Press bias and policy agendas: an analysis of
the 2016 presidential campaign
Was there biased media reporting on the policy agendas of the presidential candidates?
Student: Hannah Boerstra (10307826) Supervisor: dr. Gijs Schumacher Second reader: dr. Daphne van der Pas
June 22, 2018
Master thesis Political Science Track: International Relations
Abstract
The aim of this thesis is to establish whether there was biased media reporting on the policy agendas of the presidential candidates during the presidential election of 2016. Previous research has shown that candidates create their policy agenda thoughtfully and that the congruence between a candidate’s policy agenda and its reported policy agenda influences the voter’s preference for that specific candidate. However, the American media is tainted by the perception and presence of bias, bias that is able to affect the reports on the candidate’s policy agenda. The content analysis conducted in this thesis compares the campaign messages of the two presidential candidates with their media coverage to determine whether there was biased media reporting present and to what extent partisan bias could be the explanation for the differences in coverage. The results show that the newspapers’ reported policy agendas correlate more with the policy agenda of the candidate they endorsed than with the policy agenda of the opposite candidate. In contrast, the results present a visibility bias in the issue coverage of the newspapers in favor of the Republican candidate, Donald Trump. Hence, the results depict the multifaceted nature of media bias and further research should explore other structural factors influencing media coverage.
Keywords: media bias, presidential election, U.S. politics, policy agenda, newspapers, issue coverage, partisanship
List of Tables
Table 1. Number of Collected Articles per Candidate 20
Table 2. List of Policy Issues Included in the Codebook 22
Table 3. Policy Agenda of Presidential Candidates (%) 27
Table 4. Correlations between Policy Agendas and Reported Policy Agendas 28 Table 5. Campaign Speeches of Presidential Candidates Covered by the Media (%) 31
Table 6. News Articles that Discussed the Policy Issues of the Candidate (%) 32 Table 7. The Average Number of Sentences Discussing Policy within News Articles 33
Table of Contents
Abstract 2
List of Tables 3
1. Introduction 5
1.1 Media Bias in the United States 6
1.2 The Purpose and Construction of Political Campaigns 8
1.3 Media Coverage of Election Campaigns 9
1.4 Research Question - Scientific and Societal Relevance 11
2. Theoretical Framework 13
2.1 A Closer Look into Media Bias 13
2.2 Hypotheses 16
3. Methods 17
3.1 Research Design 17
3.2 Data and Unit of Analysis 18
3.3 Operationalization 21
3.4 Data Analysis 23
3.5. Empirical Limitations of the Research Design 25
4. Results 26
4.1 The Policy Agendas of the Candidates and the Reported Policy Agendas Compared 26 4.1.1 Policy Agendas of the Presidential Candidates Hillary Clinton and Donald Trump 26 4.1.2 Correlations Between the Policy Agendas of the Candidates and the Reported Policy Agendas 28
4.2 Taking Issue Coverage into Account 30
4.2.1 Total Coverage of the Campaign Speeches 30
4.2.2 Issue Coverage of the Campaign Speeches 32
5. Discussion and Conclusion 34
5.1 Findings 34
5.2 Discussion 35
5.3 Suggestions for Further Research 37
5.4 Conclusion 38
6. References 40
1. Introduction
In recent years, the issue of media bias in the United States has occupied the academic dialogue and public opinion. Citizens’ trust in the media is decreasing and research points out how media reports are consistently influenced by multiple factors, from consumer preferences to partisanship (Friedman, 2018; Groeling, 2013; Druckman & Parkin, 2005). Meanwhile, political candidates rely on the media during elections for informing the voters about the candidates’ policy agenda and plans for the country. For generations, the media has been the vital source of information about the political candidates (Farnsworth & Richter, 2010). Rather than functioning as a political bystander and simply transferring information, media often play a more decisive role. Instead, they choose to influence the interpretation and presentation of the campaign events and actively evaluate the performance and character of the candidates (Farnsworth & Richer, 2010; Druckman & Parkin, 2005). This raises the following question: apart from character and performance evaluations, to what extent does the media report on the policy agendas of the presidential candidates and most importantly, to what extent is there biased media reporting on the policy agendas of the candidates [emphasis added]?
This thesis offers a detailed analysis of the way newspapers reported on the policy agendas of the presidential candidates during the presidential election of 2016. This thesis seeks to unravel to what extent the candidates were able to put their policy agenda through the media and whether the policy issues mentioned by each candidate were reported by the media roughly in the proportions that the candidates themselves emphasized them. Consequently, by comparing different media sources, the results of this thesis establish whether there was biased media reporting on the policy agendas of the presidential candidates Hillary Clinton and Donald Trump. By studying the media reports during the presidential elections, this thesis contributes both to the literature on media bias in the United States and the literature on the presidential election of 2016.
This thesis is structured as follows. The first chapter of this thesis offers a review of the literature on media coverage of political institutions, in particular political campaigns. Furthermore, this chapter presents the research question of this study and clarifies the societal and scientific relevance. Chapter two features the theoretical framework regarding media bias and particularly highlights agenda bias, the topic of this thesis. From this theoretical framework, the hypotheses utilized for the analysis are created. The third chapter consists of the methodological framework and illustrates the detailed plan for the quantitative content analysis conducted for this study, clarifying the process of data collection, the operationalization and the data analysis. The fourth chapter presents the results retrieved from the content analysis with the use of statistical analyses. Lastly, the results are discussed and substantiated in chapter five in order to answer the research
question. In addition, the last chapter offers a recommendation for further research along with an explanation of how these results contribute to the literature about media bias.
1.1 Media Bias in the United States
Everyday news stories take millions of American citizens around the world to inform them about the political battlefields that have originated or are being continued by political players in the field of world politics. Accordingly, citizens expect the media to inform them about any injustice or wrong-doing in the political sphere and to give them an overview of what is considered of great significance by different groups in society. For these reasons, Americans pay close attention to the mass media from a young age to learn about current events and value the images and the content that media outlets share with them (Graber & Dunaway, 2017).
During elections, campaign statements and speeches shape voters’ perception of important political problems and inform the voters about the stances of the candidates on various policy issues. Evidently, the media play an important role as voters obtain a large part of this knowledge during elections via the media. The media convert these campaign messages into news stories which are then read, heard and viewed by citizens. While the media has always been deemed influential during political elections, media favorability within the coverage of political campaigns has not always been perceived as similarly influential. Earlier on, social networks and party identification served as the powerful influencers of the people’s vote, instead of the mass media. Moreover, research showed little effects on voting behavior. These factors tempered scholars’ interest to study this subject (Hayes, 2008).
In recent decades, discussions of media bias revitalized scholarly interest in bias and to what extent bias could possibly influence voters’ perceptions of candidates, parties or policy issues. Research attributed evidence of these effects and theories about media bias, media frames and media favorability were constructed. Though effects were typically modest, the findings suggested that news outlets favor different candidates or parties (Entman, 2007; Hayes, 2008; D’Alessio & Allen, 2000).
