The Educative Effects of Extreme Television Media

Abstract

This article investigates whether exposure to extreme television media informs citizens about politics. Using lab experiments with both student and non-student samples, I find that extreme media produce higher levels of political knowledge and that they also produce higher levels of negative affect among viewers compared with control groups. I also show that extreme media are at least as informative as traditional news. This research adds to the growing literature on media effects in a polarized media environment, showing that extreme television media can have a beneficial impact on at least one important area of U.S. politics: citizen competence. To account for external validity and popular conceptions on extreme media’s non-informative nature, I use cross-sectional data from the 2008 National Annenberg Election Survey finding that extreme television viewership correlates with greater political knowledge, while controlling for other known predictors.

Keywords

media effects extreme media political knowledge immigration multi-method research

Do extreme television media inform citizens? Sobieraj and Berry (2011) explain that “outrageous” media are those media hosts and outlets using hyperbolic language, personal attacks, and generally bombastic confrontations to discuss politics.¹ It is important to note the positive outcomes—if any—from viewing extreme media because most recent scholarship suggests that “outrage” or extreme media are normatively bad for U.S. politics (Berry & Sobieraj, 2013; Sobieraj & Berry, 2011). These media are important to research because they are increasing in number and audience share compared with traditional media. Political talk radio—a platform that is classified almost uniformly extreme—continues to increase in market share, whereas extreme political television content continues to emerge with strong growth from Fox News as the conservative outlet and the emergence of MSNBC as the liberal alternative (Sobieraj & Berry, 2011). The proliferation of media choice affects the ways people access their news and information, and thereafter their political behavior (Prior, 2007). We know differences in knowledge acquisition are exacerbated by media polarization essentially making those who are already knowledgeable learn even more, and those who are unknowledgeable increasingly less knowledgeable (Prior, 2007). Thus, the question becomes if Americans are given increasing choices of political media coverage that are extreme in nature, do these media produce or hinder political knowledge in their audience?

This project fills a gap in this line of research. I posit that extreme media can increase political knowledge despite having an ideological point of view. This happens for two reasons. First, viewers of extreme television media have the “ability, motivation, and opportunity” to gain information from this medium (Luskin, 1990). Extreme media seek to entertain as well as inform (Berry & Sobieraj, 2013), meaning extreme media are not altogether different than soft news in this way, and soft news research shows positive effects on knowledge (see Baum, 2003a, 2003b). The result is a direct effect coming through extreme media’s constant discussion of political topics leading viewers to become more politically informed. In addition, extreme media may also activate “affective intelligence” (Brader, 2006) putting viewers in a position to learn information through emotions like anxiety or fear (Visser, Bizer, & Krosnick, 2006). With this theoretical basis, I hypothesize that extreme media create a positive externality² in promoting political knowledge.

I use laboratory experiments to test the hypothesis that extreme television media generate political knowledge using Public Broadcasting Service (PBS), Glenn Beck, and Keith Olbermann as media treatments and Arizona’s Senate Bill (S.B.) 1070 from 2010 as the policy of interest. I find participants assigned to extreme media—Beck and Olbermann—conditions do learn information about policy versus a control group, and they do so at similar rates to more straightforward, mainstream media. There are generally modest effects when interacting knowledge with ideology, indicating there is mild polarization of knowledge among extreme media. There are modest differences between treatments for those who are politically knowledgeable, which follows the literature from soft media. Finally, I find that Beck and Olbermann produce higher responses of negative affect than the mainstream treatment or control in post-test questions, and that interacting affect with treatment conditions shows higher levels of political knowledge among some extreme condition participants. Taken together, these results show evidence that extreme media can produce political knowledge, and subtlety heighten the affective state of their viewers in a way conducive to learning. To account for external validity, I use nationally representative survey data from the National Annenberg Election Survey (NAES) demonstrating that, even while controlling for known predictors of political knowledge and the partisan slant of media outlets and hosts, extreme media are positive and significantly correlated with higher levels of civics-based political knowledge. This project’s findings have important normative and policy implications because this growing area of media—extreme, partisan, “outrageous” television talk shows—may change how citizens interact with information about politics.

The Capacity for Media to Teach

Political knowledge is important for citizens because it shapes the ways in which citizens engage politically and make competent decisions (Althaus, 1998; Delli Carpini & Keeter, 1996; Gilens, 2001). Bartels (1996, 2005, 2007) shows that citizens are often unable to piece together the policy implications of their opinions on political matters, and that their low levels of political knowledge result in sub-optimal electoral outcomes. We should, however, expect that when citizens are presented with facts about current politics and policy that they would learn those facts. This happens primarily through media (Graber, 2001, 2004), where knowledge is taken from the media environment and used to make political decisions. Thus, if a person watches a story on a current political topic, we would expect that person to be able to recall some factual information. The question then becomes, does this educative effect happen when confronted with all the bombast, hyperbole, and angst of extreme media?

Because citizens know very little about politics generally and typically put little effort into researching, a convenient and common place Americans get their information is the news media (Graber, 2004). Television is the most common medium for citizens to get their political information (Graber, 2001; Paletz, 2002). Knowledge is acquired even in shows where one might think accessible, serious political information is in short supply—such as “soft news” (Baum, 2003b). Soft news such as Oprah or The Daily Show With Jon Stewart often generate political knowledge, particularly among those who are least likely to be presented with political information in the first place (i.e., younger and/or politically uninterested citizens). Importantly, Baum and Jamison (2006) show that this increase in political knowledge corresponds to real-world political activity. Voters who consume soft news vote correctly at higher levels than similar voters who do not watch soft news. The soft news literature suggests that extreme media may also be able to teach.

