Introduction

In August 2006, political neophyte Ned Lamont scored a shocking victory over incumbent Senator Joseph Lieberman in the Connecticut Democratic primary election. Lieberman’s defeat seemed highly improbable. He was an 18-year incumbent who six years earlier had been his party’s vice presidential nominee on a ticket that won the national popular vote. Unlike most incumbents ousted in primaries, Lieberman was not implicated in a scandal, had trounced his prior opponent by a margin of nearly 2–1, and appeared ideologically compatible with his primary voters on most issues.

Instead, Lieberman’s primary defeat appeared largely to have resulted from his support for the Bush Administration—particularly on Iraq. A CBS/New York Times exit poll (CBS/NYT 2006) indicated that over three quarters of primary voters disapproved of the decision to go to war, and of those disapprovers, 60% voted for Lamont. Lamont himself argued prior to the election that Lieberman was “too likely to support the President, particularly on this war…It takes away from the Democratic voice” (Bacon 2006).Footnote 1

In this study, we examine how elite messages like Lieberman’s support of President Bush’s Iraq policies affect public opinion. In particular, we seek to explain why partisans might disproportionately fear criticism from their fellow party members while seeking praise from across the aisle. While prior research has given some basic intuition about the potential effects of different elite messages, we use national survey data and a media exposure experiment to determine exactly when and how public opinion is influenced by various partisan messages emanating from different sources and media outlets. Our core assumptions concerning the factors contributing to the persuasiveness of information are not novel. However, we offer more systematic tests than prior studies of several implications of these assumptions, at least some of which (e.g., with respect to partisan support for Iraq) are counterintuitive.

Our framework is general. However, we focus on war and related national security policies as a theoretically interesting and politically consequential application of our framework. In the former case, prior theories of public opinion and foreign policy have generally ignored the strategic incentives of media actors and their potential effects on the nature of the information upon which distinct subgroups of the public base their opinions, as well as on the persuasiveness of different types of elite rhetoric emanating from different media sources. In the latter, because typical Americans tend to know relatively little about foreign affairs (Holsti 2004)––and less than with respect to domestic policy (Edwards 1983; Sobel 1993; Canes-Wrone 2006)––they are particularly dependent upon elite cues in determining whether to support or oppose a presidential foreign policy initiative, like a military conflict. This makes individuals’ credibility assessments especially important in the realm of foreign policy.

We begin in the next section by explicating our framework. After briefly situating our argument in the literature, we present a typology of partisan messages in the news, determining the likely effects of these messages based on the characteristics of the speaker, listener, news organization, and the valence of the message itself. We examine how varying the party of the source and respondent, as well as the valence of the message and the identity of the media organization conveying it, influences the credibility, and hence the effectiveness, of such messages. We argue that the influence of partisan messages on viewers will depend upon whether: (a) the speaker shares the viewer’s party affiliation, (b) the message imposes some cost upon the speaker, and (c) the news outlet conveying the message is viewed as biased in favor of or against the message being conveyed.

To test for these effects, we conducted an experiment in which we exposed participants to a series of distinct partisan messages embedded into video and web text versions of edited news stories attributed to either CNN or the Fox News Channel (henceforth “FOX”). We subsequently investigated the treatments’ effects on our participants’ opinions regarding the president and the news stories they consumed. Finally, to bolster confidence in the external validity of our experimental results, we applied our analysis to national survey data examining public opinion related to the war in Iraq.

Partisan Cues and Public Opinion

Public Opinion and National Security

Politicians and pundits routinely assert that public support is vital for the success of national security policy, especially in the case of military conflicts. As former President Bill Clinton’s 1997 National Security Strategy document put it: “One…consideration regards the central role the American people rightfully play in how the United States wields its power abroad: the United States cannot long sustain a commitment without the support of the public” (NSC 1997). Similarly, President George W. Bush’s National Strategy for Victory in Iraq listed “Continued support of the American people” as one of six “conditions for victory” in the Iraq conflict (NSC 2005).

Yet the conditions under which the public will support a policy remain inadequately understood. The literature emphasizes either the public’s reliance on elite cues in the news (Brody 1991)––and particularly the degree of elite consensus surrounding a conflict (Brody 1991; Larson 2000)––in determining whether to support the president, or on the characteristics of the policies or conflicts themselves. The latter class of explanations emphasizes conflicts’ principal policy objectives (Jentleson 1992; Oneal et al. 1996; Eichenberg 2005), degrees of success (Feaver and Gelpi 2004; Kull and Ramsay 2001), or the numbers of and trends in U.S. casualties (Mueller 1973; Gartner and Segura 2000).

Throughout these theories, the public appears to engage in little, if any, evaluation of the content of public discourse. In the former literature, the public meekly buys the “spin” of politicians in the news, while in the latter they effectively ignore the news and focus on objective indicators, like body bags.Footnote 2 We argue that the public plays a more proactive role in deciding whether to support or oppose presidential foreign policy initiatives, actively reasoning about the content and credibility of the messages they receive in the media.

