Abstract
In this article, we examine both the content and effects of opinion shows during the 2008 presidential election. First, a content analysis shows that opinion shows devote most of their attention to attacking the opposition candidate, rather than praising the like-minded candidate. Second, analyses of panel data show that exposure to opinion shows made viewers less (more) favorable toward the opposition (like-minded) candidate. Finally, we use overtime analyses to show that coverage of the opposition candidate affects attitudes toward both candidates, whereas coverage of the like-minded candidate has negligible effects on attitudes toward either candidate.
Opinion shows have become a prominent part of the modern media landscape. Conservative talk show hosts dominate political radio by opining to their listeners. Chief among the talk show hosts is Rush Limbaugh, who refers to his listeners as dittoheads, because they are supposed to listen up and “say ditto.” Meanwhile, consumers need only to turn on their computers or their smartphones to access their favorite talking heads’ opinions on current events via the blogosphere or the twitterverse. Opinion shows have even come to dominate 24-hour cable news. Perhaps the best evidence of their popularity is the prime-time placement of opinion shows on Fox News and MSNBC. Indeed, opinion shows are some of the most successful shows on cable news outlets, with Fox News’ The O’Reilly Factor leading the way (Carter 2009).
Despite the popularity of opinion shows, scholars know very little about how those programs affect public opinion. What role do opinion shows play in media effects? Are opinion shows more persuasive than news shows? If so, how do opinion shows shape public opinion? These are the primary questions of this research. We answer these questions by examining the independent effect of watching opinion programs on cable news. We make the case that opinion show content has a distinct and powerful effect for two reasons. First, opinion show content is more likely to be negative toward the opposition. Second, opinion show content is more likely to be one-sided. Our findings seek to advance current empirical understanding of media effects in a diffuse and dynamic media environment in which the line between opinion and news is increasingly blurred.
Partisan Media Effects
The expansion of the news media marketplace over the last thirty years has provided Americans with a variety of news sources. An important consequence of this transformation was increased competition for audience share that makes it more profitable for news outlets to cater to niche audiences (Mullainathan and Shleifer 2005). Consequently, a large number of news outlets on television, radio, and the Internet appeal to political partisans by providing news coverage that has either a liberal or conservative slant.
The rise of partisan news sources allows Americans to get their news from ideologically consistent sources. Recent research provides significant evidence that Americans engage in selective exposure to partisan news sources: Republicans tend to get their news from Fox News and Rush Limbaugh, whereas Democrats are more likely to watch MSNBC and CNN (Stroud 2008). In an experimental setting, Iyengar and Hahn (2009) found that Republicans preferred news stories that were attributed to Fox News—regardless of content—whereas Democrats preferred stories that were attributed to CNN. Thus, partisans prefer to consume news from like-minded news sources.
Many scholars, pundits, and politicians have pointed to the dangers of a highly partisan news media. Partisan news sources allow people to isolate themselves from contrary arguments (Sunstein 2001), and this selective exposure can increase hostility and polarization because neither side understands the rationale for opposing viewpoints (Mutz and Martin 2001). Recent research suggests that such fears regarding partisan news sources might be warranted. Both survey and experimental evidence demonstrates right-wing talk radio affects the attitudes of listeners (Barker 1999; Jamieson and Cappella 2008). In addition, Natalie Jomini Stroud (2010) provided convincing evidence that selective exposure increases polarization, while the reverse was not supported. In other words, exposure to partisan news sources results in more polarization, not the other way around.
Much of the research on partisan media effects centers on Fox News, with a bevy of scholars showing its “Fair and Balanced” coverage is actually conservatively slanted. Aday (2010) found that Fox News was less likely to report on negative news involving Iraq and Afghanistan than other sources during George W. Bush’s administration. There is also evidence that Fox News was mostly negative toward Democratic politicians (Morris and Francia 2010). Such one-sided coverage has been found to significantly affect viewer attitudes (Morris 2007). Morris and Francia (2010) found that Fox News viewers became less favorable toward John Kerry—and more favorable toward George W. Bush—following the 2004 Party conventions. Furthermore, DellaVigna and Kaplan (2007) found that the introduction of Fox News into new media markets corresponded with increased vote shares for Republican candidates.
Largely missing from recent research on partisan media effects is the distinction between news shows and opinion shows. This shortcoming is explained in part by limitations that are characteristic of survey measures of television news exposure. Media effects research often relies on telephone surveys to measure audience attitudes, but time limitations prevent researchers from measuring all of the television sources respondents’ use on a regular basis. Consequently, researchers typically rely on respondents’ self-assessment of their primary news source on television (Morris and Francia 2010; Stroud 2011). Unfortunately, measuring viewers’ primary news source fails to account for the variety of show types that people watch. Differentiating between exposure to opinion shows and straight news programs should provide a more complete understanding of how Fox News affects audience attitudes.
We address this shortcoming in the media effects literature by examining the distinct content and effects of news and opinion shows on cable news. Specifically, we first compare the way news and opinion shows covered the election to determine whether opinion shows covered the 2008 presidential candidates differently from news shows. We then test whether exposure to news or opinion shows—at the individual level of analysis—had a stronger effect on viewer attitudes during the 2008 presidential election. As far as we know, this is the first study to examine the independent effects of opinion shows on viewer attitudes in a natural setting. 1
To increase the robustness of our discussion, we also incorporate overtime variations in partisan media coverage to explain changes in audience attitudes. If variations in content precede audience attitude change in predictable ways, it is safe to conclude that it was media content—and not characteristics of the audience—that caused viewers to change their attitudes. We use the amount of coverage on Fox News—concerning the two main presidential candidates—to predict subsequent changes in favorability toward the two candidates. Including media coverage in overtime analysis allows for a stronger test of media effects than most previous studies.
