On August 11th, 2017, a group of White Nationalists marched through the University of Virginia’s campus while chanting Nazi and white supremacist slogans. Shortly after, videos of the march were spreading rapidly around the internet. The rally, called “Unite the Right Rally” (Charlottesville Rally hereafter), was organized to protest the removal of the Robert E. Lee Statue in Charlottesville, Virginia. Several alt-right groups, including the KKK, neo-Nazi, and White Nationalists, were involved in the rally. As the protest heated, James Alex Fields Jr., a self-identified white supremacist, drove his car into a crowd of counter-protesters, resulting in 1 death and 19 injured (Heim 2017).

As a result of the tragic death of Heather Heyer, the Charlottesville Rally was well known both in the U S and globally. However, studies have not been conclusive on the actual effect of the movement. On the one hand, the rally exposed alt-rights’ white supremacist agenda and ambitions, which lead to denied access by major network providers and legal consequences (Atkinson 2018; Menn and Ingram 2017; The Daily Progress 2019). On the other hand, the Charlottesville Rally presented a strong visual representation of the power of white supremacy within the US, as well as the oppositional strength to the alt-right ideology. A group of counter-protesters, mainly formed by the University of Virginia students, were on the front line at the beginning of the rally (Heim 2017).

Behind the scene, media played a vital role for both sides. Several days before the Charlottesville Rally, organizers of the rally posted a statement on an Alt-right website and defined it as a “turning point” where the alt-right would “do something IRL (in real life)” (Law 2017). Leading up to the rally, this message played a crucial role in calling for actions from “all White advocates” to “defend White heritage” and protect the White race, its history, and “way of life” (Law 2017). Media coverage was utilized to convey messages as well as to shape the way the public perceive the rally and form political opinions (Gamson 1992; Nelson and Kinder 1996).

Through the lens of media framing (Entman 1993; Scheufele 1999), this study explores how major media outlets, particularly CNN and Fox News, presented the Charlottesville rally. Specifically, I conducted a computer-assisted content analysis of one week of Charlottesville Rally-related online news articles from the two networks in order to identify and discuss the media’s most salient thematic frames and sentiments.

The critical race perspective also informs this study. This perspective underscores the media contributions to racial conflict through the articulation of problematic representations and framing of race and racial relations in American society (Bonilla-Silva 2006; Delgado and Stefancic 2012). The evidence from the content analysis demonstrates the efforts made by the media to uphold color-blind ideology, underplay racial conflict, and subtly normalize white supremacy.

This paper addresses the following research questions:

RQ1: How did major media outlets in the US frame the Charlottesville Rally in the online articles?

RQ2: Did the framing strategies promote color-blind ideology in reporting the Charlottesville Rally?

Literature Review

Media Framing

Framing refers to the process of highlighting some parts of information about a particular event in a way that influences how audiences receive, process, and differentiate messages are conveyed through this process (Entman 1993; Scheufele 1999). As frames construct social reality, both media and individuals can participate in the framing process (Scheufele 1999). Gamson and Modigliani defined framing as “a central organizing idea or storyline that provides meaning to an unfolding strip of everting, weaving a connection among them. The frame suggests what the controversy is about, the essence of the issue” (1987, p. 143). Similarly, Entman suggests that “to frame is to select some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation” (1993, p. 52). Under the constructionist model, frames represent a set of ideas that interpret, define, and give meaning to a particular social and cultural phenomenon (Goffman 1986).

While the audience is seemingly interpreting the message in their own ways, the media is paving a certain pathway for the audience, using the frames as the bricks (Scheufele 1999). The framing and presentation of events can systematically affect how the audience understands these events. Framing issues in distinct ways may shape public opinion on certain issues (Gamson 1992; Nelson and Kinder 1996). Critical scholars argue that media framing plays a key role in promoting values that allocate power in US society, such as white privilege (Budd et al. 1999). Similarly, Entman (2007) suggests that media in the US is constantly in favor of conservative elites, whose political ideology confines the public discourse.