Of course, because most Americans know relatively little about foreign affairs (Holsti 2004), they are ill-equipped to independently assess the merits of a policy, especially in the short-term. Instead they rely on information shortcuts, or heuristic cues (Sniderman et al. 1991; Popkin 1994), most notably the opinions of trusted political elites whom they consider credible. Trust and credibility assessments, in turn, frequently hinge on one particularly accessible heuristic: party identification (Rahn 1993; Popkin 1994; Nelson and Garst 2005).Footnote 3

With relatively few partial exceptions (e.g., Zaller 1992; Morgan and Bickers 1992; Edwards and Swenson 1997; Baum 2002), most theoretical discussions of public opinion and foreign policy do not account for partisan differences in public opinion. Yet voluminous research (e.g., Lupia and McCubbins 1998; Druckman 2001a, b) shows that typical individuals are more responsive to information, such as being more susceptible to framing effects (Druckman 2001b), when they perceive the source as credible. The party affiliations of information sources (e.g., political and media elites) and receivers (citizens), in interaction with the content of the partisan messages themselves, thus can mediate the selection and implications of the information shortcuts typical individuals rely upon in making political judgments.

A Typology of Partisan Messages

We assume that the evaluative statements of partisans break down into four basic categories: (1) attacks on the other party (cross-party attacks), (2) support for one’s own party (intra-party praise), (3) support for the other party (cross-party praise), and (4) attacks on one’s own party (intra-party attacks).

Politicians expend considerable effort in seeking to shape their messages and images in the news media. The most universally accepted assumption in U.S. electoral politics is that politicians seek, first and foremost, re-election (Mayhew 1974). We generalize Mayhew’s famous observation by assuming that politicians seek re-election both for themselves and their fellow partisans. After all, winning a seat in the Congress holds dramatically different implications—both with respect to resources available for subsequent election campaigns, and for a member’s ability to influence public policy—if one is a member of the majority party (Cox and McCubbins 1993; Cox and Magar 1999). Winning election or majority party status, in turn, requires making one’s self and one’s fellow partisans look good, while casting the opposing party in a negative light. The implication for politicians’ preferences regarding media coverage is straightforward: typical politicians prefer stories that praise themselves and their fellow partisans, or criticize their opponents or the opposition party. Thus the parties will generally prefer to broadcast cross-party attacks and intra-party praise, while avoiding cross-party praise and intra-party attacks.

However, in determining each message type’s effect on viewers, it is important to note not just the content of the message itself, but also the credibility of the message or its speaker. Parties do not “inject” messages into a passive public; such messages are processed by individuals who accept or reject them depending in part on their perceived credibility (Sniderman et al. 1991; Kuklinski and Hurley 1994; Druckman 2001a). One source of credibility for a message is the belief that the speaker and listener have common interests (Crawford and Sobel 1982; Calvert 1985). This suggests that statements by a listener’s own party will be regarded as more credible than those of the opposing party, all else equal. Our first hypothesis follows:

H1: Partisan Credibility

Approval of the president among members of a given party will be more strongly influenced by presidential evaluations from their fellow partisans than by evaluations from members of the other party.

Another important source of credibility derives from the interaction of source and message: whether the message is costly to the speaker (Spence 1973). Typical individuals regard messages that are harmful to the interests of the speaker as more credible than those that impose no costs (so-called “cheap talk”).Footnote 4 In the context of partisan communication, messages by partisan speakers that appear to damage their own party or help the other party should be regarded as more credible than messages that help their own party or damage the other party. Such costly messages should be at least somewhat credible regardless of the party affiliation of the listener. Our second hypothesis follows:

H2: Costly Credibility

Evaluations that impose a cost on the speaker’s own party will have a stronger effect on individuals’ propensity to support the president than will equivalent “cheap talk” evaluations.

Table 1 summarizes the relative credibility of different partisan messages about the president based on their partisan and costly credibility for viewers of each party. It demonstrates the relatively weak persuasive power of “politics as usual” statements (i.e., intra-party praise or cross-party attacks). Such statements by members of the presidential (non-presidential) party serve only to rally their own followers, who probably already approved (disapproved) of the president prior to the statement (Baum 2002).

Table 1 Partisan and costly credibility, by party of speaker and viewer
Full size table

In contrast, non-presidential party praise should be exceptionally persuasive and beneficial to the president, especially among non-presidential party members. Similarly, if members of the president’s own party attack him, the effects on public opinion should be dramatic (but negative), especially among the president’s fellow partisans. In both cases, if available, the media demand for such stories virtually ensures they will receive coverage, further magnifying their potential impact on public opinion.

The Mediating Effect of the Press on Credibility

In recent years, as media have fragmented and some news outlets have begun to cater to partisan audience niches (Hamilton 2003), we argue the underlying preferences and routines of news organizations have shifted markedly. These changes have widened the gap between the true nature and extent of elite rhetoric and public perceptions of such rhetoric. (For empirical evidence in this regard, see Groeling and Baum 2008). While, for the most part, traditional journalistic norms and preferences have persevered, their applicability clearly varies across media outlets, particularly for the norm of offering balanced coverage (Tuchman 1972; Graber 1997)––that is, covering “both sides” in a story whenever possible. Increasingly, sophisticated and motivated consumers are able to seek out news sources —from cable news to partisan web sites to political talk radio—that reflect their own ideological preferences.