We chose to focus most of our attention on Fox News for methodological reasons. For one, Fox News had a large number of shows included in the survey we use for analysis, which allowed us to compare news shows with opinion shows. In addition, Fox News released a press statement, in which they make a clear distinction between their news and opinion programs (Stelter 2009). This classification indicates the expectations of the Fox News Network in regard to its programming. Although our primary interest is Fox News, we also examine MSNBC’s news and opinion programs where possible to provide comparison. It is important to note that our conclusions about MSNBC are limited given the lack of an objective categorization of news and opinion programs for this network.
Theory: How Opinions Change Opinions
Opinion shows present one-sided content similar to opinion columns and editorials in newspapers or political talk radio shows. More prominent newspapers such as the New York Times or the Wall Street Journal present editorials that typically slant to the left and right, respectively. An ideologically one-sided Op-Ed section can be defended on the grounds that news organizations establish a wall of separation between news and opinion.
Although this wall of separation between opinion and news may be theoretically useful, it does not diminish the distinct effects of exposure to opinion shows versus news shows. Opinion shows should be persuasive because they present one-sided debates to viewers. The format for opinion shows is typified by a charismatic host who reports his or her opinions on the headlines of the day often alongside “expert” guests who share the same views as the host. The ideological opinion content that characterizes these sorts of shows is not bound by the professional norms that require journalists to present balanced discussion of political issues. Instead, opinion shows often present the opinions of the host(s) without any reference to opposing viewpoints. For example, on The O’Reilly Factor, host Bill O’Reilly begins each show with a “talking points memo” in which he expresses his opinions on the top story of the day.
Given the difference in content between opinion shows and news shows, it is reasonable to expect that audience members of opinion shows are exposed to one-sided coverage. Moreover, it is precisely this sort of one-sided information flows we can expect to have a significant persuasive effect on attitudes (Zaller 1992). In fact, past research suggests that it is the opinion side of the “wall of separation” that proves more persuasive. For example, Dalton, Beck, and Huckfeldt (1998) found that the slant of opinion columns and editorials predict readers’ candidate preferences more reliably than straight news articles. In a similar vein, other studies have found that opinion content and commentary are better predictors of issue opinions than straight news (Entman 1989; Page, Shapiro, and Dempsey 1987). Taking an experimental approach, Feldman (2011b) demonstrated that television opinion shows move public opinion in the direction of the host’s rhetoric.
This persuasive effect should be elevated by the common frames of reference invoked by coverage. Frames are persuasive because they make some considerations appear more relevant than others (Nelson, Clawson, and Oxley 1997). In addition, frames are more persuasive when they are strong, salient, and come from a trusted source (Chong and Druckman 2007; Druckman 2001). News shows might be less effective at framing because they present competing frames by, for example, playing clips of candidate speeches or featuring political analysts from each side of the aisle to discuss a political issue. This assures that viewers will be exposed to strong frames on each side of the issue, which makes each frame markedly less persuasive (Chong and Druckman 2007).
Beyond the one-sided coverage, there are other reasons to believe that opinion shows have a more significant effect on viewers than news shows. First, opinion show hosts act as opinion leaders by interpreting the news. Similar to opinion columns and editorials in newspapers, opinion shows provide cues that help viewers understand the political world. As Dalton, Beck, and Huckfeldt (1998, 124) stated, “More than just framing events, the press provides political cues that may significantly influence the opinions of readers.” The effects of such cues should be particularly pronounced given the credibility of these hosts. Prior research has found that like-minded opinion show hosts are perceived as credible by their viewers (Feldman 2011a). This added credibility increases the persuasive power of the opinion show host. For each of the reasons set forth above, we propose the following hypothesis:
Hypothesis 1: Opinion show coverage has a stronger effect on viewer attitudes than news shows.
More specifically, in the context of a presidential election, opinion shows are likely to have distinct effects on the out-party candidate as opposed to the in-party candidate. According to Jamieson and Cappella (2008), one of the strategies of partisan sources is to make the out-party candidate appear ideologically extreme. Consider the narrative of the 2008 presidential election: Fox News framed Barack Obama as an extremist outside the bounds of mainstream politics with ties to Bill Ayers, Muslims, and Rev. Jeremiah Wright. This attack mentality is evident in content analyses of election coverage showing that partisan news sources tend to spend more time attacking the opposition candidate than they do praising the in-party candidate. For example, Morris and Francia (2010) found that Fox News’ coverage of the 2004 Party conventions was more likely to attack the Democratic candidate (John Kerry) than to praise George W. Bush. Fox News painted John Kerry as an extreme Massachusetts liberal out of touch with the average American (Jamieson and Cappella 2008). It appears that the attacks on Kerry were successful, as Fox News viewers came to perceive him as more liberal following the Democratic and Republican conventions (Morris and Francia 2010). Meanwhile, watching Fox News had no effect on perceptions of Bush’s ideology. Given that opinion shows are free to serve as attack dogs on the opposition, we propose the following hypotheses:
Hypothesis 2a: Exposure to opinion show coverage increases the likelihood viewers perceive the out-party candidate to be ideologically extreme.
Hypothesis 2b: Exposure to opinion show coverage has no effect on viewer perceptions of the in-party candidate’s ideology.