Constant exposure to selected information on certain social groups can also cause people to adjust their perception, judgments, and behaviors towards those groups, especially toward racial groups (Arendt 2013; Foreman et al. 2016). With selective associations, the stereotypes in the audience’s minds will be created and reinforced (Nelson et al. 1997). This effect is also strengthened by the frequent access to the media (Foreman et al. 2016).

Media outlets may also enjoy power over leading the public course toward a favorable direction where they “declare the underlying causes and likely consequences of a problem and establish criteria for evaluating potential remedies for the problem” (Nelson et al. 1997, pp. 567–568). For example, a study found that news pieces featuring black crime suspects are more likely to quote a pro-prosecution slant compared to their White counterparts (Entman 1992). As such, these news stories emphasize the criminality of certain racial groups, while downplaying others, which in turn attributes responsibility for crime only to some racial groups.

Seeking a more powerful discourse, social movements often find alliance with the media. Social movements often rely on the media to communicate their goals to the audience. At the same time, the media also shows the need to gather “stories” from social events and movements. Nevertheless, this relationship is not merely symbiotic. It often leads to an unequal power dynamic such that social movement organizations lack the means to conduct media campaigns and consequently rely heavily on what resources they can get (Baylor 1996; Gamson and Modigliani 1987). Therefore, the media enjoys an influential role in directing the results of social movements and social events.

Framing the Color-Blind Ideology

As American society transitioned from Jim Crow to the post-civil rights era, expressing racial bias and racism became less socially acceptable. However, such a transition does not merely mark the end of racism (Bonilla-Silva 2015). Studies have shown that racial inequalities persist in every aspect of social life, such as labor market inequality (Drakulich 2015), residential segregation (Bonilla-Silva and Embrick 2007), and health disparities (Bryant-Davis et al. 2009; Gee et al. 2009; Kim et al. 2010). Color-blind racism, as Bonilla-Silva puts it, functions as a racial ideology in justifying the existence of race-based social realties (Bonilla-Silva 2006). By avoiding talking about race and racism, Americans, especially White Americans, find ways to explain, normalize, and even deny the existence of the systemic racism beneath it.

Notably, the color-blind ideology does not only manifest itself at the individual level; instead, it is embedded within the social structure and utilized by social institutions. The media has played a critical role in conveying the color-blind ideology. One of the common strategies involves racial stereotyping. For example, the media often depicts African Americans as criminals (Drew 2011; Entman 1992; Foreman et al. 2016; Nishikawa et al. 2009). Not only does such a strategy strengthen hegemonic racial discourse through oversimplified representations, but it also disguises the structural challenges racial minorities are facing. Seemingly contrarily, the media sometimes takes the stand of African Americans being fully assimilated, proving the absence of racial bias (Campbell et al. 2012). By employing these two strategies, the media is able to direct the public discourse of racially-related affairs toward a non-racial explanation. That is, it allows the ignorance of structural racism while practicing racial discrimination without recognition.

The availability of multiple channels for distributing color-blind ideology also magnifies the effects of media framing. News reporting, especially political reporting, is a case widely examined by researchers (Caliendo and Mcllwain 2016; Squires and Jackson 2010). Studies have shown that racial minority politicians are often racialized in the news coverage while their White counterparts remain unmarked (Caliendo and Mcllwain 2016; Squires and Jackson 2010). Other researchers have revealed that commercials frequently use the racialized image of minorities to reinforce the dominant position of White male (M. Kim and Chung 2005). Numerous television and films adopted the strategy of hiring diverse casts as an abstract liberal solution to racism in the media (Bonilla-Silva and Ashe 2014). Sports media draws on a discourse that perpetuates racial stereotypes when talking about sports, too (van Sterkenburg et al. 2019, 2010). Such a multi-dimensional information network makes it almost inevitable for the audience to absorb color-blind ideology in one way or another.