Extreme media serve as an important access point for political information because these shows and hosts put a premium on entertainment (Berry & Sobieraj, 2013). As a result, viewers with less education may be able to glean more information from them than other, denser material on the same topics. Research suggests less educated participants are better able to use television to decrease the differences in knowledge between the highest and lowest educated groups (Jerit, Barabas, & Bolsen, 2006). Yet, the knowledge increases from television are still not uniform with discrepancies occurring between the lowest and highest educated in society (Baum, 2003a, 2003b; Graber, 1988; Prior, 2003). Jerit et al. (2006) find that newspapers, given their more dense nature, exacerbate differences in education levels with the highest educated being able to recall more surveillance knowledge. Importantly, despite the differences in effects based on media type, newspapers and traditional news media in these studies are generally able to increase knowledge over pre-test or control conditions.

A main critique of opinionated, extreme media is that the information presented is biased in ways detrimental to democracy (Barker, 2002; Berry & Sobieraj, 2013; Jamieson & Capella, 2010; Sobieraj & Berry, 2011). However, in spite of the hosts’ ideological predilections, learning can happen through extreme media because discussions about policy and politicians constantly occur. Hosts like Glenn Beck, Sean Hannity, Rachel Maddow, or Ed Shultz may be slanted in their discussion, but they still discuss the most salient political topics of the day. For instance, they often talk about policy issues and members of Congress who draw their ire. The health care debate in 2009 and 2010, for example, was the source of countless hours of extreme media (Chefets, 2010). During this debate, citizens who would have otherwise never heard the names Nancy Pelosi, Harry Reid, or known anything about Congressional procedure (conference committees or “deem-and-pass”) would have been confronted with these people and terms. Thus, although opinions may be affected in a negative way (see Barker, 2002), important political information is also being dispensed to viewers. Currently, there is little literature on the substantive effect of extreme media—particularly television³—on political knowledge. Forgette and Morris (2006) test the effect of CNN’s Crossfire—an early version of high-conflict cable media—against CNN’s traditional State of the Union coverage on attitudes toward government officials and institutions. Participants watching Crossfire exhibited lower evaluations and trust in these institutions compared with control group viewers. At present, there is little to no research assessing the capacity for these media to educate. Yet, individuals who encounter extreme media meet Luskin’s (1990) basic requirement for knowledge acquisition; they have the ability, motivation, and opportunity to acquire and process political information.⁴ Thus, my primary hypothesis is that viewers in extreme media conditions should exhibit higher levels of political knowledge compared with a control group, and that knowledge should be at least as high as viewers of mainstream media like PBS.

Consumers of extreme media may also learn through this medium because it heightens their emotional state through anxiety and fear when compared with more traditional news such as PBS. There is an increasing amount of scholarship about emotional learning in political environments (see Huddy, Feldman, & Cassese, 2007; Marcus, 2002). Although at one time reasoned, rational thought without emotions was considered to be the only method of political learning and decision making, it is now clear that emotions are important and perhaps necessary in decision making (Huddy et al., 2007; Marcus, 2002; Marcus, Sullivan, Theiss-Morse, & Stevens, 2005). Anxious citizens are more likely to seek out information to increase their knowledge on political topics. Furthermore, the presence of a political threat increases the likelihood that citizens become engaged politically (Miller & Krosnick, 2004). Psychologists demonstrate that emotions—specifically anxiety—can motivate people to learn and retain information (Damasio, 1994; Eysenck, 1992; Lazarus & Lazarus, 1994).

Fear and anxiety are not the only possible motivating emotions. Positive emotions such as enthusiasm are also important for learning. Research by Brader (2006) on the 2004 presidential election outlines how the Bush campaign advertisements were able to produce both anxiety and enthusiasm. The results of these ads were that voters who viewed them were both motivated to learn more about the campaign, and remembered more of what they learned. Similarly, Marcus and Mackuen (1993) and Marcus (2002) find that positive emotions can play important roles in information seeking and retention in election settings. However, research shows that extreme media are often about creating anxiety or fear in viewers (Sobieraj & Berry, 2011), so I expect extreme media will heighten the affective state of their viewers when compared with viewers of more mainstream media making them more likely to learn.

Despite these reasons to expect a positive finding, extreme media may not be beneficial because they may encourage or prime misperceptions about the policy they discuss. Prior (2007) claims there are issues with increased media polarization because media increasingly cater to niche markets, and benefits generated by media will also be restricted in those niches. Those who consume extreme media, according to Prior (2007), may be missing key pieces of information needed to make informed decisions or acquire knowledge. Partisanship can also play a role in what people learn about objective information as partisans are better able to learn information that conforms to their worldview (Jerit & Barabas, 2012). Misperceptions are demonstrated to be persistent in highly detailed policy areas (Kuklinski, Quirk, Jerit, Schwieder, & Rich, 2000), and difficult to overcome because people anchor in their prior beliefs (Nyhan & Reifler, 2010). For example, research on Rush Limbaugh listeners in the 1990s suggests that they were more likely than non-listeners to have misperceptions about Clinton administration policy proposals (Barker, 2002). Furthermore, Levendusky (2013) shows extreme media polarize attitudes among those who are ideologically similar to the hosts they watch. This could result in a rejection of pertinent information if there is heterogeneity in ideology or partisanship of the host and viewer. Extreme media are a likely place for misperceptions and attitudinal anchoring that mitigates the degree to which they can successfully inform their viewers.