We do not argue that opinion shows, or partisan news media more generally, have no effects on attitudes toward the in-party candidates. Instead, we argue that it is largely coverage of the out-party candidate that moves viewer attitudes. Attacking the opposition is likely to rally support around the one person (the in-party candidate) who can keep the out-party candidate from winning. Indeed, motivated reasoning suggests that people process information in ways that support what they want to believe (Kunda 1990). In this case, negative information about the opposition should illustrate why the other side is bad while also providing the impetus viewers need to feel good about their own candidate (Holbert, Garrett, and Gleason 2010). In addition, research on campaign advertising finds that negative messages are more powerful than their positive counterparts (Baumeister et al. 2001). For these reasons, negative messages about the out-party candidate should be very persuasive. In tandem with opinion shows’ effects on viewer perceptions of the out-party candidate’s ideological extremism (Hypotheses 2a and 2b), we also test the following hypotheses:
Hypothesis 3a: Exposure to opinion show coverage of the out-party candidate increases the likelihood viewers positively perceive the in-party candidate.
Hypothesis 3b: Exposure to opinion show coverage of the out-party candidate increases the likelihood viewers negatively perceive the out-party candidate.
Data and Method
In this study, we use panel and rolling cross-sectional (RCS) data from the 2008 National Annenberg Election Survey (NAES). The RCS data are a repeated cross section in which the cross-section design is spread out in a controlled way over time. The 2008 RCS was conducted from December 17, 2007 until November 3, 2008 and included 57,967 total adult respondents in the United States. To assure random selection, respondents were contacted by telephone using random-digit dialing technology. 2
The major weakness of the RCS design is that it does not capture attitude change within individuals. For this reason, we rely on web-based panel data for our individual-level analysis. Respondents were recruited from Knowledge Networks, which maintains a nationally representative sample. 3 There were a total of 28,985 participants in the online panel interviewed over five waves with each wave lasting several months, but we only use three of the waves in our analysis. The Spring wave (n = 17,747) includes interviews conducted during the party primaries from January 1 to March 31, 2008. The next round of interviewing is the Summer wave (n = 20,052) conducted from April 2 to August 29. Finally, the Fall wave (n = 19,241) was administered from August 29 until Election Day (November 4, 2008).
The dependent variables of interest are respondent attitudes toward the 2008 presidential candidates. Favorability is measured using a feeling thermometer that scores individual attitudes toward McCain or Obama from 0 (very unfavorable) to 100 (very favorable). We also include measures of respondent perceptions of the ideological leanings of the candidates. Respondents were asked to rate the ideology of Obama and McCain on a seven-point scale from extremely liberal to extremely conservative.
To measure exposure to television news sources, respondents were asked to indicate which programs they watched from a long list of news and entertainment shows. Specifically, respondents were asked the following question: “Which of the following programs do you watch regularly on television? Please check any that you watch at least once last month.” We used responses to create binary measures representing exposure to Broadcast news, CNN, and news and opinion programs on Fox News and MSNBC. Each variable indicates whether respondents watched at least one news (opinion) program on that station. To classify shows as news or opinion shows, we rely on the classification provided by Fox News in 2009 (Stelter 2009). 4 According to Fox News, news hours are from “9:00 a.m. to 4:00 p.m. and 6:00 to 8:00 p.m. on weekdays.” During the Spring wave, 50 percent of all Fox News viewers reported watching at least one opinion show, and this increased to 55 percent during the Fall wave. 5 Given that MSNBC only had three programs included in the panel survey list during the Spring wave, we include binary variables indicating that respondents watched: MSNBC Live, Hardball with Chris Matthews, and Countdown with Keith Olberman. It is important to note that including separate measures of each program on MSNBC might produce weaker effects than if we included all programs in the same variable, as we do with Fox News, CNN, and Broadcast. For this reason, we caution readers against comparing the effect of MSNBC with the effects of other programs. 6
In addition to news exposure, we use standard demographic variables, such as political interest, party identification, ideology, and the number of hours the respondent watched television. More detailed information about all of the variables used in this analysis can be found in the appendix.
Content Analysis
Content analysis data come from the Campaign Coverage Index (CCI), gathered by the Project for Excellence in Journalism (PEJ). The CCI is a subset of the larger News Coverage Index, but only includes stories that concerned the presidential campaign. 7 On a daily basis, the CCI coded all three nightly news programs on Broadcast and three out of the four evening programs on CNN and Fox News, but only two out of the four programs on MSNBC. The CCI also analyzed thirty-minute segments of daytime programming on two out of three networks each day. As this coding structure rotates on a daily basis, daily variations in content will result from the coding scheme. Therefore, it is best to use weekly measures of content because daily variations will average out over a week long period.
Our primary variable of interest is the candidate covered in each story. The CCI coded stories for the candidates discussed, including John McCain and Barack Obama. 8 There can be more than one lead newsmaker in a story, meaning that some stories discuss both McCain and Obama. We created separate variables measuring the number of stories that discussed McCain and the number of stories discussing Obama. These variables are used in our test of Hypotheses 3a and 3b that focus on coverage of in-party candidate versus out-party candidate.