The focus of media effect has shifted with the rise of the digital era. Studies on social media have noted that the racial divide online is as prominent as offline (Cisneros and Nakayama 2015; Daniels 2013; Sharma 2013; Sommier et al. 2019). For instance, Cisneros and Nakayama (2015) examined the online response on Twitter to the election of 2014 Miss America. Their study, revealed the marked difference between online and offline narratives. That is, while the offline world celebrating the first Indian American Miss America as a sign of a progressive post-racial society, social media platforms enacted blunt racist expressions and reinforces white supremacy. Yet, the internet is more than a mere reflection of reality. Critical race researchers have recognized that racial relation is embedded within the technology itself; in turn, it participates in shaping and reinforcing racial relations beyond internet (Daniels 2013; Sharma 2013; Sommier et al. 2019). As such, seeking a comprehensive framework to further examine the internet-racial complex becomes more important than ever.

Framing the Charlottesville Rally

Agreeing upon the nature of the Charlottesville Rally as an alt-right movement (Atkinson 2018; Hartzell 2018; Pei 2017), researchers have examined several aspects of the movement through media framing theories. Scholars have paid attention to how the public discourse has been directed and changed during and after the Charlottesville Rally (Atkinson 2018; Hartzell 2018). While the “alt-right” ideology is no stranger to US society, the effort to promote themselves into mainstream public discourse has never stopped (Atkinson 2018; Hartzell 2018). The Charlottesville Rally helped to reveal the “alt-right’s” strategies that attempt to normalize other political discourses such as white nationalism and populism (Hartzell 2018; Perry 2018). The response from President Donald Trump is also seen as an action of mourning “the cultural erosion of whiteness” (Perry 2018, p. 63).

The most common approach to study framing Charlottesville Rally focuses on social media. For example, adopting a conflict frame, one study found that Twitter was used as the main stage to motivate hostile reactions and promote material violence (Klein 2019). Both main actors—the alt-right and antifascist groups—were found to participate in promoting the “rightness of their actions” (Klein 2019, p. 315). The author further argued that both sides framed their counterparts with subtexts providing foundations for extremism and violence (Klein 2019). Similarly, another study examined the structure of the online conversation of the Charlottesville Rally (Tien et al. 2019). Using Twitter data, researchers concluded that the Twitter network is strongly associated with the audience’s political orientation and media followership (Tien et al. 2019). Although the authors in the second study found that the media followership with Fox News was strongly associated with the right-leaned audience, both studies focused on social media and captured how individuals participate in the public discourse.

Contribution of This Study

While most studies on the Charlottesville Rally address violence and white supremacy at least to some extent, few have linked the event to color-blind racism in the US (Hartzell 2018). White supremacy and color-blind racism go hand in hand with each other (Simpson 2008). This paper attempts to contribute to the existing literature in the following ways. First, this paper re-emphasizes the critical role that the media has played in shaping our perceptions of the Charlottesville Rally through racist representations, which was seemingly neutral but actually perpetuates social inequality, especially racial inequality in the US. Second, this paper attempts to join the new wave of critical study in the era of digitalization by extending the examination of the media effect from traditional forms to online platform, with a shift from social media toward online reports. This approach not only captures the media effects from a relatively understudied platform, but also pushes critical theory to join the conversation in the digital age. Third, this paper reexamines the political leanings that the two major media outlets posit. More importantly, this paper aims to bridge the study of media and color-blind racism, as an attempt to reveal the prevalence of color-blind ideology, as well as echoing the call of a more critical interrogation of media as a social institution. In addition to theoretical contributions, this paper also introduces a new methodology into the predominantly qualitative field of content analysis. The application of unsupervised machine learning algorithms opens up the possibility of quantifying the media content on a large scale, which echoes the unceasing tide of digitalization of social life.

To further bridge media framing theories and critical race studies, as well as fill in the literature gaps, this study addresses the two main research questions by testing the following hypotheses. To answer the first question of how major media outlets in the US framed the Charlottesville Rally in online coverage, the first hypothesis examines the nature of Charlottesville Rally in the media and the next two compare two major media outlets with different audiences. To be more specific, topics refer to the general themes whereas the wording preferences also cover the ranking of the keywords under each topic.

Hypothesis 1

Media outlets framed the Charlottesville Rally as a violent event lead by white supremacists.