This project fills an important gap in the literature because previous research on extreme and polarized media mainly focuses on attitudinal polarization (for instance, Levendusky, 2013) and behavioral implications (see DellaVigna & Kaplan, 2007; Dilliplane, 2014). The effect of extreme media on political knowledge is still an open question, and popular and academic accounts of these media suggest that knowledge is in fact unlikely to be produced through this format (“Fox News Viewers Know Less,”2011; Jamieson & Capella, 2010; Sobieraj & Berry, 2011). More to the point, Prior (2013), in his thorough review of the literature on polarized media, demonstrates that there is a dearth of empirical evidence on the capacity of these new media to create or hinder knowledge. As a result, need to examine the capacity for extreme media to produce knowledge is pressing.

An Experiment on Immigration Policy Knowledge

Using a laboratory experiment, I test my main hypothesis by exposing treatment groups to extreme and non-extreme media video clips. In these experiments, I use actual, real-world media content about the 2010 Arizona immigration law, S.B. 1070, to test participants’ “surveillance knowledge” (Jerit et al., 2006).⁵ I use immigration because it is a salient topic often discussed by extreme and non-extreme media, and because policy-specific knowledge is generally not well retained by the general public (Delli Carpini & Keeter, 1996; Lupia & McCubbins, 1998). Immigration as an issue area presents a good test for theories on information acquisition through extreme media because the hosts will allow their partisan predispositions to frame their coverage. A liberal host will frame the debate about the Arizona immigration law so that ideologically opposed participants might be turned off and not learn any policy-specific information. Similarly, liberal participants who watch conservative extreme media might be turned off by that host’s frame as well. Therefore, there is a real chance that no knowledge acquisition occurs for these participants because they may not believe the information coming from an ideologically opposing figure (i.e., Jerit & Barabas, 2012; Taber & Lodge, 2006).

Previous research on news media frames and immigration also demonstrates significant effects on opinions about immigration (Domke, McCoy, & Torres, 1999; Dunaway, Branton, & Abrajano, 2010).⁶ In addition, proximity to large immigrant populations affects the ways Americans feel about immigrants (Tolbert & Hero, 2001), as does education (Hood & Morris, 1998), and ideology and partisanship (Citrin, Green, Muste, & Wong, 1997; Hood & Morris, 1997, 1998). Domke et al. (1999), for instance, framed news stories as “material” or “ethical” and produce differences in attitudes. In the literature, most studies about immigration focus on opinion change, so a notable omission in immigration research is investigating the ways media can increase political knowledge about this dense policy area. Sides and Citrin (2007) use an experiment to test the effect of new information about the size of the immigrant population in the United States on attitudes, but the primary finding is that citizens are simply unknowledgeable about immigration generally. Thus, immigration is a useful policy area for two reasons: (a) Although salient, it is dense and most citizens are uninformed in this policy area, and (b) immigration is ideologically charged enough to demonstrate the (potential) problems with extreme media as a disseminator of political knowledge.

Experimental Procedures

There are two sets of experimental participants in this project. The first set is a convenience student sample from a large research university in the southeastern United States. The second set of participants is comprised of workers from Amazon.com’s Mechanical Turk or “MTurk.” Nationally representative samples are clearly the best samples to use, but both student samples and MTurk workers, or “MTurkers,” are popular sources for experimental participants (Berinsky, Huber, & Lenz, 2012; Druckman & Kam, 2011). Student samples are useful because the treatment effects generally replicate consistently as with more representative samples, and differences between the populations are minimal (Druckman & Kam, 2011). However, for replication, MTurkers provide a legitimate pool of potential participants as Berinsky et al. (2012) state

that relative to other convenience samples often used in experimental research in political science, MTurk subjects are often more representative of the general population and substantially less expensive to recruit. MTurk subjects appear to respond to experimental stimuli in a manner consistent with prior research. (p. 366)

The university student participants and MTurkers were randomized into one of three treatment conditions or a control condition where they were instructed to watch a video that appeared in an online environment. Student participants were offered extra credit for their participation, whereas the MTurkers were paid a nominal fee via MTurk. The videos were hosted on YouTube so that they would approximate likely online viewing experiences. The control group watched a video of a bird that has been demonstrated in numerous experiments to not effect attitudes or beliefs in social science research (i.e., Richey, 2012). The treatment groups watched videos showing a clip of Glenn Beck as seen on Fox News or Countdown With Keith Olbermann from MSNBC, or a clip from the PBS NewsHour program. Glenn Beck and Keith Olbermann were selected for the following reasons: (a) Beck and Olbermann are noted by Sobieraj and Berry (2011) as the most extreme hosts in their study period and (b) both Beck and Olbermann have sufficient content on Arizona’s immigration law. The PBS clip serves as the traditional media content control (Dilliplane, 2011). See the online appendix for a full transcript and table of specific topics mentioned in each video.