To understand the nature of media bias, one should know the reasons behind it. In the United States, the debate around media bias centers on the popular notion that the media is liberal and therefore ideologically biased. Ideological bias refers to the world views of journalists, editors and owners taking part of a media organization and how these world views influence the news stories these actors publish (D’Alessio & Allen, 2000). Here, ‘liberal’ is understood as an important American value or belief and the opposite of another important value, ‘conservative’. While the Republican Party is labeled as conservative, Democrats identify themselves as liberal. Therefore,
the American media is considered to be in favor of the Democratic Party. In addition, partisan bias emerges when news media report a distorted portrayal of reality that systematically favors one party over the other (Groeling, 2013).
Public perception took over this idea of liberal bias in the media, but research into ideological and partisan bias in the American media shows no consensus (e.g. Grosecloce & Milyo, 2005; Niven, 2003; Soroka, 2012; Waldman & Devitt, 1998; Grabe & Bucy, 2009; Butler & Schofield, 2010). Because of the lack of comprehensive empirical evidence for media bias, scholars argue how this perception of bias manifested due to Republican politicians who accused the media industry of favoring the Democratic Party (Watts, Domke, Shah & Fan, 1999; Entman, 2007; Patterson and Wolfgang, 1996). Likewise, Watts, Domke, Shah and Fan mention how public perception of liberal bias is the result of constant news self-coverage on the general topic of media bias (Watts, Domke, Shah & Fan, 1999). In contrast, liberal critics argue that the American media is biased in favor of the conservatives, because the American journalists are employed by business people who will likely be conservative and have the opportunity to replace liberal journalists or editors when they want to (D’Alessio & Allen, 2000).
Notwithstanding these contradicting results and viewpoints on ideological bias, an element of partisan bias does characterize American newspapers. The content of newspapers is divided between the editorial pages of a newspaper and the pages devoted to the news. Editors articulate their political beliefs within the opinion pages which are not supposed to influence the overall news coverage. During elections, the editors of American newspapers regularly endorse a specific candidate from a political party within the editorial section. Thus, the editors inform their readers about which political party they believe ought to be in charge over a certain political institution. Various studies evaluated whether this partisanship influenced the news coverage of the newspapers during elections and although the impact varied, partisan endorsements were reflected in the absolute scores of bias (Kahn & Kenney, 2002; Brandenburg, 2006; Barret & Barrington: 2005).
Nonetheless, the roots of media bias are not always ideological or based on partisanship. For instance, prior research identified economic reasons for media outlets to focus on specific political parties, candidates or policy issues. Dalton, Beck, Huckfeldt and Koetzle (2010) studied whether the agenda of the media was controlled by the media as an individual actor or rather by a transactional process between different actors (the media, the people, the political actors) where consumer preferences influenced the agenda of the media. They referred to the media industry as an economic market-place (Dalton, Beck, Huckfeldt & Koetzle, 2010). This introduces the opportunity for bias. Media outlets are companies that try to fit the demand of the consumer and
cater to the consumer’s needs. Therefore, the media is able to pay disproportional attention to certain policy issues, specific candidates or other items as these subjects are frequently discussed or popular among citizens (Entman, 2007). For instance, charismatic candidates receive more coverage than opposite candidates and well-known candidates will also be covered extensively, especially during primaries, as the voters simply are less aware of the other candidates (Graber, 1991). Hence, the media outlets are able to act in response to the preferences of their consumers instead of acting out of ideological opinions of journalists and editors (Sutter, 2001). Interestingly, consumer preferences recurrently explain the partisan bias of newspapers. For example, Larcinese, Puglisi and Snyder Jr (2011) concluded that the newspapers within their study catered to the partisan beliefs of the consumers in the coverage of particular issues.
Although the debate about the nature of media bias is indecisive and inconclusive, in many occasions the media does report biased on political events (Entman, 2007). During elections, media reporting is crucially important for political candidates and biased reports are able to inflict their chances to win. To understand how the media manage to influence public opinion and voters’ perceptions of candidates with biased media reporting on the candidates, it is vital to comprehend the purpose of political campaigns and in what way the political candidates try to sway and inform voters during their campaign.
1.2. The Purpose and Construction of Political Campaigns
In the most general sense, campaigns can be best described as “organized efforts to inform, persuade and mobilize” (Farrell & Schmidt-Beck, 2002: 128). However, a political campaign is concerned with influencing the process and outcome of policy making and governance [emphasis added]. A political campaign implicates various actors (for instance candidates, parties, individuals and interest organizations) engaging in political decision-making by influencing public opinion. For citizens, political campaigns should expose the differences between candidates regarding personality, interests, stances and all other factors that may determine their performance in office. Moreover, elections in which candidates compete with each other are vital for democracy, as it provides the citizens with the opportunity to choose and evaluate their leadership (Kaplan Park & Ridout, 2006). In recent decades, politicians and other political actors progressively value political campaigning and view campaigns as an essential part of their process to influence policy making and to obtain a position within the government. Professionalization of political campaigns has increased the opportunities for political candidates and each year billions of dollars are spent by candidates to achieve their political goals (Farrell & Schmidt-Beck, 2002).
Despite the professionalization of political campaigns, the content of the campaign messages is still the vital aspect of an election campaign. As stated before, a competitive campaign provides the voters with an overview of the differences between the candidates, most importantly regarding policy issues. Candidates choose the policy issues they will promote during a campaign based on the premise that they will collect the biggest number of votes with these specific issues (Amorós & Puy, 2011). They determine what the dominant themes of their campaign will be by making an inventory of the issues dominating public opinion and try to offer solutions for the problems that seem to occupy the public debate (Trent & Friedenberg, 2008).
Furthermore, Petrocik proposed the theory of issue ownership, hereby referring to the idea that the public considers certain policy issues to be best addressed by Democrats and other policy issues to be best handled by Republicans. The reputation of these policy issues and their relations to political parties have emerged over time and are constantly evaluated (Petrocik, 1996). Accordingly, this division in policy issues determines the policy agenda of the candidates of both political parties. The candidate of the Democratic Party will voice their capability in relation to policy issues that are considered to be better handled by Democrats and Republican candidates will try to do the same (Petrocik, Benoit & Hansen, 2003). All in all, candidates choose their policy agenda deliberately and consciously to achieve the biggest number of votes. The next section discusses the media coverage of elections to determine to what extent this process of transferring stances on policy issues from the candidates to the voters is distorted or influenced by the media
1.3. Media Coverage of Election Campaigns
While candidates esteem to set the policy agenda they envisioned during an election, the power to set the agenda is in the hands of another actor: the media (Farnsworth & Lichter, 2010). Candidates and parties try to shape their media coverage by tailoring their messages to correspond with the desires of the reporters, but eventually the media outlets decide what is shared with the public. They choose whether to highlight certain policy issues, to frame events in particular ways and in which lights they portray the different candidates. As voters base their evaluations and opinions on the presidential candidates on the issues and character evaluations emphasized by the media, these choices interfere with the final vote each citizen casts (Druckman & Parkin, 2005; Farnsworth & Lichter, 2010).