Recent research (Baum and Gussin 2008), in turn, suggests that media outlet labels, and the ideological reputations their “brand names” carry, serve as important judgmental heuristic cues which consumers employ to help interpret both the meanings and implications of partisan messages in the media. As a consequence, we argue that the nature of the media’s influence on policy has evolved from what scholars often refer to as the “CNN Effect,” which emphasized the importance of the 24-hour news cycle and live coverage of events, to what we refer to as an emerging “FOX Effect.” The latter effect concerns the implications of perceived partisan favoritism, combined with the effects of self-selection and credibility-based discounting by audiences.

Table 2 disaggregates the expected credibility of messages attacking and praising a Republican President (like President Bush) based on the perceived partisan leanings of the network airing the story. As before, credibility stems from perceived partisan common interest, combined with the costliness of the statement to each party’s perceived interests. We anticipate that perceived credibility enhances the persuasive power of a message. Consequently, we expect that a Congressional Democratic criticism of a Republican president will be more persuasive to viewers if it appears on a network they consider conservative––and hence sharing the interests of the Republican president––than if the same message appears on a network they consider liberal––and hence not sharing the Republican president’s interests.

Table 2 Credibility impact of network attribution
Full size table

Assuming Republicans think they have common interests with networks they consider conservative, while Democrats perceive common interests with networks they perceive as liberal, the patterns in Table 2 suggest the following hypothesis:

H3: Partisan Media

Statements critical of a conservative (liberal) president should be more credible to viewers when they appear on a news source perceived as conservative (liberal), relative to a news source perceived as liberal (conservative). Conversely, statements praising a conservative (liberal) president should be more credible on a news source perceived as liberal (conservative), relative to a news source perceived as conservative (liberal).

If, as predicted by H3, viewers find rhetoric perceived as costly for a given news outlet (i.e., contrary to the outlet’s perceived partisan leaning) more valuable and persuasive than other rhetoric, one empirical manifestation ought to be a relatively greater propensity to discount, or counter-argue (e.g., criticize as “biased”), “cheap talk” rhetoric (i.e., supportive of the outlet’s perceived ideological leaning). This suggests another hypothesis.

H4: Selective Acceptance

Individuals will be more critical of statements opposing a conservative president from sources they perceive as liberal, relative to the same news from sources they perceive as conservative. Conversely, they will be less critical of statements supporting a conservative president from sources they perceive as liberal, relative to the same news from sources they perceive as conservative.

Experimental Examination of Message Effects

Design

We first test our predictions through an online experiment designed to explore the effects of intra- and inter-party attacks on and praise of the president (2 evaluation sources × 2 evaluation types) attributed to either FOX or CNN (2 networks). This yields a total of 8 possible treatments.Footnote 5 The treatments consisted of a streaming video regarding the NSA domestic spying scandal, followed by a static web text report on the war in Iraq.Footnote 6

As with any media exposure experiment, we faced a tradeoff between external validity and greater control. For our comparisons of the effects of network reputations, we were able to maximize control by testing the effects of treatments that were precisely identical across networks––except for the relevant network-identifying information.Footnote 7 However, because we have already demonstrated the opinion effects of partisan rhetoric elsewhere (Groeling and Baum 2008, Baum and Groeling forthcoming), we elected to sacrifice some degree of experimental control in testing our rhetoric hypotheses in favor of enhancing external validity in our video treatments. That said, in a separate pilot study, we found no significant differences in mean thermometer scores for the figures cited in each party’s praise and criticism treatments (video or text). As one would expect, respondents from each party rated their own partisans more highly than figures from the other party, but these differences were symmetric across the parties. In other words, Republican identifiers rated Republican members of Congress (MCs) at about the same level as Democrats rated Democratic MCs, while Republicans rated Democratic MCs about the same as Democrats rated Republican MCs.Footnote 8 After watching and reading the video and text stories, participants filled out a survey asking them to indicate which aspect of the news reports they found most interesting, and answered some questions about their political attitudes. (The full texts of the webpage and video treatments are available at http://www.sscnet.ucla.edu/comm/groeling/warstories/Welcome.html.)

Our experiment included 1610 participants drawn from UCLA communication studies (55%) and political science (45%) courses taught between Spring 2006 and Winter 2007. We offered modest extra credit for participating. Twenty-one percent identified themselves as Republicans (including leaners), while 53% identified as Democrats (including leaners). Independents and third-party identifiers accounted for the remaining 24% of our participants. Table 3 presents our population characteristics for the overall sample, as well as for Republican and Democratic sub-samples.

Table 3 Summary of participants’ characteristics (means and standard deviations)
Full size table

Of course, some research (Sears 1986) has famously called into question the generalizability of experimental findings based on student population samples. As noted, we are cognizant of this concern and have sought to enhance our confidence in the generalizability of our findings by subsequently applying our analysis to a national survey. It is also worth noting that we have successfully replicated several variants of this experiment across numerous distinct student population samples between 1999 and 2007, drawn from eight different universities that, in total, include every region of the country. Consequently, while we cannot discount entirely the possibility that any student sample is inherently biased, we can confidently conclude that our results are highly robust across numerous distinct student population samples with quite distinct characteristics.Footnote 9

Key Variables

The main dependent variable for this experiment is participants’ approval of President Bush’s handling of national security (“Do you approve or disapprove of the way George W. Bush is handling national security?”), which we employ to test H1–H3. To test H4 (selective acceptance), we measure whether participants criticized the balance of the stories they viewed in response to an open-ended question asking: “What did you find most interesting about either or both of the news reports you just watched and read?”