Table 1 displays the pertinent demographic summary statistics for the participants in both the MTurk sample and the university student sample. The student participants skew toward traditional college age, are somewhat liberal, are more likely to be women, and half the sample is non-White (Black/African American is the most common ethnicity among minority participants). The MTurkers, however, were almost entirely White, evenly divided on gender, similar ideologically to the student sample, and overall not as knowledgeable as the students about politics based on a pre-test knowledge battery. Based on t-tests, there are no significant differences between the treatment conditions in either sample. Randomization successfully balanced the participants, and unit homogeneity was achieved.

Table 1.

Experimental Participants’ Demographic Summary Statistics.

Variable	M	SD	n
Student participants
White	0.273	0.446	686
Black	0.423	0.494	686
Latino	0.15	0.357	686
Male	0.352	0.478	650
Age	20.393	4.928	631
Pre-test knowledge	3.397	1.017	638
Ideology	4.616	1.548	649
MTurk participants
White	0.824	0.381	131
Black	0.068	0.253	131
Latino	0.07	0.256	128
Male	0.5	0.501	128
Age	33.714	7.607	126
Pre-test knowledge	2.769	1.349	130
Ideology	4.275	1.867	127

Note. SD = standard deviation; MTurk = Amazon.com’s Mechanical Turk.

Each video treatment was roughly 4 min long and began with the typical introduction used by the shows as they appear on television. Each video discusses the same information and uses both the host and several “cutaway” clips of other individuals. Unlike other experiments (Mutz, 2007; that is, Mutz & Reeves, 2005), the treatment in this experiment is not a facsimile of a show or a certain phrase. These clips were the actual aired segments of the show itself. I use actual clips because the effect of extreme media is produced by a combination of the introduction, guests, and the manner in which the hosts discuss the information and so on. For this reason the treatments are the clips themselves. This may limit the types of causal claims I can make based on these shows. For instance, I cannot isolate the particular words or phrases that contribute or inhibit increases in knowledge or misperceptions. This is a limitation, but it is one that is required if we want to study real-life content, which always contains multiple images and phrases.⁷

The questions used to construct the dependent variable—political knowledge—are constructed to be surveillance knowledge questions (Jerit et al., 2006). These types of questions are useful because they gauge if the participants truly extract information from the treatment. The information to answer these questions was contained in the clips, but was also available knowledge to those who may be politically engaged. To ascertain the treatment effects of knowledge contained within the shows, I ask specific questions about the topic covered in all these clips: Arizona’s S.B. 1070. The surveillance knowledge questions are as follows: (a) “Which state passed a bill to address illegal immigration?” (b) “What is the name or number of the bill that addresses illegal immigration in Arizona?” (c) “Does the immigration law in Arizona require anyone who is asked to present their identification for immigration purposes?” and (d) “Which political party is responsible for passing the law in Arizona concerning immigration?”

The first two questions were open ended where the participants typed their answers. Answers accepted as correct and coded as one—zero for all others—for the first question was, “Arizona.” For the second question, I accepted “S.B. 1070” or “1070.” The last two questions the participants answered via radial buttons on the computer screen. The third question was coded correct if answered as “yes,” and the fourth was coded correct if the participants selected “Republican” from their answer choices. The information to answer these questions was contained in each clip used as a treatment.⁸

Experimental Results

The main dependent variable in lab experiment is political knowledge. It is measured as a surveillance knowledge variable and is measured with the five questions on Arizona’s immigration law previously described.⁹ To gain a measure of political knowledge, I conduct a principal components analysis (PCA) using the four questions because this allows us to see if there is a fundamental concept of “knowledge” with this scale. The results demonstrate that the questions strongly load as one component with an eigenvalue well above 1, so I use that component measure as the dependent variable.¹⁰ Figure 1 displays the treatment effects of the media content on the PCA political knowledge measure with 95% confidence intervals (CIs).

Figure 1.

Treatment effects for media on S.B. 1070 knowledge.

As demonstrated in Figure 1, it is clear that the PBS treatment is positive and significant compared with the control group (p < .05, one-tailed test). This finding is both intuitive and expected. PBS is a mainstream, reputable news source associated with producing unbiased news reports (Dilliplane, 2011; Groseclose & Milyo, 2005). Looking at the CIs between the sets of participants, we can see that there is much more variation for the MTurk worker sample. This is likely a product of sample size. The important finding in this figure is that both the extreme media treatments—Beck and Olbermann—are also positive and significant compared with the control in both samples (p < .05, one-tailed test). These findings are important for two reasons. First, the dominant supposition in the literature at this time is that extreme media are not informative, or that whatever knowledge Beck or Olbermann produce would be corralled among like-minded partisans (Jerit & Barabas, 2012). If the literature’s position on extreme media’s informational capacity is correct, there should be a null finding because the partisans in each extreme treatment would cancel each other out. These results show that this is not the case. Extreme media are at least as informative as traditional news such as PBS. Second, these results demonstrate that extreme media can produce a positive externality in the form of political knowledge on a technical area of policy. By replicating these results across two samples—adult via MTurk and a traditional experimental student sample—this finding can be considered robust. An important limitation to this finding is that knowledge is subject to decay over time, and that is not tested here. The key finding is that, despite the ideological nature of the content, these media are able to produce political knowledge.