Patterns in media coverage on presidential campaigns are steadily shifting in the United States. Content analyses show that, on average, substantive coverage prioritizing policy issues is decreasing while coverage about the personal and professional qualifications of the candidate is increasing and outpacing substantive coverage (Graber, 1991; Farnsworth & Lichter, 2010; Gilens,
Vavreck & Cohen, 2007). Moreover, studies have shown that the media only address a relatively small set of issues compared to the number of issues the candidates address during their campaigns. The media emphasize domestic issues as these are easily explainable to their public, thus paying less attention to foreign affairs. While candidates prefer to address broad and different kinds of issues during their campaign, the media choose to focus on narrow issues on which the candidates clash. Accordingly, most media outlets work via an ‘incentive model’, where the story has to appeal to the prospective audience for the story to be covered, instead of mirroring the actual campaign (the ‘campaign model’) (Graber, 1991). Consequently, criteria for newsworthiness, for example whether the event contains unexpected, sensational aspects or whether the event is of interest to many people, are important factors for media outlets when they consider which event to cover (Strömback, Karlsson & Hopmann, 2012).
Patterns in media coverage during elections have been the victim of criticism. For instance, the media tend to distribute ‘a general air of negativism’ regarding the candidates. They focus less on the positive accomplishments of the candidates and focus instead on the practice of mud-slinging between the candidates (Drew & Weaver, 2006; Farrell & Schmidt-Beck, 2002). Lastly, another complaint about the media coverage of presidential elections discredits the attention dedicated to who is ahead and what has happened on the campaign trail, portraying the campaign as a horse race. Journalists include public opinion polls in media reports to emphasize the horse race and these polls are used to identify the expected winner and loser of the election (Paletz, 2002). In addition, analysts stress the fact that the media engage too much in discussions of candidate personalities instead of policy issues. All the attention dedicated to scandals, horse race stories and polls distract the citizens from the important issues or problems facing the country in that particular period of time (Graber, 1991; Paletz, 2002).
Similarly, citizens heavily criticize the performance of the media and accuse the media of being biased, despite the lack of evidence and disputes regarding ideological bias. Citizens have repeatedly asked for more in-depth, fair and unbiased coverage that focuses on issues instead of controversies (Farnsworth & Lichter, 2010; Ericsen & Gottfried, 2016). The evidence for this dissatisfaction with the media coverage of presidential elections results from the citizens’ evaluations of the media’s performance during the election. Every election, citizens are asked to grade the media for their performance during the election in numerous polls. In the last three decades, the media’s performance received declining grades according to the Pew Research Center. In 2016, the poll resulted in record low grades. Almost forty percent of the citizens gave the press a failing grade (Pew Research Center, 2016b). To add to this, the majority of the American citizens
view the media negatively and describe the different forms of media bias as ‘major problems’ concerning media coverage (Gallup, 2018).
During the election of 2016, presidential candidate Trump claimed that the media covered him from a negative and biased perspective, deliberately denying his successes. Moreover, he argued that multiple media outlets, like the New York Times and CNN, included false information on the biased news reports on his campaign. On top of this, he attacked multiple journalists, news anchors and writers personally, calling them “sleazy” or “dishonest” (Shafer, 2016: 1). Clinton has criticized the media about her coverage during her campaign in her memoir, published last year. She argued the media was fixated with covering only one story: how she used a private email server during her time as secretary of state for official communication. Consequently, coverage of her stances on policy issues that she voiced during her campaign was outpaced (Fallows, Jones, Fang & Allsop, 2017; Clinton, 2017). Even the New York Times, a newspaper that had endorsed Clinton, had disappointed her. She argued how the newspaper could not stop talking about her email scandal and that she felt like the newspaper viewed her with hostility (Shepard, 2017).
Critics argued that the media directed their attention excessively to Donald Trump, as his controversial campaign and remarks created a phenomenon the media was glad to cover (Lafrance, 2016; Kellner, 2016). Studies that analyzed the media coverage of the presidential elections confirmed this criticism. The Berkman Klein Center at Harvard University produced a study that concluded how Donald Trump had succeeded in shaping the election agenda and how his media coverage had overwhelmingly surpassed Clinton’s media coverage (Faris et al., 2017). Furthermore, another study determined that Trump’s efforts to establish media attention were largely successful (Wells et al., 2016). This conclusion substantiates the opinion that Trump was able to shape his media coverage. However, although studies analyzed the media coverage of the presidential elections of 2016, fewer studies have compared the media’s issue coverage to the candidates’ campaign messages to study which policy issues were discussed by the two actors. The next section explains why such a study can contribute to this field of study and presents the research question used for the analysis.
1.4. Research Question - Scientific and Societal Relevance
The literature on media coverage of elections explained how the media pays less attention to issues compared to the candidates and how the amount of issue coverage is steadily declining. However, despite the declining amount of issue coverage, the differences between the issue coverage of the two candidates by the media could potentially signal media bias. The literature review illustrated the importance of the media for both the candidate and the citizen during an election campaign.
The media informs the citizens about the candidates’ plans for the country, which is valuable for both actors. This study is set up to analyze whether different media outlets treated the policy agendas of the presidential candidates differently when covering their campaign speeches with news reports. For this reason, the presidential election of 2016 between Hillary Clinton and Donald Trump serves as a case study. As explained earlier, the role of the media was heavily discussed during and after the election and this study offers a contribution to the debate whether the media reported biased on the presidential campaign. Hence, this thesis answers the following research question:
Was there biased media reporting on the policy agendas raised by presidential candidates Hillary Clinton and Donald Trump during the presidential election of 2016?
This question can be answered by identifying the policy agendas of the candidates and by establishing the correlation between those policy agendas and the reported policy agendas. To answer the research question, campaign messages of the presidential candidates are compared with news reports of different media outlets with the use of quantitative content analysis. This study focuses on both presidential candidates and various media outlets to identify possible differences between reports on the candidates by one media outlet, but also between the media outlets. These comparative elements create the opportunity to add to the existing literature on media bias and the presidential elections of 2016.