For CNN and Broadcast news, we combined the content on all of the daytime and nightly news programs. For Fox News, we created separate measures for the news and opinion programming. Unfortunately, the CCI did not code the content for all programs. The only two opinion programs that were coded on a consistent basis were The O’Reilly Factor and Hannity and Colmes, and thus we use the content of those shows to represent opinion programming. The content of all other programs on Fox News was coded as news. 9 As for MSNBC, we classify Countdown and Maddow as opinion shows based on their content and format. According to the PEJ, those programs resemble a talk show in their extensive reliance on interviews. As for their content, Countdown and Maddow were unabashedly negative toward John McCain in 2008 (PEJ 2008). 10 Finally, CCI also coded for some characteristics of the story, including tone, which we use in Table 1. 11
Television News Coverage of Obama and McCain in 2008.
The data in this table come from the Campaign Coverage Index coded by the Project for Excellence in Journalism. Although the CCI data are publicly available, we were granted special access to the tone data. The top row of each section is the number of all stories from which the percentages were taken. Percentages were rounded to the nearest whole number. The Fox opinion category only includes content from The O’Reilly Factor and Hannity and Colmes, whereas the Fox News category includes all other content coded on Fox News. The MSNBC News category includes all content coded on MSNBC except for stories airing on Countdown with Keith Olberman and The Rachel Maddow Show. Meanwhile, the MSNBC Opinion category only includes stories airing on Countdown and Maddow. Stories were only included in the analysis if they discussed at least one of the two candidates. For the tone data, the missing percentages represent the amount of neutral coverage toward the candidates. The percentage of neutral stories can be found by subtracting the sum of positive and negative percentages from 100. Difference of means tests was performed to test significance *p < .05.
Time Series
We examine aggregate weekly attitudes toward McCain and Obama for Fox News viewers using time-series analysis. To that end, we combined the CCI content data with the RCS survey data collapsed over a forty-five-week period. The content measure includes an overall news measure, opinion coverage of McCain (Obama) on Fox, and news coverage of McCain (Obama) on Fox—each recorded as the number of stories per week with the candidate as the primary focus.
In the RCS, attitudes toward John McCain and Barack Obama were measured on a scale from 0 (very unfavorable) to 10 (very favorable). Using this measure, we created a weekly average of favorability toward McCain (Obama) for Fox News viewers. The RCS survey asked respondents to list their primary source of news—if and only if they reported watching television news. Only those selecting Fox News as their primary news source were included in the weekly averages of candidate favorability. The weekly average attitudes toward McCain (Obama) are then matched to the CCI content analysis data. Using these two sets of observations over time, we are able to investigate whether changes in the average attitudes toward McCain (Obama) for Fox viewers are affected by the previous week’s average coverage of the candidate on opinion or news shows on Fox. 12
Results
We have argued that the effects of coverage stem from opinion shows’ imbalanced coverage versus news shows coverage of both candidates and use of competing frames. Before we test our hypotheses, we first present descriptive statistics on patterns of coverage in opinion shows and news shows, by tone and by type of coverage, during the 2008 election to provide firm footing for our statistical analyses. To begin, we examine whether opinion shows devoted more attention to the opposing candidate or the in-party candidate. If opinion shows are more one-sided, they should be more likely to cover one candidate without mentioning his or her opponent: single-candidate focus. 13 Meanwhile, news shows should be more likely to discuss both candidates in the same story. Table 1 shows the percentage of all campaign-related stories from June 6, 2008—following Barack Obama clinching the Democratic nomination—to Election Day that focused exclusively on Barack Obama, John McCain, or discussed both candidates in the same story. 14 Although our primary purpose is to examine differences between news and opinion show coverage, we include Broadcast and CNN for comparison with the news programs on Fox and MSNBC.
The results are consistent with our expectations as opinion shows on both Fox News and MSNBC were more likely than news shows to devote exclusive coverage to the out-party candidate. Indeed, opinion shows on Fox News had 34 percent more exclusive Obama stories than exclusive McCain stories, whereas MSNBC’s opinion shows had slightly more exclusive stories about McCain than Obama. 15 However, the news programs on Fox and MSNBC had 16 and 14 percent more exclusive stories about Obama than McCain, which is nearly identical to the 14 percent on both CNN and Broadcast news. Thus, news programs were similar across news outlets, while opinion shows focused more on the opposition candidate.
Previous research leads us to expect there might be more variation in policy coverage between sources and between programming types. 16 For example, the Center for Media and Public Affairs found that Fox News was fairly balanced in their coverage of the presidential candidates, except for the pro-McCain bias in its policy coverage (Media Monitor 2009). In addition, news shows should be more inclined to discuss policy in terms of competing frames, whereas opinion shows are more likely to have a singular candidate focus in their policy coverage.
The results indicate that for policy coverage in particular, opinion shows are much more likely to focus exclusively on the opposing candidate. More than half of the policy-related stories on Fox’s opinion shows discussed Obama, but did not discuss McCain. Likewise, nearly half of the policy stories on MSNBC’s opinion shows discussed McCain but had little discussion of Obama. 17 Unlike opinion shows, news programs tended to discuss both candidates in the same story. The news programs on Fox and MSNBC were just as likely as CNN and Broadcast news to discuss both Obama and McCain in the same story. 18 These results are consistent with the different professional standards of news and entertainment (opinion) programming. With few exceptions, news coverage was uniform across partisan and nonpartisan outlets, reflecting the common journalistic standards of covering both candidates. Lacking such a standard, opinion shows are free to focus on the opposition candidate, while largely ignoring the in-party candidate.