We employ national security approval, rather than overall presidential approval, for two reasons. First, we specifically selected our treatments to target the national security domain of politics. Consequently, we anticipate that the treatment conditions should primarily influence participants’ attitudes in this area. Second, at the time of the experiment, President Bush’s approval ratings were well below 40%. Among Democrats in our data, less than 5% indicated that they approved of the President’s overall job performance. This creates a significant floor effect; criticism of the president, however credible, could not significantly erode the president’s approval among our Democratic participants, nearly all of whom already disapproved. While most Democrats also disapproved of the President’s handling of national security, his approval rating in this area among Democrats was nonetheless over twice as high, at about 10%.Footnote 10 This leaves somewhat more room for any hypothesized negative effects of credible criticism to emerge.

For a similar reason, we also employ an expanded version of the approval question, which distinguishes between “strongly” and “somewhat” approving or disapproving, as well as permitting a response of “neither approve nor disapprove.” This allows us to observe treatment effects that a more blunt “approve vs. disapprove” question might obscure. For the Criticize Balance dependent variable, in turn, we created a dummy, coded 1 if the respondent explicitly criticized the ideological balance of the treatment to which they were exposed (in the open-ended question), and 0 otherwise.

Our main independent variables are dummies indicating which of the eight treatment conditions participants viewed. We modeled these variables as interactions between three variables: Negative (scored 1 if the treatment criticized the Bush administration), Republican Source (scored 1 if the MCs who appeared in the stories were Republicans), and, to allow us to test differences in partisan credibility, the participant’s own party affiliation. In the latter case, we created dummy variables for Republicans and Democrats, including leaners, and also for non-leaning Independents, including third-party members.Footnote 11

While random assignment ought, theoretically, to account for many potentially confounding causal factors, our student population sample differs systematically from a truly random population sample in several important ways (e.g., partisan leanings and ethnicity). Moreover, it is always possible to draw a systematically biased sample even when drawing at random, especially given relatively small treatment groups. For both reasons, we add several control variables intended to account for these systematic differences. These include ideology, campaign interest, whether participant were enrolled in a communication or political science course, ethnicity, a 10-point index of political knowledge (defined in Appendix A), age, and participants’ assessments of the ideological orientations of FOX and CNN on a liberal-to-conservative scale (from the pretest).Footnote 12

Results

We begin with tests of the Partisan (H1) and Costly (H2) Credibility Hypotheses. Model 1 in Table 4 presents our tests of these hypotheses.Footnote 13 Because our analysis examines the impact of four different types of rhetoric on three different types of partisan viewers’ decisions to approve or disapprove on a five-point scale, in Fig. 1 we employ statistical simulation software (King et al. 2000) to transform the key coefficients into probabilities of approving of President Bush’s handling of national security and chart the effects of moving from one type of treatment message to the next.

Table 4 Ordered logit and logit analyses of the correlates of approving the president’s handling of national security and criticizing the ideological balance in news story content
Full size table
Fig. 1
figure 1

Probability of disapproving of president Bush’s handling of national security, as message source and valence vary (Error bars denote 95% confidence intervals)

Full size image

Beginning with H1 (Partisan Credibility), we first compare the effects of moving from the Democratic Criticism to Democratic Praise treatments among Democratic identifiers with the corresponding changes among Republican identifiers. H1 would predict that the former effect should be larger and more significant, which is in fact what we find. In the former case, moving from criticism to praise by Democratic members of Congress (MCs) yields a 16.2 percentage point decrease in the probability of disapproving of President Bush’s handling of national security (combining “strong” and “somewhat” disapprove), a 6.7 percentage point increase in the probability of neither approving nor disapproving and a 9.1 percentage point increase in the probability of approving (p < .01 in each case).Footnote 14 In contrast, among Republican the corresponding effects are small and statistically insignificant.

Similarly, Fig. 1 also indicates that, as predicted by H1, shifting from Republican Criticism to Republican Praise affects the approval of Republicans far more than Democrats. For Republicans, this shift is associated with nearly a 22-percentage point drop in the probability of disapproving (compared to only 6 percentage points for Democrats), while the probability that Republicans will approve of the president’s performance jumps 20 percentage points (compared to only 3.3 percentage points for Democrats). While both the Republican differences are statistically significant (p < .01), the Democratic differences are, as predicted, insignificant.

The data in Fig. 1 also facilitate a further test for partisan credibility effects by allowing us to hold the message content constant while varying the party. In this instance, we anticipate that participants will view rhetoric by their fellow partisan elites as more credible, all else equal. However, it is important to note that in some cases, costly and partisan rhetoric will conflict, thereby presumably weakening the results.