In prior academic research into media bias, scholars have studied media bias by analyzing media reports on different political candidates to discover, for instance, ideological, negativity or gender bias (Entman, 2010; Niven, 2001; Moldovan, 2009; D’Alessio & Allen, 2000; Entman, 2010; Bode & Hennings, 2012). Fewer studies have however probed into agenda bias (Eberl, Boomgaarden & Wagner, 2017). When reading into this subject I encountered multiple studies that mentioned in their discussion how future research should compare campaign messages of presidential candidates with media coverage, preferably different sorts of media (Bode & Hennings, 2012; Kittilson & Fridkin, 2008; Hofeman & Daigle, 2017). For example, Bode and Hennings (2012) argued that “it would be of use to consider how media coverage compares to campaign-generated content, such as speeches and press releases, to better understand the differences in media coverage” (p. 246). Other studies that focused on media reporting on policy issues, used the party agendas, press releases or television spots to identify the policy agenda of candidates (Brandenburg, 2005; Tedesco, 2005; Petrocik, Benoit & Hansen, 2003). This study takes on a different approach and uses the campaign speeches of presidential candidates, campaign messages
directly communicated to the public, as data to determine the policy agenda of the candidates. Because of the comparison between campaign speeches and media reports, this thesis is an interesting contribution to the literature on media bias.
In terms of societal relevance of this thesis, the literature review already enunciated how media bias is perceived by American citizens as an important and troubling problem in general but particularly during the latest presidential election. However, studies into media bias concluded that voters’ perceptions of media bias are inconsistent with the findings of academic research (D’Alessio & Allen, 2000). For this reason, this thesis contributes to answering the question whether the perception of partisan bias during the presidential election of 2016 was indeed justified. In addition, addressing media bias and unraveling the extent to which bias influences media coverage informs the audience of the workings of the media. Observing media bias and understanding how the media works enables the citizen to be aware and attentive of these patterns when consuming media content.
2. Theoretical Framework
The literature review presented in the first chapter of this thesis encompasses most of the information necessary to grasp the concept of media bias and to understand media coverage of American elections. However, to understand the theoretical notions behind media bias, the next chapter offers the theoretical approaches to media bias as described by academic research. Thereafter, the combination of the literature review and the close examination of media bias literature is the source for the construction of the hypotheses for this thesis.
2.1. A Closer Look into Media Bias
Bias is best understood by reflecting on its opposite: news reporting that consists of aspects such as clarity, objectivity, accuracy and avoidance of slant. Moreover, scholars argue that news articles should always be balanced and neutral. All sides or actors involved in an event should be equally represented in news reporting. Consequently, media bias is present when news reporting deviates from this set of rules (Eberl, Boomgaarden & Wagner, 2017; Druckman & Parkin, 2005; D’Alessio & Allen, 2000; Groeling, 2013).
Moreover, bias is embodied by three concepts: framing, priming and agenda-setting. Entman describes framing as “the process of culling a few elements of perceived reality and assembling a narrative that highlights connections among them to promote a particular interpretation.” (Entman, 2007: 164). Priming is part of framing because it functions as the method for framing: priming introduces salience of apparent importance to certain issues and ideas,
therefore activating the audience to think and decide in a particular way. Agenda-setting is also a part of media bias and media framing, as media outlets define within this process the problems and issues that deserve the attention of the public and/or government. Research that studies agenda-setting identifies the interactive relationship between the political actors and media outlets in regard to setting the agenda (Entman, 2007; Entman, 2010; Druckman, 2001; Van der Pas, 2014; Van der Pas, Van der Brug & Vliegenthart, 2017).
Though these concepts relate to media bias, media bias itself is a complex concept. For one, Entman distinguished slant from media bias. Slant is present when frames used within individual news reports favor one side of the political spectrum over the other. If studies demonstrating the presence of slanted news create a pattern that persists over time and across media outlets, scholars can assess the larger implications and therefore determine how the media may be distributing political power to certain individuals, parties or other causes. This is considered media bias. Therefore, Entman formulates the definition of (content) bias as: “consistent patterns in the framing of mediated communication that promote the influence of one side in conflicts over the use of government power” (Entman, 2007: 166). For this reason, Entman calls for studies that identify slant to see if there are recurring patterns that could discover the presence of bias within the media industry. In addition, Druckman and Parkin (2005) argue that the search for the objective standard by which media bias should be assessed is futile. Instead, the most effective strategy to determine media bias is to study relative comparisons of coverage. While this strategy precludes in many instances the opportunity to derive the news media’s underlying intentions or motivations for biased reporting, it allows the scholar to reach conclusions about the difference in coverage and whether or not it affects voters (Druckman & Parkin, 2005). By examining the presidential elections of 2016 and identifying bias in the news coverage of campaign speeches of the two presidential candidates, this research could possibly fit into a recurring pattern and function as the evidence for media bias within the American media environment. It is important to emphasize that this thesis cannot prove the existence of media bias in the media industry, as it will analyze only one case study, but can however add to the existing literature on media bias and furthermore identify bias within this specific election to discover whether the media could have influenced the voter’s choice at the ballot.
Literature on media bias describes three types of media bias. For one, there is coverage or visibility bias. Here, political actors are the subject of unequal amount of coverage by media outlets. As we should not expect each political actor to be covered equally, the relative amount of coverage is the focus of research studying coverage bias by comparing the coverage of various media outlets to each other (Eberl, Boomgaarden & Wagner, 2017; D’Alessio & Allen, 2000). Secondly, tonality
or statement bias refers to the process of measuring whether evaluations of political actors present in media coverage prove to be systemically more favorable or unfavorable to certain political actors compared to evaluations of other political actors. Hence, besides the amount of coverage one candidate receives, research also identifies the tone of the coverage (Eberl, Boomgaarden & Wagner, 2017; D’Alessio & Allen, 2000; Niven, 2001; Meffert, Chung, Joiner, Waks & Garst, 2006).
Although both of these types identify bias from a certain perspective, they do not specify one important aspect of media coverage: in what way the media reports on the policy agenda of political actors and specifically political candidates. The third type described in the literature takes on this subject and analyzes whether media covers the policy agenda of the candidate in roughly the same proportions as the candidates’ policy agenda. Scholars that studied this subject refer to this type of bias as ‘agenda bias’. Eberl, Boomgaarden and Wagner define agenda bias as ‘the extent to which political actors appear in the public domain in conjunction with the topics they wish to emphasize’ (Eberl, Boomgaarden & Wagner, 2017: 1128). Hence, it is not only important how often political actors are mentioned or how they are evaluated in news articles, but also whether political actors are allowed to present their policy positions and issues they emphasize in media coverage (Eberl, Boomgaarden & Wagner, 2017; Brandenburg, 2005).
Research into media bias tends to disregard agenda bias, but various scholars analyzed agenda bias during election periods. For instance, Heinz Brandenburg (2005) studied to what extent there was agenda bias present during the general elections of 2002 in Ireland and compared the party agenda of different political parties to the reported party agendas of multiple newspapers. Brandenburg identified the parties that were able to put their party agenda across through the media outlets and the parties that were less able to do so by calculating correlations between the reported party agendas and actual party agendas. In addition, John Tedesco compared the issue and strategy agenda of the presidential candidates with the journalists’ agenda during the presidential elections of 2000 and 2004 and examined the influence of the interactive relationship between the candidate and the journalist on agenda-setting (Tedesco, 2001; Tedesco, 2005). Both Brandenburg as well as Tedesco created one reported policy agenda for each media outlet. They listed the number of issues the media had paid attention to within a specific period of time and tracked how frequent these issues were mentioned.