Hypothesis 2

Catering to the targeted audiences, CNN and Fox News framed the Charlottesville Rally differently by choosing different topics.

Hypothesis 3

CNN and Fox News adopted different wording preferences.

To further explore the sentiments of each topics and compare between CNN and Fox News, I test the following hypothesis:

Hypothesis 4

The topic choices influenced the level of sentiments expressed.

Methods and Data

To answer the research questions, I applied computer-assisted content analysis to explore how major media outlets frame the Charlottesville Rally. While traditional content analysis remains a valuable research tool in social studies, it also faces challenges. For example, traditional content analysis usually requires a high amount of human labor to be put into the coding process, which sets limitations on the scope and size of the content to be examined (Lewis et al. 2012; White and Marsh 2006). Manual coding may also introduce bias into the sample (Lewis et al. 2012; White and Marsh 2006).

Compared to traditional content analysis that solely relies on manual coding, computational methods have several advantages that could overcome some of these obstacles (Lewis et al. 2012). First, with the help of algorithmic measures, it is possible to analyze a broader sample, even making it possible to include the entire data of interest. As the digital form of news content has grown, so has the need for expert labor to analyze this information. While manual coding requires relatively high human labor resources, computational methods are less resource-dependent. Second, topic models are able to identify the hidden structures within the collection of documents with a more accurate prediction than human coding based on probabilistic models (Blei 2012). Moreover, computer-assisted content analysis allows for the utilization of multiple tools to provide additional information that is nearly impossible to uncover using manual coding (Lewis et al. 2012). Lastly, as the big data era arrives, the need for social scientists to apply new technologies to utilize data cost-effectively becomes more pressing.

Topic Modeling

Topic Modeling is a suite of computerized language processing algorithms that “aim to discover and annotate large archives of documents with thematic information” (Blei 2012). Among all the topic modeling techniques, Latent Dirichlet Allocation (LDA) is one of the most common and broadly used topic models across disciplines and topics. In general, LDA captures the common topics in the documents and assigns the document to a certain theme with the highest probability (Blei 2012). In other words, not only does LDA provide thematic allocation to each article in the sample, but it also offers detailed and visualized results that illustrate the distribution of the themes. After repeating the same process with each item in the documents, LDA generates a number of topics with a number of documents under the topic collection. That is, documents in the same collection share the same topic, yet they may exhibit the topic in different proportions.

For the purpose of this paper, LDA is able to discover (1) most relevant topics for each media outlet, (2) ranking of keyword frequencies, and (3) the probability of an article to be assigned under a specific topic.

Sentiment Analysis

Besides themes of the news articles, the latent attitude conveyed through the publication itself also plays a vital role for framing content for the readers (Eshbaugh-Soha 2010; Farnsworth and Lichter 2010; Gentzkow and Shapiro 2010; Young and Soroka 2012). While LDA provides information on topic of choices, it lacks of the ability to detect the attitude orientation behind topics and keywords. As such, the tone of the news articles is identified using the Lexicoder Sentiment Dictionary (LSD). LSD is a comprehensive dictionary that is aimed primarily at news content and other materials (Young and Soroka 2012). Combining effective lexical resources from political science, linguistics, and psychology, LSD scores positive and negative tones of the target documents. It counts the number of positive and negative words based on the dictionary and assigns a sentiment score to each item in the sample. Comprised of three different established dictionaries, LSD includes both positive words and negative words. For example, positive words include “benevolence,” “glory,” and “respect,” whereas negative words include “malevolence,” “chaos,” and “anxiety” (Young and Soroka 2012). It also excludes terms that may not have a clear sentiment, such as “increase” or “decrease” (Young and Soroka 2012). Such an approach discovers the tone and attitude that would otherwise be invisible in the article. Researchers have shown that LSD has a high explanatory power across a number of most popular dictionaries, while being consistent with human coding (Young and Soroka 2012).

The results from the LSD meaningfully supplement to the LDA results by (1) providing sentiment scores for each article and each media, (2) making it possible to link topics and sentiment scores, (3) detecting latent attitude in the article which would be difficult to discover otherwise, and (4) revealing the whole picture of the frames utilized in reporting the Charlottesville Rally.