As predicted by H1, shifting from Republican to Democratic praise––where partisan and costly credibility are not in conflict––decreases by 6.7 percentage points (p < .10) the probability that Democratic participants will disapprove of the president’s national security performance, while increasing by 4 points (p < .10) the probability that they will approve. Presumably because costly and partisan credibility conflict for Democrats when moving from Democratic to Republican criticism, this latter shift is associated with smaller and statistically insignificant effects.

Among Republicans, and also consistent with H1, shifting from Democratic to Republican criticism––where partisan and costly credibility do not conflict––is associated with about a 12 (10) percentage point increase (decrease) in the probability of disapproving (approving) (p < .10 in both cases). Conversely, shifting from Democratic to Republican praise––where partisan and costly credibility do conflict––has no significant effect on approval ratings.

Turning next to H2 (Costly Credibility), we compare the effects of moving from “cheap” (Democratic) criticism to “cheap” (Republican) praise with the effects of moving from “costly” (Republican) criticism to “costly” (Democratic) praise. In this instance, we focus our analysis on Independents, for whom partisan credibility presumably plays no offsetting role in credibility assessments, thereby allowing us to isolate the effects of costly credibility.

The results once again strongly support the hypothesis. Among Independents, moving from costly—Republican—criticism to costly—Democratic—praise is associated with about a 14 (8) percentage point decrease (increase) in the probability of disapproving (approving) (p < .01 in both cases). Conversely, moving from cheap—Democratic—criticism to cheap—Republican—praise is associated with a small and statistically insignificant effect on approval ratings.

We next investigate our media-outlet hypotheses, where, as noted, variations in our treatment conditions are more precisely controlled. We begin with the Partisan Media Hypothesis (H3), which predicts that, due to its relatively greater costly credibility, viewers will find criticism (praise) of President Bush more credible when it appears on a news source they perceive as conservative (liberal). Model 2 in Table 4 tests this hypothesis. Table 5 converts the coefficients into probabilities, as well as assessing the magnitude and significance of the difference in support for President Bush’s handling of national security, as a given message moves from a liberal to a conservative network. In the top half of Fig. 2, in turn, we separately plot the probabilities of disapproving (strongly or weakly) among participants exposed to costly or cheap rhetoric.Footnote 15

Table 5 Effects of perceived outlet ideology on probability of approving of president’s handling of national security, as treatment varies from liberal to conservative network
Full size table
Fig. 2
figure 2

Probability of disapproving of president Bush’s handling of national security as news source and message valence vary (** p < .01, * p < .05, ^ p < .10)

Full size image

The results in Table 5 and Fig. 2 strongly support our hypothesis. Costly outlet communication mattered far more than cheap talk. Among participants exposed to cheap talk (any praise on a conservative network or any criticism on a liberal network) moving from the praise to criticism conditions is associated with a small and insignificant effect on disapproval of the president’s handling of national security. The corresponding effect among participants exposed to costly communication (moving from praise on a liberal network to criticism on a conservative network) is a highly significant (p < .01) 30 percentage point increase in the probability of disapproving (strongly or weakly), from .47 to .77. In other words, as predicted, the ideological reputations of the networks mediate the persuasive power of the information they present to consumers. Messages perceived by our participants as running counter to the perceived ideological interests of the outlets to which they were exposed had a far greater effect on their attitudes toward the president than messages perceived as self-serving for the networks, given their presumed ideological orientations.

Finally, we turn to H4 (Selective Acceptance), which holds that, all else equal, people are more prone to critically evaluate, or counter-argue, information perceived as supportive of a news outlet’s presumed ideological orientation (cheap talk), relative to information that challenges an outlet’s presumed orientation (costly talk). Model 3 in Table 4 presents the results of our test of this hypothesis. In the bottom half of Fig. 2, we again transform the key coefficients into predicted probabilities that participants criticized the experiment’s news stories as biased.

Once again, the results strongly support our hypothesis. Participants were 12 percentage points more likely to criticize rhetoric critical of the president when it appeared on a network that they considered to be liberal, relative to the identical rhetoric appearing on a network perceived to be conservative (.34 vs. .22, p < .10). Conversely, participants were 17 percentage points less likely to criticize rhetoric supportive of the president when it appeared on a network they perceived as liberal, relative to the same rhetoric on a network perceived as conservative (.18 vs. .35, p < .05). Interestingly, looking across the criticism and praise treatments, we see that the probabilities of criticizing both types of cheap talk are nearly identical––.34 for criticism on a liberal network and .35 for praise on a conservative network––as are the probabilities of criticizing costly talk––.22 for criticism on a conservative network and .19 for praise on a liberal network. This strongly suggests that viewer credibility assessments derive in significant measure from ex ante assumptions regarding the ideological orientations of news outlets. Such assessments, in turn, appear to heavily influence consumers’ propensity to counter-argue different types of rhetoric (cheap vs. costly talk).

News Consumption and Attitudes toward Iraq

We turn next to an empirical investigation of national public opinion regarding the Iraq war. The goal is to determine whether the patterns that emerged in our experiment generalize to a real-world context. Specifically, we investigate the effects of the credibility assessments of different types of consumers (Democrats and Republicans) vis-à-vis different media outlets (FOX vs. CNN) on attitudes toward the war. This investigation tests the external validity of our theoretical framework, and, in particular––given the distinct perceived partisan leanings of FOX and CNN––the effects of partisan media (H3) and selective acceptance (H4).