As discussed before, during elections voters get the chance to compare the policy stances of the candidates with each other to see whether a candidate fits their own beliefs and stances. Candidates wish to convince the voter they are better capable to handle a certain issue compared to the opposite candidate (Petrocik, Benoit & Hansen, 2003; Brandenburg, 2005). News outlets are in the position to suggest to voters and news audiences the specific policy issues on which they
should assess the different political candidates and parties by linking these political actors to certain policy issues (Eberl, Boomgaarden & Wagner, 2017; Brandenburg, 2005). Consequently, it can damage the political candidates when the media links the candidates to issues that are not ‘owned’ or emphasized by the political party they belong to (Petrocik, 1996). As literature pointed out that candidates very carefully create their policy agenda and fill this agenda with issues that are perceived as best handled by their political party and will advantage them at the polls, we can assume that the media’s policy agenda will benefit the candidate when the reported agenda correlates strongly with the agenda of the candidate. This assumption is based on the research of Eberl, Boomgaarden and Wagner (2017) who confirmed the hypothesis that when the reported party agenda of media outlets is congruent with the actual party agenda of a specific party, this will increase the willingness of voters to vote for that party. Hence, when the congruency with the policy agenda of the candidates differs between the candidates within one newspaper’s coverage, this difference can be influential – one of the premises of media bias (D’Alessio and Allen, 2000; Entman, 2007).
2.2 Hypotheses
In addition to studying to what extent there was biased media reporting on the policy agenda of Clinton and Trump, this thesis also tests whether partisan bias could explain possible evidence for media bias. The literature review discussed the contradicting results on ideological and partisan bias but also illustrated the effect of endorsements on bias in media coverage. Accordingly, this thesis analyzes newspapers that endorsed one of the two presidential candidates to examine whether these endorsements can explain any difference in the issue coverage of the candidates. Prior research demonstrated that the impact of endorsements varies. However, Kahn and Kenney analyzed the effect of endorsements on the news coverage by newspapers and concluded that these endorsements shaped the news coverage about the candidates. The candidate that was endorsed by the newspaper received more favorable coverage than the other candidate(s) (Kahn & Kenney, 2002). The study of Barret and Barrington (2005) amplified these results as they found out that newspapers that endorsed a particular candidate generally published more favorable photographs of this candidate. By combining these results with the assumption that candidates will benefit from issue coverage that is roughly in the same proportions as their policy agenda, the following two adjacent hypotheses are created:
H1.1: The correlation between Trump’s reported policy agenda as published by the newspapers that endorsed Trump and Trump’s own policy agenda will be higher than the correlation between their reported policy agenda of Clinton and Clinton’s own policy agenda.
H1.2: The correlation between Clinton’s reported policy agenda as published by the newspapers that endorsed Clinton and Clinton’s own policy agenda will be higher than the correlation between their reported policy agenda of Trump and Trump’s own policy agenda.
3. Methods
This chapter describes the method used to answer the research question and to test the hypotheses. The first section discusses the research design, while the other sections demonstrate the process of collecting data, the operationalization of the content analysis and explain the process of interpreting the data by using statistical analyses. The last section considers the empirical limitations of the method.
3.1 Research Design
To answer the research question and test the hypotheses raised in this thesis, I used content analysis as a method. Content analysis is used widely to analyze the content of political messages. William Benoit offers the following definition for content analysis: “the measurement of dimensions of the content of a message in a context. Content analysis can be employed to describe a group of related messages, draw inferences about the sources who produced those messages, or draw inferences about the reception of those messages by their audience” (Benoit, 2011: 269). Content analysis focuses on the verbal part of these messages – ideas and arguments expressed in words. However, the visual elements of messages can be just as important and multiple studies have delved into analyzing political images (Grabe & Bucy, 2009; Benoit & McHale, 2003). Content analysis has several other advantages. For example, it allows the researcher to draw conclusions from content evidence without relying on the participation of the communicators of this content. Also, content is largely archived and is therefore available for research beyond its production and consumption (Riffe, Lacy & Fico, 1998).
The content analysis is performed by a human coder, the most frequently used technique for content analysis (Neuendorf, 2011). Because the design of the content analysis performed in this thesis is subject to the use of interpretation, human coders are more fitted to perform the analysis compared to computer-assistance (Krippendorff, 1989). The content analysis performed in this thesis is quantitative instead of qualitative. During quantitative content analysis, communication content is systematically assigned to categories according to rules, and these categories are analyzed using statistical methods (Riff, Lacy & Fico, 2014). Hence, the content analysis of this thesis consists of three parts: collecting the communication content, categorizing
the data with the use of a codebook and lastly, running statistical analyses to determine the answer to the research question. I chose quantitative content analysis as the method for this thesis primarily because of the sample size used for this thesis. To determine media bias during a presidential election, one of the qualifications should be that the bias is consistent (D’Alessio & Allen, 2000). Although this thesis does not analyze whether agenda bias was present during multiple elections, it does analyze the presence of bias by examining the last four months of the campaign. Consequently, this leads to a quite large sample size of speeches and articles and quantitative analysis enables the author to present patterns and take conclusions out of large amounts of data (Riffe, Lacy & Fico, 1998; Neuendorf, 2011).
3.2 Data and Unit of Analysis
For the content analysis, I chose to focus on the period between July 28th 2016 until the day before Election Day, November 7th 2016. I chose July 28th 2016 because this was the date Hillary Clinton accepted the Presidential nomination at the Democratic convention in Philadelphia, Pennsylvania. At this time, Donald Trump had already accepted his nomination a few days before. As this study focuses on the period of time both candidates were campaigning, November 7th is chosen as the last day. The choice for this period of time was inspired by other studies that analyzed the presidential elections of 2016 and also chose to analyze a period of time similar to the period chosen for this thesis (Tedesco & Dunn, 2018; Hofeman & Daigle, 2017).
To study agenda bias during the presidential elections of 2016, both campaign messages as media coverage on these campaign messages are analyzed. For the analysis of the campaign messages, I collected all of the campaign speeches held by Clinton and Trump during this specific period of time. Several databases had already listed and collected these speeches1. Press releases or other (online) campaign messages of both candidates were not included, because during campaign speeches the presidential candidates have the ability to discuss certain policy issues in-depth and these speeches are intended to directly influence the voters as they are held in front of the public. Consequently, these speeches contain the policy issues that the candidate would like to inform the public about and therefore suffice for extracting the candidate’s policy agenda. In addition, Hofeman and Daigle (2017) compared campaign messages with news coverage during the presidential elections of 2016 and chose online campaign content as data for their analysis but advised in their discussion that further research should incorporate a more traditional form of campaign messaging, like campaign speeches. The transcripts of the campaign speeches were
derived from the internet and downloaded and saved. Campaign events where Trump or Clinton only addressed or welcomed the people gathered shortly were not included in the sample. Speeches made on the campaign trail by other persons than the presidential nominees were filtered out as well. Within this process, I collected 50 speeches of Clinton and 63 speeches of Trump.