Data

The media reports in this paper came from two major media outlets—CNN and FOX News, representing primary digital news coverage accessible through their websites. According to the Pew Research Center, the audiences of these two media outlets tend to have different political orientations, with the CNN audience leaning liberal and the FOX News audience leaning conservative (Pew Research Center 2016). I include both news outlets in this study in order to compare strategies to reach different audiences.

While both CNN and FOX News appear in traditional T.V., digital video, and news article formats, this study examined only online news articles. According to a report released by the Pew Research Center in 2018, about a third of Americans in the survey prefer online news, with a 6% increase from 2016 (Mitchell 2018). Moreover, the web became the most popular platform for those who prefer to read their news (Mitchell 2018). As such, this study aims to contribute to how web-based news articles may convey a certain attitude.

News articles for both media outlets were from their websites. I used the keywords “Charlottesville” and “Unite the Right Rally” in the search engine provided by each website. To specify the target time period for the search, I further restricted the timeline from August 11 to 18, 2017, intending to cover the beginning of the Charlottesville Rally and one week after the rally. In order to refine the results, the sample did not include op-eds, daily briefs, T.V. interview transcripts, and timeline reviews. The sample keeps the information about title, report date, author, and report contents. The final sample size was 404 news reports in total, with 234 pieces from CNN and 170 from FOX News.

Findings

Topic Models

As the first step to conduct topic models, I created a list of “bad words” to be removed from the dataset, including common words embedded in online articles (such as “date” and “advertisement”) and prepositions and articles (such as “on,” “the,” and “a”). This step further cleaned the dataset and minimized potential noise from meaningless words. The trimmed sample had 1,229,776 elements in total, with 661,518 elements from CNN and 224,400 from Fox News (Figs. 1, 2).

Fig. 1
figure 1

Wordcloud for the whole sample (N = 404)

Full size image
Fig. 2
figure 2

Words list with frequencies greater than 150 (N = 404)

Full size image

I first examined the word frequencies using the whole sample, including articles from both CNN and Fox News. A word cloud shows that “Trump,” “white supremacist,” and “violence” were at the center of the dataset. Specifically, word counts for the whole sample show that the most used word, “Trump,” appears for 2879 times (0.23%), followed by 2093 counts of “white” (0.17%), 840 counts of “supremacist” (0.07%), and 818 times of “violence” (0.07%). In comparison, the average frequency for the words used in the trimmed sample is 8.9 times. After breaking down the dataset by media sources, both subsamples share the same top four words. The CNN subsample contains 0.34% of “Trump,” 0.22% of “white,” and 0.08% of “supremacist” and “violence.” In contrast, the Fox News subsample includes 0.30% of “Trump,” 0.29% of “white,” 0.13% of “supremacist,” and 0.14% of “violence,” after adjusting by the total number of elements by media source.

The word frequency and word cloud offer a descriptive visualization of the dataset, which confirms the accepted view of the Charlottesville Rally as a violent protest with a highly political nature perpetuating white supremacy in the US (Atkinson 2018; Hartzell 2018; Heim 2017; Klein 2019; Perry 2018; Tien et al. 2019).

Secondly, I performed standard LDA on the whole dataset and the two subsamples. The results provide two parts of information: topics and keywords. Topics are assigned based on the general theme of the article where as keywords under each topic suggests the wording preferences under each theme. The topic models extracted four topics with 2000 iterations of Gibbs sampling. Table 1 presents the results for the whole sample, CNN subsample, and Fox News subsample. After exploring LDA from two-topic to six-topic groupings, the four-topic model provides the most interpretable results with the least word overlap. I then assigned labels to summarize the contents of each topic model. For the whole sample, Topic 1 focuses on political issues, highlighting politicians (e.g., Trump, Obama, Bannon, and Clinton), political parties (e.g., Republican and Democrat), and other politically relevant words (e.g., campaign and left). Topic 2 is about white supremacy, which included both explicit terms such as “white,” “supremacist,” and terms that highly related, such as “neo-Nazi,” “nationalist,” and “bigotry.” Topic 3 features racial conflict by talking about “Black,” “race,” and “racism.” Additionally, the word “war” is relatively prominent. Topic 4 presents the most detailed information about the Charlottesville Rally, including the name of the victim Heather Heyer and neutral terms such as “law” and “protest.” Notably, all topics share some commonality of words across topics.