For our dependent variable in this analysis, we employ the following question from a June 2005 survey (Pew Center 2005): “How well is the U.S. military effort in Iraq going?” (coded 1 = very or fairly well, and 0 = not too well or not at all well).Footnote 16 We compare responses to this question across individuals with different partisan affiliations who claim to get most of their news about politics and international affairs from CNN or FOX.Footnote 17

Unlike our experiment, in this survey we have no way to determine precisely what information FOX or CNN viewers actually consumed. Fortunately, we can derive some insight from a Project for Excellence in Journalism (PEJ) study (PEJ 2005), which content analyzed cable news coverage of the war in Iraq in the year preceding the Pew survey. PEJ found that FOX was nearly twice as likely as CNN to air stories with an “overwhelmingly” positive tone (38% of war-related stories for FOX, vs. 20% for CNN). Conversely, CNN aired nearly twice as many segments as FOX with negative tones (23% vs. 14%). Overall, CNN aired slightly more negative than positive segments, while FOX aired more than twice as many positive as negative segments.Footnote 18 The overtly pro-conservative Media Research Center similarly found, in its own study, that between May 15 and July 21, 2006, Fox aired nearly twice as many stories about successes in Iraq as CNN and MSNBC combined (McCormack et al. 2006).

The implications of our model depend also on consumer perceptions. In fact, evidence suggests Americans are polarized in their opinions regarding the ideological slant of FOX to a greater extent than with respect to CNN. For instance, one survey (Pew Center 2006) found a much larger gap between liberals and conservatives in rating the believability of FOX, relative to CNN. Liberal and conservative respondents differed by only three percentage points in their probabilities of saying they believe “all or most” of the news on CNN (28 vs. 25%, respectively). The gap for FOX was over five times larger (16 vs. 32%). In other words, liberals rate CNN as similarly believable as conservatives rate FOX (28 vs. 32%): yet, liberals rate FOX as considerably less believable than conservatives rate CNN (16 vs. 25%).

If FOX is indeed more likely than CNN to praise the war, the implications differ for respondents with different partisan affiliations. For Democrats, pro-war news on FOX will likely be dismissed as non-credible, while equivalent information on CNN will likely be accepted as reliable. Conversely, among Republicans, exposure to the disproportionately-positive war news on FOX will be more strongly associated with believing things are going well in Iraq, relative to consuming CNN’s near-equal mix of praise and criticism. Moreover, given that liberals (who are relatively more likely to be Democrats) are more skeptical of FOX than conservatives (who are more likely to be Republicans) are of CNN, the implication is that liberals (and Democrats) are more likely to discount pro-war content on FOX than conservatives (and Republicans) are to dismiss anti-war content on CNN. This suggests two Iraq-specific hypotheses, both of which follow from our more general Partisan Media (H3) and Selective Acceptance (H4) Hypotheses:

H5

Democrats who rely on CNN are more likely than their FOX-watching counterparts to believe the war in Iraq is going well.

H6

Republicans who rely on FOX are more likely than their CNN-watching counterparts to believe the war in Iraq is going well.

Differences in the characteristics of respondents who choose to watch FOX and CNN, rather than––or in addition to––the content of news to which they are exposed, could be driving any observed relationship between outlet preferences and war attitudes. To some extent, we anticipate this is the case; our model presumes that partisans will take advantage of the opportunity to self-select into friendly environments. In fact, there is evidence of such a pattern. Democratic and Republican FOX watchers in this survey are more conservative than their counterparts who prefer CNN (by .42 and .46 points on the 5-point ideology scale, for Democrats and Republicans, respectively). To further assess this possibility, we investigate whether, and to what extent, knowledge about or attitudes toward the war, net to other factors, influence the decision to watch either network. We begin with FOX, shown in Models 1-4 in Table 6.

Table 6 Logit Analyses of correlates of relying on FOX or CNN as primary source of political and international news
Full size table

Model 1 in Table 6 includes dummy variables for whether the respondent is liberal, conservative, Republican or Democrat. The results indicate that Republicans and conservatives are significantly more likely to rely on FOX, while being liberal or a Democrat has no significant effect. In Model 2, we add three dummies to our base model, derived from questions asking respondents about the appropriateness of the Democratic and Republican party positions on: (1) national security and foreign policy, (2) economic policies (e.g., taxes), and (3) social issues (e.g., abortion). For each issue area, we subtracted the Democratic from the Republican “appropriateness” variable, to yield a scale measuring whether the respondent thinks one or the other party has a relatively more correct position on the issue. (See Appendix B for question wording and coding of key casual variables.) The results indicate that only the economic policy item significantly influences the tendency to prefer FOX.Footnote 19

In Model 3 of Table 6, we add several indicators of attitudes toward the media. While attitudes regarding the overall quality of the news media have no effect, nearly all of the specific media outlet favorability indicators are significant. Believing the news media are too critical of President Bush also positively influences the propensity to prefer FOX, though the coefficient is insignificant. None of the relative partisan issue correctness variables remain significant.