Data of media coverage was included by collecting news articles written by American newspapers. I chose newspaper content as media coverage for several reasons. Although research demonstrated that the popularity of newspapers as a source of information for citizens is declining, a study into American citizens’ media behavior concluded that citizens rely as much on newspapers as on social media for information. Moreover, together with television news, citizens trust the American newspapers the most out of all media sources “to provide mostly accurate and politically balanced news.” (Gallup, 2018: 3). By analyzing newspapers instead of any other news source, this study tests whether American newspapers deserve citizens’ trust in delivering balanced news. Lastly, newspapers tend to direct more attention to policy issues when reporting about elections when compared to other media outlets such as television news (Jalalzai, 2006) and is therefore more suitable for the purposes of this study.
To determine whether possible agenda bias could be explained by partisanship of the newspapers, I included three newspapers that endorsed one of the two candidates. Two newspapers that endorsed Hillary Clinton during the general election (The Washington Post and the New York Times) and one newspaper that endorsed Donald Trump during the primary election (the New York Post) were chosen. The editorial board of the New York Times endorsed Clinton on September 24th 2016 by summing up her qualifications and mentioning that they “believe Mr. Trump to be the worst nominee put forward by a major party in modern American history” (2016: 1). Three weeks later, the editorial board of the Washington Post endorsed Clinton as well, declaring they “endorse her without hesitation” (2016: 1). The editorial board of the New York Post did not endorse a candidate during the general election but did endorse Donald Trump during the Republican primary election and argued that Donald Trump possessed the skills and the values to “live up to his campaign slogan: to make America great again” (2016: 1). In 2017, the New York Times enjoyed the highest circulation of all newspapers in the United States, whereas the New York Post possessed the fourth largest circulation and the Washington Post the sixth largest circulation (McIntyre, 2017).
For the purposes of this thesis, I included only the articles written about a specific campaign speech in the content analysis. This choice creates the opportunity to analyze in what way each speech was covered by the newspapers and to what extent the newspapers covered the content of the speech. The articles were derived from LexisNexis. To collect the articles written
about a specific speech, I used a combination of two search terms: the name of the presidential candidate together with the place where the speech was held (for instance: Donald Trump AND Youngstown). When these search terms were too broad (for instance, the search term ‘Washington, DC’ resulted in 79 articles), I added the exact location of the speech. The three consecutive days after the day of the speech was the period used to search for these articles. Then, all three newspapers were given up separately as a source to search for articles that contained these two search terms. Then, I screened the articles to see whether the article was indeed referring to the campaign speech, filtering out articles that did not discuss the speech but mentioned both search terms for other purposes. Moreover, articles that were published on the website of the newspaper were omitted, only articles that were published in the actual newspaper were included. Articles that discussed or referred to the speech within the text of the article were then downloaded onto the computer. In some cases, I had to download articles twice because they referred to multiple speeches in one article. However, this did not occur often. The number of articles collected per newspaper and per candidate are described in table 1 below.
Table 1. Number of Collected Articles per Candidate
Newspapers
Candidates New York Times Washington Post New York Post
Donald Trump 51 55 27
Hillary Clinton 50 41 10
The combination of the transcripts of the selected campaign speeches and the collected news articles translates to the data used in the analysis of this thesis. The unit of analysis is one sentence. During the content analysis, I counted the number of sentences that included policy issues per speech or news article. However, I made a distinction between different policy issues. Sometimes one sentence could entail multiple references to different policy issues, therefore the number of policy issues that were mentioned within a speech or news article were also counted. Consequently, when the results are presented the unit of analysis will sometimes be the number of sentences but the results can also be presented as the number of policy mentions within the sentences discussing policy. In addition to the collected data, I used literature, news articles and reports on the presidential election of 2016 to put the results from the content analysis into a broader perspective and to substantiate them in the discussion section of this thesis. The next section discusses the operationalization of the research design.
3.3 Operationalization
To analyze the data and identify to what extent there was agenda bias present in the media coverage of the presidential candidates, the next step for the content analysis is to operationalize the method by creating a codebook. Within this codebook, different subjects or policy propositions that candidates speak about are coded into topics (‘policy issues’).
When conducting context analysis, categories are employed to link numeric values to dimensions of messages. Categories should be relevant, exhaustive and mutually exclusive and the parts of the text can only be linked to one category (Benoit, 2011). Categories can be set up deductively or inductively. When the categories are derived inductively, the researcher will study the texts to produce a list of categories or topics. In addition, the categories can also be derived deductively. This means that the categories will be set up before reading or studying the content and are extracted from theory or previous research (Benoit, 2011; Elo & Kyngäs, 2007). I created the list of issues used for this thesis mostly deductively - the list is based on other content analyses that focused on issues while analyzing media coverage, campaign speeches or party programs (Petrocik, 1996; Bode & Hennings, 2012; Tedesco, 2005). In addition, the issues listed in the codebook were carefully selected through the examination of public opinion polls rating the most important issues of the nation and the presidential campaign in 2016, for instance the polls of Pew Research Center. In July 2016, they released a list of the top voting issues for the election, emphasizing the importance of salient issues like the economy, foreign policy and health care, but also listing relevant issues like the appointment of Supreme Court justices and immigration (Pew Research Center, 2016a). By including these issues, the codebook is based on the relevant issues during this specific year. During the content analysis, one issue was added to the codebook (‘government functioning’) because I noticed Donald Trump paid a lot of attention to this issue and therefore could not be excluded from the list. For this reason, the codebook was also created partly inductively.
The list encompasses seventeen categories that range from the economy to foreign policy and healthcare to gun policy. Identity and minority issues like the treatment of ethnic minorities, LGBT-rights and women’s rights were coded into three different categories, as each of these issues comprehended an important part of the debate during the election of 2016 (Lilla, 2017; Pew Research Center, 2016a). Besides healthcare, which I coded into a separate category, all other social issues were coded together within the category ‘social issues’. Policy issues regarding the environment or energy industry were coded into one category. All issues that emphasized law and order, justice or American values were assembled under the category ‘civil order’. Table 2 offers
the different categories or ‘policy issues’ listed in the codebook. During the content analysis, statements on issues that did not fit into one of these categories were not considered.