Table 1 LDA topic assignment with top 10 keywords
Full size table

Comparing the total sample with the two subsamples, the keywords do not vary. The CNN subsample highly resembles the four topics captured in the full sample, whereas the Fox News subsample does not present topics as clearly. Specifically, the keywords LDA captured from the Fox News subsample do not align with those from the total sample. For instance, Topic 1 in this subsample contains the keywords “Trump” and “Republican,” which fits the political theme. At the same time, it also has terms such as “neo-Nazi,” “white,” “supremacist,” and “hatred,” signaling inconsistency for the purpose of topic interpretation. Still, I aligned the four topics under the Fox News with the whole sample by the most similar wording usage under each topic for the purpose of further comparison (Fig. 3).

Fig. 3
figure 3

LSD sentiment analysis distribution by sources (N = 404)

Full size image

Additionally, the positions of each keyword under each topic are meaningful. Under each topic, the display of the keywords follows the pattern from the highest frequency to the lowest. In other words, the more times a keyword appears in under a certain topic, the higher that keyword ranks in the list. For example, “Trump” is the most common word under Topic 1. This position indicates that in all the articles assigned under Topic 1, the term “Trump” is the most repeated.

Pairing the topic assignment and the word positions together, the two subsamples reveal unique wording preferences under the corresponding topics. Topic 3 results suggest that both subsamples focus on the key terms “war,” using words such as “symbol” and “violent” to describe the elements of the event. Beyond the similarities, CNN and Fox News are distinguishable for that the CNN subsample seems to attribute the event to “culture.” In contrast, the Fox News subsample redeems the event as “controversial.” Additionally, the term “patriot” only appears under the Fox News subsample.

Topic models answer the first research question, that is, both media outlets of interest—CNN and Fox News—adopted fairly similar frames to talk about the Charlottesville Rally. The descriptive word frequencies chart and word cloud both provide evidence to support Hypothesis 1; that is, the media outlets frame the Charlottesville Rally as a violent event lead by white supremacists. The results, however, reject Hypothesis 2, which claims that CNN and Fox News framed the Charlottesville Rally differently by emphasizing different topics. Politics, white supremacy, racial conflict, and the car incident are the four most prominent topics that appear in the online news articles during the week after the event happened. While the Fox News subsample has a mix of words spreading across the topics, the general interpretation does not change dramatically. Not only did both media outlets focus on highly similar topics in reporting the event, but the wording preferences are highly similar. Most of the keywords show up in both media outlets. However, there are some unique words in both subsamples observed. CNN subsample highlights “moral” and “culture” whereas Fox News emphasizes on “patriot” in their articles. In addition, even though the four topics seem to be highly similar, the ranking of the top keywords vary by the media sources. For instance, under Topic 3 (i.e., racial conflict), CNN mentions “black” the most whereas Fox News focuses on “violence.” Such differences in the ranking of top keywords are observed for most keywords. These differences support Hypothesis 3, which claims that CNN and Fox News adopted different wording preferences.

Sentiment Analysis

Unlike the results in the topic models, the sentiment carried in the CNN and Fox News subsamples differs significantly. Besides the positive sentiment and negative sentiment scores from LSD, I also generated a measure of the logged sentiment ratio for each article [Logged Sentiment Ratio = Log(Positive sentiment/Negative sentiment)]. The logged sentiment ratio aims to balance out those articles with strong emotions for both positive and negative sentiments. I applied a logarithm to normalize the distribution of the measure. Table 2 provides the descriptive statistics for each news outlet by sentiment. The CNN subsample shows relatively high positive (M = 28.28, SD = 19.27) and negative emotion scores (M = 49.14, SD = 31.55) in comparison to the FOX News subsample (Positive Mean = 16.21, Negative Mean = 31.26). Welch t-tests also confirm statistically significant between-group differences between the two subsamples for positive sentiments [t(403.87) = 7.24, p = 0.000], negative sentiments [t(403.77) = 6.52, p = 0.000], and logged sentiment ratio [t(311.86) = 2.86, p = 0.005] (Table 2).

