terminology and concepts to describe and
characterize incidents involving the deaths of unarmed people of color and subsequent legal
proceedings.
The data presented in Table 5 was entered into the online interface of Vassarstats (Lowry, 2018)
and the analysis reported as seen in Table 6.
Table 6
Vassarstats Chi-square Test of Association 1
Police Race Shooting Death Justice Totals
National 3768 827 3007 1427 720 9749
Local 1019 233 355 527 147 2281
Totals 4787 1060 3362 1954 867 12030
Specifying α = 0.01, df = 4, and X2critical = 13.28, a test statistic of X
2
obtained = 261.44 was returned
using the Vassarstats online statistical analysis suite (http://vassarstats.net/newcs.html). An
associated p-value 0.0001 was obtained. Accordingly, the null hypothesis was rejected in favor
of the alternative hypothesis: Local and national print media employ different terminology and
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
concepts to describe and characterize incidents involving the deaths of unarmed people of color
and subsequent legal proceedings.
Table 7
Vassarstats Chi-square Test of Association 2
Police Race Shooting Death Justice
National -2.9% -3.7% +10.4% -9.9% +2.5%
Local +12.3% +15.9% -44.3% +42.2% -10.6%
Table 7 identifies the degree to which observed cell frequencies differ from the values needed to
support the null hypothesis. For instance, the observed frequency of concepts in the local print
media shooting cluster is 44.3% too small relative to the national print media shooting cluster
under the null hypothesis. Further examination suggests that the local newspapers say less about
the shooting itself and more about the death of the victim than the national newspapers.
Limitations
There are several limitations that potentially impact the findings of this study. First, the lack of
reliable data regarding the number of people in the United States who are shot by law enforcement
personnel each year is notable (Sinyangwe, 2015; McCarthy, 2015a; McCarthy, 2015b; Barry &
Jones, 2014; Lowery, 2014). This not only applies to victims who were unarmed, but also to all
people who were victims of these shootings. Although attempts have been made to collect data
on “arrest-related deaths” (McCarthy, 2015b, p. 2), they have not been successful to date. The
average person in the United States is more likely to learn about these events from either the
traditional media or social media than he or she is to hear it from law enforcement (McCarthy,
2015a). While such reporting may be done, there are a variety of city, county, and/or state
requirements that do not lead to a comprehensive national picture of this issue.
Without a reliable count of officer-involved shootings, it is difficult to compile a list of the deaths
involving unarmed victims and, further, the racial identity of those victims. This also leaves out
the deaths of people who were not shot by law enforcement personnel but nevertheless died while
in their care. Further, it does not address deaths attributed to security personnel who are not
Koepke, Thomas, & Manning
technically members of the law enforcement community. For purposes of this study, the authors
attempted to create a list of victims that was as accurate as possible. In addition to the steps
outlined in the discussion of methodology, the authors reached out to the National Association for
the Advancement of Colored People (NAACP), the Black Lives Matter movement, the Marshall
Project, and the Southern Poverty Law Center as potential sources for a reliable list of victims.
None of those organizations was able to assist with such a list. Without the certainty of a complete
accounting of victims, it is possible that some newspaper coverage that could have altered the
outcome of the data analysis was missed. However, the potential for missing data should not deter
scholars from a study of this type that can provide a description of what can be seen in the known
data.
A second limitation is the number of victims who did not receive coverage in either a local or a
national newspaper. Eight of the ninety-seven victims received no coverage at all in the
newspapers chosen for the study. The other stories did not necessarily receive equal coverage in
all the newspapers under review, either in quantity or quality. Both of these could have led to
biased outcomes in the data analysis. While the lack of coverage of some of these cases could
potentially open a dialog on the lack of concern for the death of an unarmed person of color, the
authors do not feel that such a gap in information negates the descriptive findings of this study.
Finally, a number of the newspaper stories were attributed to The Associated Press or other wire
services rather than original works written by staff reporters. While this has the possibility to skew
the results, it should also be noted that the choice of which, if any, articles to include regarding
each victim is made by the individual newspapers. Thus, the choice of how to present the
information to the audience remains a local one.
Conclusions
As the reader can see from both the graphical and statistical analysis of this data, the authors have
demonstrated that there is a marked difference between local coverage of these deaths and national
coverage of the same stories. This further suggests that there may be a difference in audience
understanding of these events due, at least in part, to the difference in coverage. If one agrees that
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
public opinion has an effect on the actions of others, such as law enforcement and security
personnel, then one must understand how those opinions are formed.
There is much to be learned about race relations in the United States, as well as public sentiment
regarding law enforcement and security personnel, by analyzing newspaper articles. The authors
are planning additional research on newspaper coverage of the deaths of unarmed people of color
that will address the sex of the victim, the coverage that exists before and after an incident happens
locally, a comparison of local coverage when the law enforcement member responsible is indicted
versus not indicted, and a comparison of coverage by geographical region in the US. The authors
would like to invite others to join them on this journey. Unfortunately, it appears that there is no
end in sight to the names of victims to be included in future research. At least eight unarmed people
of color have been killed by law enforcement personnel in the first six months of 2018. The names
Danny Ray Thomas, Decynthia Clements, Stephon Clark, Saheed Vassell, Diante Yarber, Claudia
Gomez Gonzalez, Antwon Rose, Jr., and Anthony Marcel Green have already been added to the
list from which these researchers work.
References
Alexander, B. (2016). Introduction: "Hands up! Don't shoot!": Policing race in America. Cultural
Studies - Critical Methodologies, 16(3), 239-244.
Alexander, J.C. (2003). The sociology of news. New York: W.W. Norton & Company.