Interaction effects

There are two important interactions to assess when considering the effect of extreme media: ideology and prior knowledge (Jerit & Barabas, 2012). I use the student-participant data given the large number of participants for statistical power.¹¹ The first interaction is the effect of the ideological orientation of the viewers. As the motivated reasoning literature demonstrates, individuals will use their prior attitudes to color the new information they acquire (i.e., see Jerit & Barabas, 2012; Nyhan & Reifler, 2010). Liberals should learn more from Olbermann and perhaps PBS, but less if they are in the Beck treatment. Similarly, conservatives should learn more from Beck or PBS when compared with Olbermann. Beck and Olbermann should have the largest effects among their ideologically consistent viewers. Moderates should learn more from Beck and Olbermann than PBS because of the entertainment associated with these hosts (Baum, 2003a). To assess this possibility, I break down the participants into their self-identified ideological groups—conservatives, moderates, and liberals—and plot the treatment effects on knowledge of Arizona’s immigration law in Figure 2.

Figure 2.

Student sample treatment effects for media on S.B. 1070 knowledge, by ideology.

The patterns for treatment effects in Figure 2 comport with the basic expectations, particularly among liberal participants. There is a statistically significant difference between Beck and Olbermann using a one-tailed test with 90% CIs. Liberal Beck participants exhibit lower levels of political knowledge, suggesting that they are not learning or are unwilling to learn from Beck. Given the results among moderates, which reflect the general finding from the sample, the Beck–liberals interaction finding lends evidence for polarization among partisans. Among conservatives, Olbermann is associated with slightly less knowledge, but not to a statistically significant degree even with a directional hypothesis and 90% CIs. This could be an artifact of the lower number of conservatives in the sample compared with liberals.¹² Overall, the implications from these graphs suggest there are modest polarization effects with extreme media, which follows from previous work on attitudes (Levendusky, 2013). However, the modesty of the results is important because we might expect to see much more significant effects. It is difficult for people to overcome their prior attitudes (Taber & Lodge, 2006), so finding that effects are not as large as one might assume is more evidence that partisan media are not as pernicious as some make them out to be.

The second interaction is the mediating effect of participants’ political knowledge and can be found in Figure 3. It is likely that the participants with varying levels of civics-based, general political knowledge are affected differently by extreme media.¹³ Based on Jerit et al. (2006) and Baum (2003a, 2003b), those with lower political knowledge should learn more from the extreme treatments because of their attention to entertainment as well as news. Those with the most political knowledge should be able to learn from any news source, but those with average knowledge should learn more from PBS given the more direct nature of content delivery.

Figure 3.

Student sample treatment effects for media on S.B. 1070 knowledge, by pre-test knowledge.

I divide the sample into three categories based on their political knowledge score from the pre-test political knowledge battery in the study—high pre-test knowledge, mean pre-test knowledge, and low pre-test knowledge. The mean pre-test political knowledge mean for the sample was 3.3 based on the five-question scale with a standard deviation (SD) of 1.016. Every participant who scored at least 1 SD above the mean I coded as “high” pre-test knowledge, whereas all those falling at least 1 SD below the mean are coded as “low” political knowledge.

As with the previous interaction effect with ideology, these results are modest but informative. For the high pre-test knowledge participants, participants in each of the treatment groups score significantly higher than the control (p < .05, one-tailed test). However, once again, none of the extreme treatments are significantly different than the PBS treatment showing highly knowledgeable viewers are able to learn from all media. This demonstrates again that Beck and Olbermann are at least as informative as PBS on immigration, which is an addition to the literature on extreme media. With the low and mean pre-test knowledge participants, there is some variability between both the treatments and control, but also the treatments themselves. In the mean pre-knowledge graph, PBS is the treatment where participants exhibited the most knowledge. Beck is significantly lower than PBS (p < .05, one-tailed test), and Olbermann is not different than the control. Yet, in the low pre-test knowledge graph, the means for Beck and Olbermann are positive and significant (p < .10, one-tailed test) relative to PBS and the control. These findings show the entertainment or bombast associated with these extreme television media may help participants be engaged with the material.

The differences in political knowledge acquisition by levels of prior political knowledge mirror what previous studies show with other media. Soft news, the cable-cousin to extreme media because of the premium placed on entertainment, increases knowledge among those who are least knowledgeable (Baum, 2003a, 2003b). In addition, we know that high-knowledge individuals are simply good at learning new political information (Jerit et al., 2006). The conclusions of these tests are that the effect of prior political knowledge on mediated learning works with extreme media similarly to other types of media in U.S. politics.

The role of affect

One reason I hypothesize extreme media may be able to produce learning is their capacity to produce affective responses from their viewers, which should cause them to pay more attention to the content. Anxiety (Marcus & Mackuen, 1993), fear (Brader, 2006; Valentino, Hutchings, Banks, & Davis, 2008), and anger (Isbell, Lair, & Rovenpor, 2013; Parker & Isbell, 2010) are negative emotions, but all engender positive cognitive conditions for learning as long as the affective state is not overwhelming. Thus, subtle negative affective conditions can produce “affective intelligence” (see Brader, 2006). To discern the affective state of participants in this experiment, I ask participants to register their emotional state in the post-test battery. The choices are “angry,” “anxious,” “fearful,” “happy,” “calm,” “enthusiastic,” or “none of the above.” If participants selected “angry,” “anxious,” or “fearful,” I coded them as having a “negative affect” (1 = selected at least one emotion associated with negative affect, 0 = did not report any of these emotions). The modal category for this variable is “0,” meaning that most participants did not report categories associated with negative affect.