Table 2 Descriptive statistics on negative sentiment, positive sentiment, and logged sentiment ratio (N = 404)
Full size table

Combining the results from both topic models and sentiment analysis, I further explored the relationship between the assigned topics and the sentiment scores. I hypothesized that the topic of each article would influence the expressed sentiments of the article. In other words, reporters tend to show different attitudes when writing news articles on various topics. I employed ANOVA to test for differences in positive sentiments, negative sentiments, and the logged sentiment ratios, treating the topics (i.e., politics, white supremacy, racial conflict, and car incident) as a between-subject variable.

The ANOVA results suggest that there are statistical differences in the usage of positive tones when talking about different topics for the whole sample [F(3,400) = 7.37, p = 0.000], the CNN subsample [F(3,230) = 7.90, p = 0.000], and the Fox News subsample [F(3,166) = 9.65, p = 0.000]. Similarly, the results for the measurement of logged sentiment ratio are also significant for the whole sample [F(3,398) = 8.93, p = 0.000], CNN subsample [F(3,230) = 12.27, p = 0.000], and Fox News subsample [F(3,164) = 11.30, p = 0.000]. Interestingly, only the Fox News subsample presents the significant difference between topics when expressing negative sentiment [F(3,166) = 5.33, p = 0.002] (Table 3).

Table 3 ANOVA of positive sentiment, negative sentiment, and net tone on top topics using CNN and Fox News subsamples
Full size table

To further investigate which topic presents more sentiments, I estimated multiple Ordinary Least Square (OLS) regression models using the topic assignment as predictors and controlling for the news outlets. The results confirm that CNN articles present more sentiments compared to Fox News. Such findings hold for positive sentiment, negative sentiment, and the logged sentiment ratio. OLS results also suggest that, for both positive sentiment and the logged sentiment ratio, compared to the reference topic of car incident, articles about racial conflict and politics carry more sentiment after controlling for media sources. More importantly, the topic related to white supremacy is statistically insignificant in sentiment expression.

The results from the sentiment analysis offer more evidence to answer the research questions. Combining topic models and sentiment analysis, the results provide support for Hypothesis 4, which claimed that topic choices influenced the level of sentiments expressed in the news reports, but the effect did not apply equally for every topic (Table 4).

Table 4 OLS regression models of topic assignments on sentiments (N = 404)
Full size table

Discussion

The main contribution of this paper lies in the exploration of how color-blind racism manifests itself in media framing strategies during the digital era. By adjusting the topic focuses, wording preferences, and sentiment expression, major media outlets participate in directing the public discourse around the Charlottesville Rally.

How did major media outlets in the US frame the Charlottesville Rally? While this overarching research question contains more dimensions than the scope of this research, this paper offers one way to approach it. With the online news articles from two major news outlets during a 7-day period after the Charlottesville Rally, this paper has shown that the most prominent topics do not vary significantly between news sources. In other words, the major news outlets provide similar information on the Charlottesville Rally during the time period of interest. It suggests that when the media outlets present the Charlottesville Rally, the audience-reach is not the focal point. These findings seemingly disagree with existing literature (Pew Research Center 2016). Yet, several factors may have contributed to the invariance.

On the one hand, all the articles retrieved from the websites are around the same event, which limits the number of possible ways to talk about it to some extent. One the other hand, it implies that the audience-reach for CNN and Fox News is not as different as most people thought, especially on the topic of Charlottesville Rally. Notably, most existing literature regarding audience difference examined television programs and social media (Caliendo and Mcllwain 2016; Klein 2019; Squires and Jackson 2010; Tien et al. 2019). As society is fully embracing digitalization, the presumptions of media also encounter challenges and need reexamination.