Table 2. List of Policy Issues Included in the Codebook Economy Terrorism Foreign policy Healthcare Gun policy Immigration Social issues Education Supreme Court
Treatment of ethnic minorities LGBT-rights Trade Environment Women rights Civil order Infrastructure Government functioning
The complete codebook is located in the appendix (Appendix 1) of this study. After developing the codebook, I applied the codebook to the different speeches and news articles. Although the candidates mentioned these issues in many ways, for example by attacking their opponent’s position on a specific issue, during the content analysis only the sentences that proclaimed the stance of the candidate on a particular issue were coded. Consequently, most of the sentences began with “I believe (…)”, “We’re going to (…)”, “We need (…)”, “When I’m your President (…)”, etcetera. Every sentence indicated what the candidate envisioned for a specific policy issue. The candidate often mentioned two or more different policy issues in one sentence. For instance, the sentence: “We're going to do everything we can to catch up to the rest of the world: affordable childcare, paid family leave, earned sick days, jobs with rising incomes.”, was coded as both ‘economy’ (because the candidate mentioned jobs with rising incomes) and ‘social issue’ (because the candidate mentioned childcare and paid family leave). This way, I counted both the number of sentences that contained policy issues and the amount of policy issues mentions.
The news articles were coded slightly different than the campaign speeches. Because the article was never completely reflecting on the campaign speech, first I counted the number of sentences that actually touched on the specific campaign speech. Only sentences that mentioned the speech (“Thursday’s address”, “Clinton’s speech in Tampa”) and discussed the content of the speech were counted. Multiple times the article did not discuss the content of Clinton’s speech, but paid attention to the guest speakers or introductory speeches that took part at the same venue.
As these specific speeches were not coded, I did not take the sentences referring to these speeches into account. The sentences that discussed the speech were read and filtered to separate the sentences that discussed policy from the sentences that did not discuss policy. The number of sentences that did discuss policy issues were counted, as well as the different policy issues that were mentioned within these different sentences. Within the content analysis of these articles, I included both the sentences that displayed the candidate’s opinion or plans for this issue and the sentences wherein the author mentioned that a policy issue was the subject of the speech or that a speech was characterized by a specific policy issue were coded. Including these sentences as policy mentions attains to creating a complete picture of what kind of policy issues these newspapers are linking to the presidential candidate within their news articles. This process created an overview of the frequencies each policy issue was mentioned in every speech and news article collected.
The content analysis of this study was performed by one coder, the author of this thesis. Although the inter-coder reliability test usually is performed to test whether multiple coders within one study agree on the coding scheme when applying this codebook to the content of interest, I decided to run the test to determine to what extent the content analysis depended on the interpretation of the coder. When the coder has to keep track of multiple categories at the same time, as is the case within this study, there is a greater chance for coding errors (Campbell, Quincy, Osserman & Pederson, 2013). To test whether the same results could be replicated by another coder, an independent coder performed the inter-coder reliability test (Appendix II). The independent coder coded a sample of twenty speeches (ten speeches of Hillary Clinton and ten speeches of Donald Trump). Due to the limited time available to the author and the independent coder, the independent coder was asked to keep track of the policy issue ‘economy’. The Krippendorff alpha, calculated after receiving the results, turned out to be 0.85. Literature on the inter-coder reliability test argues that a score of 0.80 is acceptable in most situations while a score of 0.90 is acceptable in all situations, and how lower criteria should be used for indices known to be more conservative, like the Krippendorff alpha (Lombard, Snyder-Duch & Bracken, 2002; Matthes & Kohring, 2008). Therefore, the score of 0.85 is accepted.
3.4 Data Analysis
When the relevant news articles and campaign speeches were collected and analyzed using the codebook, I used the collected data to perform multiple statistical analyses. The first step was the creation of the policy agendas of the candidates and the reported policy agendas. These agendas were created by presenting the total amount that each policy issue was mentioned within the content (speeches or articles) as a percentage of the total number of policy issues mentioned within
the content. Other studies identified one reported policy agenda for each newspaper, accumulating all the different policy issues that were mentioned within the articles (Brandenburg, 2005; Eberl, Boomgaarden & Wagner, 2017; Tedesco, 2005). Instead, I created different reported policy agendas for each presidential candidate from the collected content.
To test the hypotheses, I calculated correlations within the data analysis to establish to what extent the three newspapers’ issue coverage correlated with the content of the policy agendas raised within the campaign speeches of Clinton and Trump. The correlation between the reported policy agenda of each candidate and the candidate’s policy agenda was calculated. The use of this statistical method was inspired by other scholars that studied agenda bias (Brandenburg, 2005; Eberl, Boomgaarden & Wagner, 2017; Tedesco, 2001; Tedesco, 2005). In this process, I used the Pearson’s correlation, measuring how similarly the two policy agendas shared common rankings (Riffe, Lacy & Fico, 1998).
Moreover, I used a variance of descriptive statistics in multiple sections of chapter four to demonstrate the differences in issue coverage between the three newspapers. Hence, measures like the mean, standard deviation and percentages are presented in chapter four. In addition, I conducted multiple Chi-square test of independence to determine whether the differences in issue coverage were significant. Chi-square indicates whether there is a significant difference between two distributions measured at the nominal level (Riffe, Lacy & Fico, 1998). In the first Chi-square test, I compared the number of speeches that was covered by each newspaper with an article between candidates to see whether one of the candidates was significantly covered more by the individual newspapers compared to the other candidate. For this reason, I coded the data into nominal variables. In the first test, each speech was specified as being of ‘Clinton’ or ‘Trump’ and was coded as ‘covered’ or ‘not covered’ by each newspaper. In the second test, the articles of each newspaper written about the campaign speeches were compared between candidates to determine whether policy issues were significantly mentioned more in connection to one candidate compared to the other candidate. For this reason, I coded each article of the three newspapers as containing ‘policy’ or ‘no policy’ and whether the article was about ‘Clinton’ or ‘Trump’. With these datasets, I performed the Chi-square tests between the variables ‘candidate’ and ‘coverage’ for each newspaper as well as between the variables ‘candidate’ and ‘policy’ for each newspaper.
The results are all based on quantitative analysis of the collected data, but within the discussion section the results of other studies regarding this subject are taken into account to discuss and illustrate the results obtained from the statistical analysis.
3.5 Empirical Limitations of the Research Design
During this thesis, I made multiple choices that affect the generalizability of this study. This research design is subject to several limitations. For one, this thesis focusses on one case, the presidential election of 2016. Hence, the results of this study are not compared with other presidential elections to identify a pattern of behavior for different media outlets. As each presidential election proceeds under different circumstances and media coverage is heavily influenced by the characteristics of the particular candidates of that year, this limits the generalizability of this study.