Of course, discussing both candidates in one story does not equate to balanced coverage. Fox’s news programs might discuss both candidates, but in doing so, they bash Obama and praise McCain. For this reason, we compare the tone of coverage toward the candidates. 19 It is important to note that we make no claims regarding the bias of any news outlets or programs. Media bias must be considered in the context of real-world events, and it is not our purpose to establish an objective standard for media tone. 20 At most, any difference in tone across news outlets should be considered a relative bias—in that one outlet is more positive or negative toward a candidate than another outlet. We include coverage of CNN and Broadcast for comparison with news programs on partisan networks.
The results in the bottom portion of Table 1 show the percentage of the stories on each source that were positive or negative toward Barack Obama and John McCain. Compared with news coverage, opinion shows were more negative toward the opposing candidate. Fox’s opinion programming had 17 percent more negative stories than their news shows (p < .1). Meanwhile, MSNBC’s opinion shows had 40 percent more negative stories about McCain than its news coverage (p < .05). Indeed, the CCI did not code a single positive McCain story on either Countdown or Maddow for the entire coding period. 21
Interestingly, the opinion shows were slightly more negative to the in-party candidate than news shows, but the differences were not significant. Unlike the opinion shows on their networks, news programs on Fox and MSNBC were not overly negative toward the out-party candidates, at least relative to CNN and Broadcast news. Obama received the same net negative coverage (positive–negative) on Fox’s news programs that he received on CNN. Conversely, MSNBC’s news programs were 37 percent net negative toward McCain, which is identical to the net coverage he received on Broadcast news.
In sum, we point to three important findings regarding television news coverage of the 2008 presidential election. First, opinion shows were more likely to focus exclusively on the out-party candidate, whereas news programs tended to discuss both candidates in the same story. Second, opinion show coverage of the out-party candidate was more negative than news show coverage, but there was no difference between news and opinion show coverage of the in-party candidate. Third, the news programs on Fox and MSNBC were no more negative toward the out-party candidate than CNN or Broadcast news. Altogether, these results suggest that news coverage is similar across various sources, whereas opinion shows devote most of their attention to attacking the opposition.
Media Effects on Candidate Perceptions
How does Fox News affect viewers’ political attitudes? In this section, we examine the effects of Fox’s news and opinion shows on candidate favorability and perceptions of the candidate’s ideology. First, we address Hypothesis 1, which predicts that opinion shows will have a stronger effect on attitudes in comparison with news shows. Table 2 shows the results from ordinary least squares (OLS) regression models with each candidate’s favorability and perceived ideology as the dependent variables. 22 The first four columns predict candidate favorability on a scale from 0 (unfavorable) to 100 (favorable), whereas the last four columns predict respondent’s perceptions of the candidates’ ideology from 1 (extremely liberal) to 7 (extremely conservative). The column subheading indicates the wave in which the dependent variable was measured. All of the models control for respondents’ attitudes during a previous wave.
OLS Regression Predicting Candidate Favorability and Ideology.
OLS = ordinary least squares. The variables measuring television exposure are binary and indicate whether the respondent watched at least one program on the television station indicated. The Fox News variable indicates that the respondent watched any program on Fox News (excluding opinion shows), whereas the Fox Opinion variable denotes the respondent watched at least one opinion show in the last month. The Previous Wave variable is a lagged measure of the dependent variable measured during the previous wave of the panel survey. When the dependent variable was measured during the Summer (Fall) wave, the Previous Wave was measured during the Spring (Summer) wave of the survey. All independent variables (other than the lagged dependent variable) were measured during the Spring wave. Detailed information on all of the control variables used in these models can be found in the appendix.
p < .05.
The Fox Opinion variable indicates that viewers watched at least one opinion show, whereas the Fox News variable indicates that the respondent watched at least one news program. During the Spring wave of the survey, 39 percent of respondents indicated that they watched at least one show on Fox News, and 51 percent of the Fox News audience watched at least one opinion show. Overall, the results suggest that watching opinion shows has a stronger effect on viewer perceptions of the candidate than news shows. Indeed, opinion shows had a significant effect on viewer attitudes in all but one model in Table 2, whereas the Fox News variable was significant in only half of the models. Furthermore, opinion shows had a stronger effect on viewer attitudes than news shows in all but one model. Respondents watching Fox’s opinion shows were on average more than two points more favorable to McCain in the Summer (2.32) and Fall (2.20) waves. Conversely, Fox’s opinion show viewers were on average (3.16) markedly less favorable toward Obama during the Summer wave and also (1.79) less favorable during the Fall. Fox’s news show viewers were more favorable to McCain in both waves (0.832 in the Summer and 1.18 in the Fall), and (1.74) less favorable to Obama during the Summer wave. Thus, Fox’s news programs appear to affect viewer attitudes on the individual level, but Fox’s opinion shows have a stronger and more consistent effect. 23
There is also evidence that liberal opinion shows can affect viewer attitudes. Watching Countdown with Keith Olberman decreased favorability toward McCain during the Summer (2.82) and Fall (1.47) waves, but only increased favorability toward Obama during the Summer (3.13) wave. By comparison, MSNBC Live and Hardball with Chris Matthews had no discernable effect on viewer favorability toward either candidate. Altogether, these results support Hypothesis 1, as opinion shows on both Fox News and MSNBC had a stronger and more consistent effect than news shows.
We now turn to Hypotheses 2a and 2b, which predict the effects of watching opinion shows on viewer perceptions of candidate ideology. Again, we predict that exposure to opinion shows will make viewers more likely to think the out-party candidate is ideologically extreme, but will have minimal effects on perceptions of the in-party candidate. Specifically, watching Fox News should make viewers more likely to think Obama is extremely liberal, while exposure to Countdown should make viewers think McCain is extremely conservative.