The almost unanimous wording choices are also noteworthy. The choices of words, phrases, and sentences that go into a story may also lead the public discourse toward certain routes (Nelson et al. 1997). First, both media outlets seem to underplay the role of racists and racism in the event. Rather than using the straight-forward word “racist” and “racism,” media outlets used more convoluted words, such as “nationalist,” “neo-Nazi,” “supremacist,” and “alt-right.” The usage of code-words challenges the public audiences’ perception and knowledge of these abstract concepts, which creates a comfortable space for the public to distance themselves from these labels. This frame strategically offers the mainstream population with a rhetorical construction of the image of the “blameworthy people” in the violent event without necessarily identifying themselves as one of the privileged groups. Ultimately, it serves as the base to uphold the pro-white ideology and white dominance in the color-blind scheme (Hartzell 2018).

Indeed, the ranking of the keywords largely depends on the news sources. Categorizing the event as a “moral” or “culture” debate seems to be a different frame than calling the event actors as “patriots.” To this end, CNN and Fox News may have expressed their different standing points. However, how much these keywords represent the framing strategies remains unclear. Such uncertainty may come from the nature of the LDA as a frequency-based algorithm. The keywords rankings are correlated to the topic assignments. Future research may pursue further examinations on the unique top keywords or by using other topic models.

Under the assumption that the report on the car incident should be the most neutral for its fact-based nature, it is interesting that the topic of white supremacy presents significantly less sentiment compared to racial conflict. The disparity of sentiment level on white supremacy and racial conflict, despite the substantial topic overlap, signals the reluctance of the media outlets to take a side when the topic gets to the core of the pro-white ideology. While emotion and attitudes may leave an impression on the audience’s interpretation of the issue, and ultimately impact the public opinion in a subtle way, too (Iyengar 1990; Nelson et al. 1997; Nelson and Kinder 1996), suppressing the sentiment may also trivialize the topic of concern. The insignificance of the white supremacy topic compared to the car incident suggests that news articles attempted to suppress emotions when talking about this subject.

The fact that the sentiment involvement of white supremacy and the most fact-based topic on car incident further reveals the framing strategy which soft-pedaled the role of white supremacy in the event. Not only does such finding confirm the existence of media framing, but it also reveals media’s powerful role in how to present certain topics. As critical race theorists have argued, racism is no stranger to US society (Delgado and Stefancic 2012). With the emergence of color-blind racism in the post-civil rights era, racism took on a new form and formed a false allure (Bonilla-Silva 2003, 2006; López 2015). It promotes an understanding of race and racism that obscures discrimination against racial minorities and magnifies the ostensible mistreatment of the Whites (López 2015). Along with the development of the notion of color-blindness, it has been utilized strategically so that color-blindness justifies the racial status quo by deemphasizing race itself. The findings that both media outlets adjusted their sentiment expression on the topic of racial conflict and white supremacy present us with a vivid example of how race is defocused in the rhetoric and ultimately in the public discourse. The minimal sentiment expression on the topic of white supremacy is no more than a pragmatic mode of crisis management (Martinot 2010).

Finally, given the evidence presented in this paper, there is a need to examine the effect of media framing and public discourse further. Although the evidence mentioned above reveals the role of the media plays in fueling color-blind racism, the true impact of these elements on the audience remains unknown. The absence of a thorough investigation of the real effect of mass media may compromise the sociological theories that center social institutions at the heart of social inequalities. Future research may benefit from a mixed-methods approach. Indeed, the algorithm applied in this paper is an unsupervised approach. With semi-supervised methods, the accuracy and explaining power would increase dramatically. The traditional content analysis with human coders will also bring in qualitative study on this topic. Moreover, while online news articles are still receiving a considerable share in the news distribution, the digitalization of information does stop here. Other platforms, such as social media, podcasts, as well as live streaming programs, are equally crucial in exploring the true effect of the internet and media on racial relations in the US.

The Charlottesville Rally was not the first alt-right event; neither would it be the last alt-right event. Understanding the media’s role in contemporary society, especially how it influences racial relations and reinforce racial inequalities, would lead to a deeper understanding of social structure as well as direct us toward a just society.