Finally, in Model 4, we drop the highly insignificant news quality scale and add several indicators of knowledge about, and attitudes toward, the war in Iraq. These include dummies measuring whether the respondent: (1) reports following the Iraq conflict (fairly or very) closely (2) knows the approximate number of U.S. casualties, and (3) believes invading Iraq was the right thing to do. The third question represents the best available indicator of retrospective evaluations of the conflict, as distinct from contemporaneous estimates of the war’s progress.Footnote 20 In fact, none of these indicators significantly influence respondents’ propensity to prefer FOX; as before, neither does preferring the Republican Party’s policies regarding national security and foreign affairs.

We replicated Model 4 (not shown), dropping the Iraq and foreign policy questions, except whether or not invading Iraq was the right thing to do. The Iraq attitude indicator remained insignificant. In other words, once attitudes toward the news media and general partisanship and ideology are controlled, neither a general preference for Republican policies on national security/foreign affairs nor knowledge of and attitudes toward the war significantly predict respondents’ propensity to prefer FOX for news about politics and international affairs. These results indicate that attitudes toward the mainstream news media and partisan/ideological predispositions mediate respondents’ propensity to prefer FOX far more than do attitudes regarding Iraq.

Turning to CNN, Models 5–8 in Table 6 replicate Models 1–4, with preference for CNN as the dependent variable. The key results are similar to those for FOX. Hence, we do not discuss the CNN results in detail. There are, however, several differences. Most notably, interest in Iraq is positively associated with preferring CNN, while being liberal is negatively related to a preference for CNN. Attitudes about the overall quality of the mainstream news media here matter more than specific outlet favorability ratings. However, factual knowledge about and attitudes toward Iraq have no discernable effect on propensity to watch CNN. Overall, the Pseudo R 2 values on the fully specified models (4 and 8) suggest that the causal variables offer only about half as much predictive power for CNN as for FOX. Finally, comparing Pseudo R 2 values across Models 3 and 4 for FOX, and Models 7 and 8 for CNN, indicates that the Iraq knowledge and attitudes items add little explanatory power to the models (about +.005 for FOX and +.01 for CNN).

In light of these results, we tentatively conclude that selection effects based on attitudes toward Iraq––the key potential selection effect for our purposes––do not appear to be fundamentally driving the decision to consume FOX or CNN. However, the question remains as to whether the information encountered by viewers, in interaction with their partisan predispositions, influences their attitudes toward Iraq. To test Hypotheses 5 and 6, we interact respondents’ partisan affiliations with their preferred sources of news about politics and international affairs. We also include a standard battery of demographic and political control variables, as well as controls for overall trust in the news media and interest in the war.Footnote 21 Table 7 reports the results of a logit model testing the effects of news preferences and partisan affiliation on attitudes toward the Iraq War.Footnote 22 In Fig. 3, we transform the results into probabilities that the respondent believes things are “going well” in Iraq as the primary source of news about politics and international affairs varies from CNN to FOX.

Table 7 Logit analysis of likelihood of believing the conflict in Iraq is “going well,” as news source and party identification vary
Full size table
Fig. 3
figure 3

Probability of believing the Iraq conflict is “going well” as source of war news and party identification vary (** p < .01, * p < .05; Error bars denote 95% confidence intervals)

Full size image

Consistent with most polling data (Jacobson 2006), Republicans in this survey are far more supportive of the war than Democrats. However, after controlling for a variety of correlates of attitudes toward Iraq, no statistically significant difference emerges between Democrats and Republicans who prefer CNN. In contrast, Republican FOX viewers are 54 percentage points more likely than Democrats to believe the Iraq war is going well (.83 vs. .29, p < .01).

Testing our hypotheses requires comparing differences within partisan groups. Given the PEJ analysis of FOX and CNN news coverage, these results indicate that Democrats watching mostly positive coverage of the war on FOX actually decreased their assessment of the war’s progress by a highly-significant 28 percentage points, relative to their peers who consumed relatively balanced coverage on CNN (from .52 to .24 for CNN and FOX viewers, respectively, p < .01). In other words, Democrats who rely on CNN are 28 percentage points more likely than their FOX-watching counterparts to believe the Iraq war is going well. This supports H5. Republicans, on the other hand, were more strongly influenced by FOX’s relatively positive coverage, increasing their assessment of the war’s progress by 24 points (from .57 to .81 for CNN and FOX viewers, respectively, p < .05). Stated differently, Republicans relying on FOX are 24 points more likely than their CNN-watching counterparts to believe the war is going well. These results support H6.

Notwithstanding our prior exploration of the correlates of preferring FOX and CNN, we cannot rule out entirely the possibility that the moderate ideological difference between Republican CNN and FOX watchers might help account for at least part of the greater optimism regarding the war among Republican FOX watchers. However, self-selection based on ideological affinity cannot account for lower optimism among the somewhat more conservative (relative to their CNN-watching counterparts) Democratic FOX watchers. This latter pattern is precisely the opposite of what one would predict if differences in the ideological preferences of CNN and FOX viewers were driving our results. Yet it is precisely what one would predict as a consequence of partisan credibility and selective acceptance. We thus conclude that while selection effects doubtless matter, so too do differences in the information to which viewers of FOX and CNN are exposed, mediated by their partisan predispositions to accept or reject messages with particular valences.