Barry, R., & Jones, C. (2014, December 3). Hundreds of police killings are uncounted in federal
stats. Retrieved from Wall Street Journal: https://www.wsj.com/articles/hundreds-of-
police-killings-are-uncounted-in-federal-stats
Beeferman, D., Berger, A., & Lafferty, J. (1997). A model of lexical attraction and repulsion. In
Proceedings of the eighth conference on European chapter of the Association for
Computational Linguistics, pp. 373-380.
Cowart, H., Saunders, L., & Blackstone, G. (2016). Picture a protest: Analyzing media images
tweeted from Ferguson. Social Media + Society (Oct-Dec), 1-9.
Dill, R., & Wu, H. (2009). Coverage of Katrina in local, regional, national newspapers. Newspaper
Research Journal, 30(1), 6-20.
Koepke, Thomas, & Manning
Ehrlich, M. (2005). Shattered glass, movies, and the free press myth. Journal of Communication
Inquiry, 29(2), 103-118.
Entman, R. (2005). Media and democracy without party competition. In J. Curran, & M. Gurevitch
(Eds.), Mass media and society (4th ed., pp. 251-270). London: Oxford University Press.
Gitlin, T. (1979). Prime time ideology: The hegemonic process in television. Social Problems,
26(3), 251-266.
Gitlin, T. (2003). The whole world is watching: Mass media in the making and unmaking of the
New Left. Berkeley: University of California Press.
Holody, K., Park, S., & Zhang, X. (2013). Racialization of the Virginia Tech shootings. Journalism
Studies, 14(4), 568-586. doi:10.1080/1461670X.2012.726499
Hunt, E. (2016, December 17). What is fake news? How to spot it and what you can do to stop
it. Retrieved May 4, 2018 from The Guardian:
https://www.theguardian.com/media/2016/dec/18/what-is-fake-news-pizzagate
Iyengar, S., & Simon, A. (1993). News coverage of the Gulf crisis and public opinion.
Communication Research, 20(3), 365-383.
Jacobsen Koepke, D. (2009). Do you know what it means? A frame analysis of the New Orleans
Times-Picayune's articles for the year following Hurricane Katrina. (Doctoral dissertation).
Retrieved from https://search.proquest.com. doi:3371265
Klahm IV, C., Papp, J., & Rubino, L. (2016). Police shootings in black and white: Exploring
newspaper coverages of officer-involved shootings. Sociology of Crime, Law and
Deviance, 21, 197-217.
Lee, M., & Solomon, N. (1990). Unreliable sources: A guide to detecting bias in news media. New
York: Carol Publishing Group.
Leydesdorff, L. & Hellsten, L. (2006). Measuring the meaning of words in contexts: an automated
analysis of controversies about 'monarch butterflies', 'frankenfoods', and 'stem cells'.
Scientometrics, 67(2): 231-258.
Lowery, W. (2014, September 8). How many police shootings a year? No one knows. Retrieved
from Washington Post: https://www.washingtonpost.com/news/post-
nation/wp/2014/09/08/how-many-police-shootings-a-year-no-one-knows
Lowry, R. (2018). VassarStats: Website for statistical computation. Retrieved
fromwww.vassarstats.net.
Research in Social Sciences and Technology (RESSAT) 2019: 4 (1), 30-50
McCarthy, T. (2015a, March 21). The counted: Inside the search for the real number of police
killings in the US. Retrieved from The Guardian: https://www.theguardian.com/us-
news/2015/mar/21/police-killings-us-government-statistics
McCarthy, T. (2015b, March 18). The uncounted: Why the US can't keep track of people killed by
police. Retrieved from The Guardian: https://www.theguardian.com/us-
news/2015/mar/18/police-killings-government-data-count
Meyers, M. (2004). Crack mothers in the news: A narrative of paternalistic racism. Journal of
Communication Inquiry, 28(3), 194-216.
Mills, C. (2017). Framing Ferguson: Fox News and the construction of US racism. Race & Class,
58(4), 39-56.
Nisbett, R.E. & Wilson, T.D. (1977). Telling more than we can know: verbal reports on mental
processes, Psychological Review, 84(3): 231-259.
Pipkins, M. (2017). "I feared for my life": Law enforcement's appeal to murderous empathy. Race
and Justice. doi:10.1177/2153368717697103
Sinyangwe, S. (2015, May 26). Why do US police keep killing unarmed black men? Retrieved
from BBC: http://www.bbc.com/news/world-us-canada-32740523
Smith, A. & Humphreys, M. (2006). Evaluation of unsupervised semantic mapping of natural
language with Leximancer concept mapping, Behavior Research Methods, 38(2): 262-279.
Thomas, D. (2014). Searching for significance in unstructured data: Text mining with Leximancer.
European Education Research Journal, 13(2), 235-256.
Thomas, D. (2018). Technology in mathematics and science distance education: Automated
textual analysis of articles and proceedings papers using Leximancer. The Eurasia
Proceedings of Educational & Social Sciences, 9, 28-35.
Thomas, D., Handy, H., & Stols, G. (2017). Mathematics education in South Africa: Many
perspectives, many voices. In M. Shelley, & M. Pehlivan (Eds.), Education Research
Highlights in Mathematics, Science and Technology 2017 (pp. 32-41). Ames, IA: ISRES
Publishing.
Vakili, A. (2016). Framing of race by local and national newspapers in the Baltimore unrest: An
analysis of The New York Times and The Baltimore Sun's coverage of Freddie Gray's race.
(Master's thesis). Retrieved from https://search.proquest.com. doi:10119550
Koepke, Thomas, & Manning
VanGorp, B. (2007). The constructionist approach to framing: Bringing culture back in. Journal
of Communication, 57(1), 60-78.
Zimitat, C. (2006). A lexical analysis of 1995, 2000 and 2005 Ascilite conference papers.
Proceedings of the 23rd Annual Ascilite Conference, pp. 947-951.