The descriptive data support our expectations as 42 percent of the Fox opinion show audience believed Obama was extremely liberal in the Spring wave, whereas 58 percent of the same audience believed Obama was extremely liberal during the Fall wave. Meanwhile, only 12 percent of Countdown viewers thought McCain was extremely conservative during the Spring wave compared with 26 percent of viewers during the Fall wave. However, Fox viewers became only 2 percent more likely to think McCain was extremely conservative, while across the partisan aisle, Countdown viewers became only 2 percent more likely to label Obama as extremely liberal.
The results presented in the last four columns of Table 2 provide further support for Hypotheses 2a and 2b. Exposure to Fox’s opinion shows made viewers more likely to think Obama was liberal during the Summer and Fall waves of the survey. Watching Countdown and Hardball made viewers more likely to think McCain was conservative during the Summer wave but that effect disappeared during the Fall wave. Although watching Hardball did not have any effects on candidate favorability, it had a small effect on McCain’s perceived ideology. For the most part, these results support Hypothesis 2a, as watching opinion shows made viewers more likely to think the out-party candidate was ideologically extreme.
As for Hypothesis 2b, watching Fox’s opinion shows made viewers slightly more likely to think McCain was more liberal but the effect was small and limited to the Summer wave. On the other side, watching Countdown and Hardball had no relationship to viewer perceptions of Obama’s ideology. These results provide mixed evidence for Hypothesis 2b, as opinion shows had null effects on perceptions of the in-party candidate’s ideology in three out of four models. Interestingly, the audiences for Fox’s news programs and MSNBC Live were less likely to view the in-party candidate as ideologically extreme. Perhaps this results from the slightly more positive coverage given to in-party candidates on partisan news shows.
Overtime Analyses
We now turn to aggregate-level analysis to examine how variations in Fox’s coverage of the candidates influenced audience attitudes. Recall, we conduct a time-series analysis using weekly average attitudes toward Obama and McCain matched to CCI content data. First, we test Hypothesis 3a, which posits that opinion show coverage of the out-party candidate has significant effects on audience attitudes toward the like-minded candidate. Table 3 shows the results from a time-series analysis with Fox viewers’ attitudes toward McCain as the dependent variable and lagged coverage of Obama on Fox opinion shows and news shows as the independent variables. To account for variation due to coverage on other sources, we also include coverage of Obama for all other television sources. As the assumption of spatial and temporal independence is dubious with this sort of data, we estimate robust standard errors.
Fox Opinion Shows and Fox News Shows Coverage of Obama Predicts Attitudes toward McCain Over Time.
Robust standard errors are in parentheses. News and opinion coverage variables are lagged.
p < .05.
The previous week’s coverage of Obama on Fox News, for both opinion and news shows, significantly predicts current attitudes toward McCain over time. Coverage of Obama on opinion shows and news shows is positively related to attitudes, suggesting that as coverage of the out-party candidate increases, feelings toward the in-party candidate become more favorable over time. These results support Hypothesis 3a. Also interesting is the negative and significant relationship between overall coverage of Obama on other sources in the previous week and attitudes toward McCain. As coverage of Obama on sources other than Fox increases, favorability toward McCain decreases over time—the opposite effect of Fox coverage. Moreover, this negative relationship indicates that the effects of Fox coverage on viewer attitudes do not happen in a vacuum: as other sources cover Obama more one week, Fox viewers are less favorable toward McCain the next week. Still, this suggests a powerful and distinct effect for Fox coverage.
Next, we test Hypothesis 3b, the claim that coverage of the out-party candidate affects favorability toward the out-party candidate over time. Results are displayed in Table 4.
Fox Opinion Shows and Fox News Shows Coverage of Obama Predicts Attitudes toward Obama Over Time.
Robust standard errors are in parentheses. News and opinion coverage variables are lagged.
p < .05.
Again, the results support our hypotheses. The previous week’s coverage of Obama on Fox’s opinion and news shows significantly predict Fox News viewers’ attitudes toward Obama over time. As would be expected, this relationship is negative, which suggests that as coverage of Obama on Fox’s news and opinion shows increases, subsequent attitudes toward Obama become less favorable over time. We predicted that opinion show coverage would be both persuasive and significant for viewers’ attitudes toward the out-party candidate, and the data support our expectations. The results suggest that Fox weekly coverage of Obama on both news and opinion programs has a significant and distinct effect on audience attitudes the following week.
Overall, the overtime results suggest that both news and opinion content affect audience favorability toward the candidates. Similar to the individual-level results, both news and opinion shows had a significant effect, but the magnitude of the relationship between opinion coverage and attitudes is slightly larger. These results are consistent with Hypothesis 1 predicting that opinion shows have a larger effect on attitudes than news shows. 24 In addition, the significance of both news and opinion show coverage by candidate suggests that the amount of coverage of the other candidate matters regardless of program type.
Although not a direct test of our hypotheses, to better understand the effects of opinion coverage versus news coverage on Fox viewer attitudes, we also estimate time-series models that examine the overtime effects of McCain coverage. For the sake of brevity, results are included in the supplementary files. First, we estimate the effects of McCain coverage on Fox news programs, coverage on Fox opinion programs, and coverage of McCain on all other sources on Fox viewers’ attitudes toward McCain. Interestingly, the only significant coefficient in this model is coverage of McCain on all other networks. This positive and significant relationship suggests that as coverage of McCain on other sources increased, so too does favorability toward McCain. This relationship does not hold true for coverage of McCain on Fox news and opinion shows. Second, we estimate the effects of McCain coverage on Fox news programs, coverage on Fox opinion programs, and coverage of McCain on all other sources on Fox viewers’ attitudes toward Obama. Although none of the coefficients meet the threshold of significance, it is nonetheless noteworthy—both sets of results suggest that Fox coverage of McCain really has no effect on Fox viewers’ attitudes. In other words, when it comes to the in-party candidate, Fox coverage has null effects. Moreover, oddly enough, it is the coverage of McCain on other networks that has an effect on attitudes toward McCain for Fox viewers.