As expected, the extreme treatments are successful at making participants more likely to report a negative affective state post-treatment. Figure 4 reports these results.¹⁴

Figure 4.

Student sample treatment effects on self-reported negative affect.

In Figure 4, both Beck and Olbermann produce statistically significantly more negative affect responses in the post-test than the PBS treatment or the control using a one-tailed test with a 95% CI. There is no difference at all between the control and PBS indicating that respondents did not have different affective responses with the content in those conditions. This may help explain why these media are informative, but—importantly—the effect here is substantively subtle. As these media produce small, but discernable, increases in negative affect associated with learning, they may overcome some of the issues research is now demonstrating with partisan media. The type of knowledge being tested here is less ideological than other types of information, and when combined with subtle increases in cognition-promoting affect, we see that political knowledge can be acquired. With the increases in affect established, and a good theoretical ground to support the idea that this affective state assists in gaining political knowledge, a test of treatment and affect is required. To settle this issue, I estimate the model in Table 2.

Table 2.

OLS Estimates for Media and Affect on Political Knowledge.

Variable	Coefficient	SE	Coefficient	SE
PBS	0.358**	(0.135)	0.311*	(0.160)
Beck	0.189^†	(0.135)	0.244	(0.165)
Olbermann	0.342**	(0.136)	0.127	(0.171)
Neg. Affect	0.278**	(0.101)	0.120	(0.208)
PBS × Affect	—	—	0.167	(0.297)
Beck × Affect	—	—	−0.076	(0.282)
Olbermann × Affect	—	—	0.542*	(0.287)
Intercept	−0.316	(0.097)	−0.885	(0.109)
	n	614	n	614
	R ²	.022	R ²	.038

Note. OLS = ordinary least squares; SE = standard error; PBS = Public Broadcasting Service.

†

p < .10. *p < .05. **p < .01; one-tailed test.

The first model demonstrates the fundamental findings—that extreme media can produce political knowledge—but Table 2 also shows post-treatment negative affect is correlated with increased political knowledge. In Model 2, I test the interaction of extreme political media and negative affect on political knowledge. Using a one-tailed test, participants in the Olbermann treatment who reported negative affect have significantly higher knowledge scores than the control. The PBS and Beck treatments are insignificant, and Beck is in the opposite direction. These results suggest, but do not confirm, that there may be differences in affective intelligence between liberal and conservative hosts. The main point remains: Extreme political media can educate, and there is evidence, at least for Olbermann, that knowledge is increased due to the affective state these media engender for viewers.

External Validity

With experiments there is always a concern about the external validity of findings. The experimental results reported above make the causal case that, in a laboratory environment, extreme media can inform their viewers. However, there is a consistent question about the validity of this among the general public (Beaujon, 2012; Berry & Sobieraj, 2013; “Fox News Viewers Know Less,”2011; Rapoza, 2011; Sobieraj & Berry, 2011) with many claiming that these media do not educate about politics generally. To account for those concerns, I use NAES data from 2008 to test if there is a correlation between extreme television consumption and political knowledge among the general U.S. public.¹⁵ It should be noted that I do not use the NAES panel because of the specificity of the media consumption variables. The rolling cross-section has much more detailed media questions, so I sacrifice causal inference for better variables of interest. This data set is superior to the National Election Study (NES) because the NES only asks questions about the frequency of media consumption and type. NES questions do not probe the hosts or substance of the shows as the NAES does. The result is a level of specification with media-independent variables that has yet to be seen in other survey data.

The number of observations is quite large with this data set because of the rolling sample the NAES uses.¹⁶ The dependent variable is Political Knowledge, and is coded 0 to 4. These questions are the NAES version of the traditional Delli Carpini and Keeter (1996) knowledge index common in the NES or other national surveys.¹⁷ This type of knowledge can be termed general or civics-based knowledge, and has been the academic hallmark of a well-informed citizen for almost 20 years. This type of knowledge is different than the knowledge tested in the experiments. However, the point of this test is to establish that extreme media are associated with higher levels of political knowledge overall, which lends external validity to previously reported experimental findings.

The first independent variable of interest is Extreme Host. Extreme Host is coded 1 if the host has a distinct ideological direction and uses ad hominem or bombastic attacks on their opponents, and is coded 0 if they do not. For instance, Glenn Beck and Keith Olbermann are coded as extreme, but Wolf Blitzer is not.¹⁸ To account for the differences between liberal and conservative extreme hosts, I break the first independent variable into Liberal Extreme Host and Conservative Extreme Host.¹⁹ I am also able to include two important control independent variables: Non-Extreme Liberal TV and Non-Extreme Conservative TV. Dilliplane (2011) demonstrates that distinctly partisan media produce differential effects on political behavior. Congruence in media choice and ideology produces political activity, whereas incongruence does not. Including controls for the partisanship or ideological direction of media tests my claim that people are not simply paying more attention because of the self-selection into partisan media (Stroud, 2008), but “extremeness” of the host is significant as well. The Non-Extreme Liberal TV and Non-Extreme Conservative TV variables are based on coding by Dilliplane (2011), and coded here as 1 if the host is partisan and not extreme. All non-partisan, non-extreme television media are coded 0 to represent the omitted category. Where Dilliplane references the partisan bias in the news, I label this variable with the partisan ideological direction for ease of interpretation because there is ample evidence that ideological polarization and homogeneity now make partisanship and ideology synonymous in many cases (Abramowitz, 2010).

