Constructed Characters: How the Elite Media Communicates Political Figures

• In Event: Elites in the News

Abstract

Politicians and laypeople alike often argue that the media spends too much time discussing trivial matters, such as personal characteristics and scandals, and not enough time covering what really matters, policy. Politicians from both sides of the aisle regularly attack the press for being biased, superficial, or fake. While it is easy to see why political figures often want to believe the worse about the media, a question emerges: Is the press really biased or are they simply doing their jobs to keep politicians accountable? The purpose of the current study is to investigate the basis of these attitudes and to discover how the media covers political figures, specifically U.S. presidential candidates.
On the one hand, past research has established on a topical level how the media covers political campaigns and figures. Multiple studies covering multiple elections and media venues have consistently found the majority of the media coverage focused on the ‘horse-race’ of elections (e.g. polls, strategies, campaign events). Followed by the character of candidates with policy issues being one of the least covered topics (Benoit, Stein, & Hansen, 2008; D’Angelo, Calderone, & Territola, 2005; Kern, 2001). Additionally, researchers have consistently found that media coverage of political candidates and leaders to be more negative than positive in tone (Cohen, 2008; Farnsworth & Lichter, 2012). Although from these studies we have learned general coverage, there is little information about specific topics of communication.
Expanding on this past work, the current study moves beyond broad topics to understand more specifically how the media discusses political candidates. Instead of using pre-determined topics, here an applied data-driven approach was used to discover the topics in two large media corpora. The aim of this project was to demonstrate how political candidates are communicated by elite media sources and how that communication relates to perceptions of the media.
Two corpora of New York Times articles and CNN transcripts, over 2 million references to major party presidential candidates were identified from the last five election cycles. For each reference, the immediate context (15 words before and after) were extracted to capture what was being communicated about the candidate. Individual word frequencies were calculated for each reference’s context which were the basis of analysis.
The first analysis focused on determining which words were most closely associated with each candidate. Words were found by correlating the frequencies of the words with a dummy-coded variable for each candidate (e.g. words more likely to be used to describe Donald Trump than other candidates). Across corpora, words most likely to be associated any given candidate tended to reflect their past records or perceived scandals. For example, mentions of Hillary Clinton were more likely to be accompanied by the words email, fbi, comey, and foundation while mentions of Donald Trump were more likely to associated with the words tower, apprentice, casino, and twitter. Consistent with common perceptions and past research, words relating to the candidate’s policy positions were absent (with George W. Bush’s foreign policy being the sole exception).
To test the robustness of the trends, the Meaning Extraction Method was used (MEM; Chung & Pennebaker, 2008), a topic modeling technique which used Principal Components Analysis to uncover meaningful ways in which words co-occur. As in past studies, ‘horse-race’ coverage emerged with topics such as campaign events, strategy, staff, and polling appearing as some of the most frequent topics, However, this topic approach revealed that policy issues were also central to the media coverage of the presidential campaigns. In both the NYT and CNN, foreign policy, health care, the economy, social issues, and terrorism all emerged as meaningful, common topics. Coverage of candidate attributes emerged through topics about the candidates’ families and scandals. Together, the results of the topic models replicate past studies while at the same time providing greater specificity about the content of the communicated media messages.
In sum, the current study contributes to existing literature by providing a novel data-driven, contextualized approach to better understanding media coverage of political figures. Politicians may wish they could always speak to the public directly, but often what the public hears in mediated through the press. Therefore it is vital to understand how the press is communicating information from and about political figures. In a democracy, the press has the necessary function of keeping politicians honest and keeping voters informed; this work shows that the (elite) media is fulfilling those roles by communicating candidate’s policy positions and informing voters about potentially disqualifying behavior even if more trivial campaign strategy issues elicit more coverage.

Author

• Kayla Nicole Jordan, University of Texas at Austin