Secondly, I used one news source – newspapers - to study media bias in the coverage of the policy agendas. However, other media sources like televised news, blogs or radio news are also relevant to examine media bias. Additionally, online news sources, like Breitbart or Huffington Post are interesting sources to study agenda bias as the use of these news sources together with social media is becoming increasingly popular among citizens (Gottfried, Barthel & Mitchell, 2017). Furthermore, for this study the articles of only three different newspapers were used. Initially, this study was supposed to study two newspapers that endorsed Clinton or are considered left-wing and two newspapers that endorsed Trump or are considered right-wing, however, due to issues with the LexisNexis database (certain right-wing newspapers were not available) and the limited time available, only the New York Post is analyzed. However, further research into agenda bias should widen the scope of newspapers when the time enables them to do so to create a bigger picture of the reach of the results presented in this thesis. Instead of focusing on left-wing and right-wing newspapers, it would also be interesting to analyze a newspaper that ideologically is more located within the center and refrained from endorsing a presidential candidate, like the USA Today.
Another limitation is the fact that I was not able to count or track the total amount of sentences per speech or per news article. Therefore, I could not determine the exact proportion of the number of sentences that mentioned policy issues as a part of the total number of sentences used for the speech or news article. Furthermore, I did not code the content of the articles and speeches that did not discuss policy to determine what the candidates and journalists discussed besides policy issues because the subject of this thesis focuses only on the policy agenda of the candidates. However, incorporating other subjects and categories besides policy issues in the codebook and determining the tone of the news articles could have been valuable to determine bias within the media coverage. Lastly, I made the choice to create a sample by collecting the news articles written directly about each individual speech published in the three consecutive days after a speech. Nonetheless, this sample of articles is quite small compared with the total amount of
articles written about the candidates during the election and incorporating other news articles discussing the policy issues of the candidates in the content analysis could broaden the scope of the analysis.
4. Results
This chapter comprises of two sections: the first section presents the policy agendas of the candidates together with the correlations between these policy agendas and the reported policy agendas of the newspapers in order to test the hypotheses. The second section offers additional information about the total amount of issue coverage within the news articles. Although unprecedented, the content analysis resulted in interesting numbers regarding issue coverage that put the results presented in the first section of this chapter into perspective. Therefore, these results are included as they are considered valuable for assessing media bias. The results presented in chapter four are discussed in the last chapter of this thesis.
4.1 The Policy Agendas of the Candidates and the Reported Policy Agenda Compared The first section discusses the policy agendas of Donald Trump and Hillary Clinton as well as the reported policy agendas of the different newspapers. Next to this, the section presents the correlations between the policy agendas and the reported policy agendas that are calculated to determine whether the hypotheses can be confirmed or rejected based on these correlations.
4.1.1 The Policy Agenda of the Presidential Candidates Hillary Clinton and Donald Trump The content analysis of the presidential candidates’ speeches produced the following results. First of all, Donald Trump mentioned his policy issues in 2291 sentences within the 63 speeches he gave in the chosen period of time and therefore he mentioned policy issues in averagely 36 sentences per speech. To the contrary, Hillary Clinton mentioned her policy issues in 1179 sentences in the 50 speeches she held and she mentioned policy in 23 sentences per speech on average. This means that Clinton dedicated less attention to policy issues during her speeches compared to Trump.
The content analysis of the speeches facilitates the opportunity to create the policy agenda of the candidates. The total number of mentions per policy issue are presented as a percentage of the total amount of policy mentions in table 3. This way, the different policy issues that were considered important by the candidates to affiliate themselves with are detected. Both of the candidates dedicated the majority of their attention to their plans for the American economy during their campaign speeches. Clinton even emphasized her plans for the American economy in more than 37% of all the times she referred to her policy agenda.
Table 3. Policy Agenda of the Presidential Candidates (%)
Policy issues Donald Trump Hillary Clinton
Economy 25.3 37.3 Terrorism 3.8 0.4 Foreign policy 9.3 4 Healthcare 4.3 3.4 Gun policy 1.3 2.8 Social issues 6.4 4.4 Immigration 10.8 1.2 Education 6.4 17.1 Supreme Court 1.6 0 Ethnic minorities 2.6 3.3 LGBT-rights 0.1 1.8 Trade 6.1 0.8 Environment 4.8 6.2 Women rights 0.4 4.4 Civil order 8.6 3.3 Infrastructure 4.3 7.5 Government 6.7 0.1 functioning N (= policy mentions) 1447 3054
Besides the fact that both candidates frequently referred to their plans for the economy, Donald Trump discussed his policy on immigration thoroughly and Hillary Clinton emphasized her plans for the education system repeatedly. Furthermore, the policy issue ‘government functioning’ beholds an interesting discrepancy between the two presidential candidates. Donald Trump directed almost seven percent of his attention on his policy agenda to the changes in government functioning he would enforce when in office, while Hillary Clinton refrained almost completely from talking about this issue. This is however not surprising, as the president in office at the time, Barack Obama, was a Democrat and campaigning together with Clinton (White, 2016; Rhodan, 2016). Therefore, Clinton was generally positive about Obama’s eight years in office, while Trump did everything he could to paint a picture of total despair when referencing to President Obama.
On average, Donald Trump divided his attention more equally over the different policy issues, as the standard deviation of his policy agenda is 5.8 as opposed to the standard deviation of Hillary Clinton’s policy agenda being 9. For example, Clinton paid almost no attention to government functioning, terrorism, trade and immigration. In addition, Trump refrained almost completely from talking about women’s rights and LGBT-rights. The few occasions Trump mentioned these rights, he argued the United States should protect these rights by banning Muslim immigrants from the country.
This data presents the respective policy agendas of the candidates. The question is: how well did the presidential candidates manage to put their policy agendas across through the media? To what extent correlate the reported policy agendas with the policy agendas of the candidates?
4.1.2. Correlations between the Policy Agendas of the Candidates and the Reported Policy Agendas
This section features a first look into the coverage of the New York Times, the Washington Post and the New York Post. As the total amounts of policy issues mentioned by the candidates resulted into the creation of policy agendas in table 3, the same was done for the content analysis of the news articles. Reported policy agendas were created by noting the total amount of mentions for the different policy issues as a percentage of the total amount of policy mentions in all the articles of one newspaper. While the candidates’ own policy agendas indicate which issues they would like to be associated with, the reported policy agendas of the newspapers tell us which issues they actually were associated with. Instead of including the reported policy agendas per newspaper, the content of the reported policy agendas is discussed through the examination of the correlations between these agendas and the actual policy agendas below.
Table 4. Correlations between Policy Agendas and Reported Policy Agendas Reported policy agendas
Policy agendas New York Times Washington Post New York Post
Donald Trump 0.66* 0.39 0.86**
Hillary Clinton 0.98** 0.74** -
* Correlation is significant at the 0.01 level (2-tailed) ** Correlation is significant at the 0.001 level (2-tailed)
Estimating bivariate correlations between the policy agendas of the presidential candidates and the reported policy agendas allows the opportunity to make confident statements about to what extent