Additionally, our empirical results suggest the possibility that credibility effects might vary across categories of actors. We found stronger support for our partisan credibility predictions when the speakers were themselves partisan, and stronger support for the costly credibility predictions among news organizations. While some of these differences might be artifacts of the statistical modeling required to test the predictions (see footnote 15), it also seems plausible that the shared interest assumption underlying partisan credibility would apply more strongly to people publicly labeling themselves as fellow partisans, relative to media outlets that are merely viewed as sympathetic to the party (especially if those outlets deny such sympathies). Similarly, it seems possible that the public would place somewhat less weight on the credibility of statements made by office-seeking politicians relative to those of media outlets, for whom credibility can literally be a matter of corporate life and death in the news marketplace. Finally, our focus on the war in Iraq––the subject of one of the most bitter and contentious partisan struggles in recent memory––may have reduced the likelihood that partisans in the electorate would attend to any statements from the opposing party, even if those statements were supportive of their personal views.

Conclusion

Scholars have a clear understanding of the concept and implications of party with respect to legislative and voting behavior. Increasingly, however, parties have become more concerned about the collective image they present to the public through the media. As a consequence, as Joe Lieberman discovered, parties are growing increasingly aggressive in their attempts to foster or enforce such unity. Our findings suggest that politicians are justified in being concerned not just about what they do, but also what they say. While much rhetoric in the public domain is rightly characterized as “cheap talk,” a party’s messages (and those of the opposing party) do have tremendous potential to affect public opinion. In many cases politicians’ messages will be lost in the modern media maelstrom. Yet we find that relatively subtle partisan messages can have large effects on opinion, even in high-salience issue areas like war and national security, and among well-informed, politically-attentive partisans on the lookout for political manipulation and bias. The task for parties is made even more difficult by the fact that journalists regard “off message” partisan statements as almost inherently more newsworthy than cohesive messages. Parties can therefore generally count on having any damage associated with such messages magnified through extensive media coverage (Groeling 2001; Groeling and Baum 2008; Baum and Groeling forthcoming).

However, the news media environment itself is clearly changing. For instance, regardless of whether CNN or FOX actually favor a particular party, the public’s increasing belief that they––particularly FOX––do so has important implications for partisan communication. As we saw in our tests of Hypotheses 3–6, ascriptions of partisanship on the part of news media strongly influence which partisan messages the public regards as credible on those media.

As noted, our experimental evidence on this point is derived from a student population sample. These findings should thus be viewed as highly suggestive, yet not definitive, and hence in our view warranting further research in this area across broader populations. Assuming our findings prove robust––as our survey data suggest is likely to be the case––for politicians attempting to influence public opinion, the contrast to most of the television era is indeed stark. New media perceived as siding with a particular party will actually be less persuasive for all, save members of the same party, in communicating anything short of attacks against that same party. Conversely, stories communicating bipartisan support reported by a “hostile” media outlet should be one of the few positive messages that remain credible to partisans from both parties.

It is therefore unsurprising that politicians have increasingly worked to shape how the public perceives different news outlets. Republican candidates have famously argued that the media as a whole are biased against their candidates, perhaps best exemplified by a popular 1992 bumper sticker saying, “Annoy the Media: Re-elect George Bush.” However, with the rise of FOX, Democrats have mounted a specific, targeted attempt to marginalize and challenge the legitimacy of what they argue is a pro-Republican news outlet. For instance, in early 2007, liberal activists pressured the Nevada Democratic Party to cancel a FOX-sponsored Democratic candidate debate. In launching the successful campaign to drop FOX as a debate sponsor, liberal blogger Chris Bowers of MyDD.com argued that, “…instead of giving [FOX] a golden opportunity to further distort the image of Democratic presidential candidates, and instead of providing them with credibility for all of their past and future attacks against Democrats, it would be best if the Nevada Democratic Party chose a different media partner to broadcast this debate” (Bowers 2007).Footnote 23

Ironically, the bipartisanship that sprang more easily from cross-cutting cleavages and overlapping party issue areas has become that much more critical for parties and politicians striving to rally public support just as the parties are becoming more ideologically polarized at the national level. Similarly, news outlets with independent reserves of credibility and prestige have themselves become less influential, or have squandered their remaining credibility in well-publicized reporting failures (e.g. election night 2000, WMD reporting) or scandals (Jayson Blair, “Rathergate,” and Reuters doctored photos are but a few recent examples). Without being able to draw on these reservoirs of credibility, American parties will likely find their opportunities to actually persuade the public increasingly few and far between.

This, in turn, will almost certainly complicate efforts to forge a bipartisan consensus behind major presidential or congressional policy initiatives. For instance, scholars (e.g., Mueller 1973; Brody 1991) have long recognized that since World War II, presidents sending American troops into harm’s way have frequently enjoyed temporary spikes in their approval ratings. Yet recent research (Baum 2002) has found that the vast majority of this so-called “rally-round-the-flag” effect is located among opposition identifiers, the very individuals who are increasingly likely to discount, if not avoid altogether, elite messages supporting the president’s actions abroad. As a consequence, it seems likely that, at least in many circumstances, future presidents will find the American public less willing than in prior decades to rally behind their president when he or she sends the nation to war.