Discussion and Implications
An increasing number of Americans engage in selective exposure to ideologically consistent news sources. This fact has caused some scholars and pundits to worry that partisan news sources such as Fox News are making their audiences more polarized (Holbert, Garrett, and Gleason 2010; Morris 2007; Stroud 2011). Other scholars have pointed to the media’s diminished capacity for attitude change in a fragmented media environment and call for a new era of minimal media effects (Bennett and Iyengar 2008). We argue that tolling the death knell for media effects is premature, and we demonstrate that partisan media effects persist even in an era of selective exposure.
Whether one falls in the minimal versus maximal media effects camp, our results demonstrate that media effects occur in this new media environment. Fox News viewers became more favorable of McCain—and less favorable of Obama—over the course of the 2008 presidential election. This supports the growing body of research suggesting that partisan news sources have important effects on American public opinion (Barker 1999; Jamieson and Cappella 2008; Stroud 2011). More narrowly, our findings are consistent with past research suggesting that Fox News viewers, in particular, hold more conservative attitudes and beliefs than the rest of the population (DellaVigna and Kaplan 2007; Kull, Ramsay, and Lewis 2003; Morris 2007). We contribute to past research by including an exogenous measure of news content, allowing us to more appropriately test for media effects on audience attitude change over time.
Of greatest import are our findings on the dramatic effects of Fox opinion shows on individual-level attitudes. On the individual level, we find evidence that opinion shows have a more consistent effect on viewer attitudes than news shows. Watching Fox’s opinion shows made viewers more favorable to McCain and less favorable to Obama. Meanwhile, watching a liberal-leaning opinion show (Countdown) had the opposite effect, making viewers more favorable to Obama and less favorable to McCain. Compared with opinion shows, exposure to partisan news programs had a weaker and less consistent effect on viewer perceptions of the candidates.
When we investigate the effects of Fox’s news and opinion programs on viewer attitudes over time, the results are similar but show a slightly larger effect for opinion programs. We predicted that Fox viewer attitudes toward the out-party candidate and in-party candidate would be significantly affected by Fox opinion show coverage of the out-party candidate—and our results support this prediction. However, it is not just Fox’s opinion shows that matter, as Fox’s news coverage of the out-party candidate also causes attitude change over time. The overtime results are consistent with the individual-level findings showing that news programs have an effect on candidate favorability, albeit a weaker effect than opinion shows. Our results demonstrate that future research on media effects should take the type of programming consumers are exposed to into account. Moreover, results of content analyses will be more meaningful if they examine the effects of opinion content as distinct from the effect of straight news.
Our results also speak to how opinion shows affect viewer attitudes toward the candidates. For the most part, it is opinion show coverage of the out-party candidate that has the largest effect on viewer attitudes. Rather than praising the in-party candidate, opinion shows spend most of their time explaining why the out-party candidate is unacceptable. This attack mentality is not isolated to Fox News, as the liberal leaning Countdown with Keith Olberman was similarly likely to attack the out-party candidate. It appears that opinion shows aim to make the opposing candidate appear far outside of mainstream political discourse, and therefore unfit for the office (Jamieson and Cappella 2008).
Attacking the opposition appears to be an effective strategy for moving public opinion. Both individual and aggregate analysis suggest that coverage of the out-party candidate results in more substantial attitude change than coverage of the in-party candidate. On the individual level, watching Fox’s opinion shows made viewers more likely to think the opposing candidate was ideologically extreme—and this effect was evident for conservative and liberal opinion shows. Conversely, exposure to opinion shows had almost no influence on perceptions of the in-party candidate’s ideology. Our aggregate-level results tell a similar story as Fox News affects viewer attitudes of the in-party candidate by way of its coverage of the out-party candidate. Coverage of Obama not only affected attitudes toward Obama but also affected viewers’ attitudes toward McCain. Meanwhile, coverage of McCain had no influence on attitudes toward McCain or Obama. Thus, our results indicate that partisan news outlets have an effect on their viewers inasmuch as they cover the opposition candidate.
Several caveats are in order. First, our analyses present a snapshot of coverage and viewer attitudes during one presidential election, an election which was characterized by a unique political environment and even more unique candidate choices. Ideally, these results would be replicated across elections. Second, we focus on Fox almost exclusively and that diminishes our ability to generalize. Additional analysis of all partisan news outlets and all nonpartisan news outlets would strengthen our understanding of opinion show effects. Similarly, we use only a sample of coded opinion and news shows—future research should include content from the universe of Fox opinion and news programs, or even better, all television news and opinion programs. Third, although we theorize there are several reasons why out-party coverage matters more when it comes to media effects, additional experimental work is needed.