The control variables are known predictors of civics-based political knowledge. Those who are more ideological may have higher levels of political knowledge because they will be more engaged, so Ideological is coded 0 to 2 as a folded scale from Annenberg ideological question (0 = moderate, 1 = somewhat ideological, 2 = very ideological). Partisanship is known to predict political knowledge because partisans will be more engaged, so Republican and Democrat are both dichotomously controlled (1 = identity with the party, 0 = independent). Education will clearly predict political knowledge (Delli Carpini & Keeter, 1996; Jerit et al., 2006), as should being male. Education ranges from 1 to 9 following the NAES coding (1 = eighth grade or lower, 2 = some high school, no diploma, 3 = high school diploma, 4 = technical or vocational school, 5 = some college, 6 = associate’s or 2-year college, 7 = 4-year degree, 8 = graduate school, no degree, 9 = graduate or professional degree). Male is dichotomous. Age and Income are also used as control with age being continuous and income being categorical at one through nine. Political Interest will significantly predict political knowledge (Delli Carpini & Keeter, 1996; Jerit et al., 2006), so it is controlled with a variable ranging from zero to three, where 0 = not interested and 3 = very interested in politics. Racial controls are White, Black, and Latino, with being White expected to positively predict political knowledge and Black and Latino expected to negatively relate to political knowledge (Verba, Schlozman, & Brady, 1995). Each of these variables is dichotomous. Finally, media consumption variables are Television, Newspaper, and Internet, which are all coded 0 to 7 representing the number of days in the week a respondent uses these media to get information. Finally, I code for listening to political talk radio with a dichotomous variable.

Because the dependent variable, General Political Knowledge, is coded 0 to 4, ordered probit model specifications are necessary (Borooah, 2002). In addition, as these data are obtained via a rolling cross-section sample over a year, I use robust standard errors clustered on the date of the survey (174 days in total). This will account for any variation over the term of the survey that may affect the results. Table 3 displays two models testing the hypothesis that extreme television media are correlated with increased levels of political knowledge. The first model is a bivariate model testing the effect of the extreme host variable on knowledge without any controls. In Model 1, this variable is clearly significant, and it suggests that there is a correlation between extreme media consumption and civic-based political knowledge. Adding in controls in Model 2 the correlation from Model 1 is still present. Extreme Host is still positive and significant indicating that those who watch extreme hosts have higher political knowledge than those who do not.

Table 3.

Ordered Probit Estimations for Political Knowledge (2008 NAES).

Variable	Model 1			Model 2
Variable	Coefficient	SE	Minimum–Maximum	Coefficient	SE	Minimum–Maximum
Extreme Host	0.478***	(0.056)	0.071	0.237**	(0.071)	0.034
Democrat	—	—	—	−0.074*	(0.023)	0.010
Republican	—	—	—	−0.0016	(0.023)	0.0002
Ideology	—	—	—	0.019*	(0.009)	0.005
Education	—	—	—	0.153***	(0.006)	0.176
Male	—	—	—	0.476***	(0.020)	0.066
Age	—	—	—	0.011***	(0.0007)	0.122
Income	—	—	—	0.064***	(0.005)	0.075
White	—	—	—	0.246***	(0.039)	0.036
Black	—	—	—	−0.231***	(0.048)	0.034
Latino	—	—	—	−0.261***	(0.091)	0.039
Interest	—	—	—	0.197***	(0.014)	0.089
TV	—	—	—	0.006	(0.006)	0.004
Newspaper	—	—	—	0.009**	(0.009)	0.009
Internet	—	—	—	0.022***	(0.003)	0.022
Radio	—	—	—	0.027	(0.018)	0.003
Cutpoint 1	−1.478	(0.012)	—	1.004	(0.070)	—
Cutpoint 2	−0.860	(0.010)	—	1.761	(0.073)	—
Cutpoint 3	−0.255	(0.009)	—	2.509	(0.074)	—
Cutpoint 4	0.579	(0.010)	—	3.513	(0.075)	—
	n = 21,455			n = 14,608
	Wald χ² = 71.37			Wald χ² = 4,075.69
	Pseudo R² = .0013			Pseudo R² = .094

Note. Robust standard errors clustered on sampling date in parentheses. NAES = National Annenberg Election Survey; SE = standard error.

†

p < .10.*p < .05. **p < .01. *** p < .001. (two-tailed)

A constant concern with ordered probit models is their interpretation because the coefficients are not directly implied effects (Greene, 2003). To assess relative effects among the variables of interest, I calculate the average change in the predicated probabilities (Long & Freese, 2006).²⁰ In Model 1, the average change in predicted probability of the dependent variable, going from the minimum to the maximum (i.e., 0 to 1) for Extreme Host is 0.071. When controlling for other known predictors in Model 2, the average change in the predicted probability for Extreme Host 0.034. This is a smaller, but nonetheless significant impact on the dependent variable. For the full results, I turn to Models 3 and 4 (see Table 4).