Despite these shortcomings, our results indicate that viewer attitudes toward the out-party candidate becomes increasingly negative as time goes on. Although it is natural to expect that partisans’ negative attitudes toward the opponent solidify over time, the implications for partisan news outlet’s coverage facilitating such negativity toward the other candidate are significant. First, it suggests that viewers’ attitudes are shaped predominantly by vitriolic coverage of the opposition. Rather than viewers consuming news from a source that confirms their support for their in-party candidate, viewers are consuming news from a source that seems to confirm their distaste for the opposition. The result is not only more polarized partisans but also partisans who are polarized because they hate the other side rather than because they are card-carrying supporters of their side.
Second, demonizing the opposition is likely to have important consequences for the winning candidate’s ability to govern. If a large segment of the base of either party believes the president is ideologically extreme, they are likely to attach that extreme label to any policies the president supports. Perhaps the rise of partisan news outlets, and consequently the increasing influence of opinion show hosts on the right and left, has contributed to polarization of the two political parties. It is possible that exposure to opinion shows on cable news and political talk radio increases the vitriol in modern politics, where any attempt to compromise is viewed as consorting with the enemy.
Finally, the results of our content analysis might relieve some fears that Fox News is blurring the lines between opinion and news. Our results suggest that Fox’s news programming is very similar in both format and tone to the news coverage on Broadcast, CNN, and even MSNBC. Of course, this does not mean Fox is not a partisan outlet. Our research focuses on just one domain of coverage and two measures of content so we are not suggesting Fox News can now join the ranks of public television. Moreover, our tone data only include a small slice of coverage, precluding us from drawing definitive conclusions about the tone of news and opinion content. Nonetheless, the results consistently show that news and opinion programs present different content to viewers. It is possible that news programs on Fox and MSNBC have subtle biases that affect viewer attitudes, but detecting any bias in news shows requires that future research separate news from opinion content.
The Fox News Network claims to have established a clear distinction between what it considers news programming versus opinion programming—and argues that the opinions expressed in opinion shows do not affect news reporting (Stelter 2009). This wall of separation between the news and editorial departments was a hallmark of news organizations for most of the twentieth century (West 2001). Fox News argues that this wall of separation still exists in its programming schedule. Furthermore, Fox News argues that the distinction between news and opinion is not lost on viewers because in their words, “The average consumer certainly knows the difference between the A section of the newspaper and the editorial page” (Stelter 2009). If viewers choose to watch an opinion show and are persuaded by the views expressed on that program, it is less insidious than outright manipulation by straight news. From this perspective, Fox News is similar to a newspaper that presents opinion content that leans overwhelmingly in one ideological direction (Dalton, Beck, and Huckfeldt 1998; Druckman and Parkin 2005). The difference is that Fox News has garnered more attention as of late, given its success and prominence on the national stage. Nonetheless, opinionated shows are a prominent part of the modern media landscape and do not appear to be going anywhere anytime soon. It is important that media scholars recognize this new force in American politics and account for the distinct content and effects of opinionated programming.
Footnotes
Appendix
Variable Descriptions
| Variables | Descriptions |
|---|---|
| Independent variables | |
| Education | Respondent’s formal education level measured on a nine-point scale from Grade 8 or lower to doctorate degree. |
| Age | Respondent’s age was categorized into four categories, including 18 to 35, 36 to 50, 51 to 64, and 65+. |
| Partisanship | Respondent’s partisanship measured via a seven-point variable ranging from 1 (strong Republican) to 7 (strong Democrat). |
| Ideology | Respondent’s ideological orientation; a seven-point variable ranging from 1 (extremely liberal) to 7 (extremely conservative). |
| Interest | Respondent’s self-reported interest in politics on a four-point scale ranging from 1 (not at all interested) to 4 (very interested). |
| Internet | Binary variable indicating whether the respondent reported getting campaign news from the Internet. |
| Radio | Binary variable indicating whether the respondent reported getting campaign news from the radio. |
| Newspaper | Binary variable indicating whether the respondent reported reading a newspaper for campaign news. |
| Hours TV | Number of hours (0–5) the respondent watched television the previous night. |
| Fox News | Variable indicates that the respondent watched any show on Fox News (excluding opinion shows). |
| Fox Opinion | Variable indicates that respondent watched at least one opinion show on Fox News, including The O’Reilly Factor, Hannity and Colmes, Hannity’s America, Fox & Friends, and Your World with Neil Cavuto. |
| CNN | Variable indicates that respondent watched at least one show on CNN. |
| Broadcast | Variable indicates that respondent watched at least one of the nightly news shows on network television, including NBC Nightly News, ABC News and World Report, and CBS News. |
| MSNBC Live | Variable indicates that respondent watched MSNBC Live during the Spring wave. |
| Countdown | Variable indicates that respondent watched Countdown with Keith Olberman during the Spring wave. |
| Hardball | Variable indicates that respondent watched Hardball with Chris Matthews during the Spring wave. |
| Dependent variables | |
| Candidate Favorability | Scores on a feeling thermometer scale ranging from 0 (very unfavorable) to 100 (very favorable). Variable was measured separately for McCain and Obama. For the overtime analysis, favorability was measured on a ten-point scale ranging from 0 (very unfavorable) to 10 (very favorable). |
| Candidate Ideology | Respondent’s perception of candidate’s ideology from 1 (extremely liberal) to 7 (extremely conservative). Variable was measured separately for McCain and Obama. |
Acknowledgements
We would like to thank the Annenberg Public Policy Center and the Project for Excellence in Journalism (PEJ) for the use of their data. We would also like to thank Matthew Eshbaugh-Soha and anonymous reviewers for their helpful comments.
Authors’ Note
A previous version of this article was presented at the 2011 meeting of the Western Political Science Association.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