Table 4.

Ordered Probit Estimations for Political Knowledge (2008 NAES), Full Specifications.

Variable	Model 3			Model 4
Variable	Coefficient	SE	Minimum–Maximum	Coefficient	SE	Minimum–Maximum
Extreme Host	0.282***	(0.070)	0.041	—	—	—
Extreme Liberal Host	—	—	—	0.346**	(0.104)	0.051
Extreme Conservative Host	—	—	—	0.198*	(0.099)	0.028
NE Liberal Host	0.138***	(0.036)	0.019	0.138***	(0.036)	0.019
NE Conservative Host	0.113***	(0.023)	0.015	0.110***	(0.023)	0.015
Democrat	−0.064**	(0.023)	0.009	−0.067**	(0.023)	0.009
Republican	−0.015	(0.024)	0.002	−0.015	(0.024)	0.002
Ideology	0.015^†	(0.009)	0.004	0.015^†	(0.009)	0.004
Education	0.154***	(0.004)	0.178	0.154***	(0.004)	0.178
Male	0.473***	(0.020)	0.066	0.473***	(0.020)	0.066
Age	0.011***	(0.0007)	0.122	0.011***	(0.0007)	0.121
Income	0.062***	(0.005)	0.074	0.063***	(0.005)	0.074
White	0.243***	(0.039)	0.036	0.243***	(0.039)	0.036
Black	−0.229***	(0.049)	0.034	−0.230***	(0.049)	0.034
Latino	−0.258**	(0.092)	0.038	−0.260**	(0.093)	0.039
Interest	0.184***	(0.014)	0.083	0.185***	(0.014)	0.083
TV	0.004	(0.006)	0.003	0.004	(0.006)	0.003
Newspaper	0.010***	(0.002)	0.010	0.010***	(0.002)	0.009
Internet	0.022***	(0.003)	0.022	0.022***	(0.003)	0.022
Radio	0.026	(0.018)	0.003	0.027	(0.018)	0.003
Cutpoint 1	0.975	(0.070)	—	0.974	(0.070)	—
Cutpoint 2	1.734	(0.073)	—	1.732	(0.073)	—
Cutpoint 3	2.483	(0.074)	—	2.482	(0.074)	—
Cutpoint 4	3.489	(0.075)	—	3.488	(0.075)	—
	n = 14,608			n = 14,608
	Wald χ² = 4,081.81			Wald χ² = 4,084.49
	Pseudo R² = .095			Pseudo R² = .095

Note. Robust standard errors clustered on sampling date in parentheses. NAES = National Annenberg Election Survey; SE = standard error.

†

p < .10. *p < .05, **p < .01. ***p < .001. (two-tailed)

In Model 3 the independent variable, Extreme Host, is positive and significant. This demonstrates that there is a significant correlation with watching extreme news media on television and increased levels of civics-based political knowledge. These findings are made more convincing by the controls for ideologically oriented—but not extreme—television. Both Non-Extreme Conservative TV and Non-Extreme Liberal TV are positive and significant as well. The omitted category is all respondents who watch news that is non-ideological and non-extreme. One might expect watching Special Report With Brit Hume or The Situation Room With Wolf Blitzer to be correlated with informed viewers on civics-based political knowledge—which is the implication of these variables. However, even while controlling for these types of shows, extreme media—such as Countdown With Keith Olbermann and The O’Reilly Factor—still correlate with increased political knowledge.

In Model 3, the average change in predicted probability of the dependent variable, going from the minimum to the maximum (i.e., 0 to 1) for Extreme Host is 0.041, whereas Non-Extreme Liberal is 0.019 and Non-Extreme Conservative is 0.015. This means that watching extreme hosts on television generates a larger relative effect on average than other non-extreme hosts. Similarly, in Model 4, even while controlling for the ideological direction of the extreme host—liberal or conservative—the correlation with higher levels of political knowledge for extreme media is still robust and not isolated to one side of the ideological spectrum. Extreme Liberal hosts have a 0.051 average change and Extreme Conservatives have a 0.028 average change in the predicted probabilities of higher levels of knowledge. Non-Extreme Liberal has a 0.019 average change and Non-Extreme Conservative has a 0.015 average change in Model 4. These estimates indicate that, whereas all media consumption variables correlate with knowledge, the extreme media variables are associated with larger substantive effects on general political knowledge. That extreme media consumption is correlated with higher levels of political knowledge when compared with individuals in the omitted category is a counter-intuitive finding based on the previous literature.

Turning to the controls for Models 3 and 4 for partisanship, being a Democrat is associated with lower political knowledge at significant levels, whereas being a Republican is negative, but insignificant. Respondents who are more ideological are correlated with increased levels of knowledge in Model 2, but not Models 3 and 4 with the full specifications. Furthermore, Ideology has a substantively small average change in predicted probabilities while holding all other variables constant. These results suggest that ideology and partisanship play little role in gaining political knowledge in this 2008 NAES sample. For the socio-economic controls, Education, Male, Income, Age, and White all positively and significantly predict having higher levels of political knowledge. Of these, Education and Age have the largest average changes in the predicted probabilities (0.178 and 0.122/0.121, respectively, in Models 3 and 4). These findings mirror other research on political knowledge (i.e., Delli Carpini & Keeter, 1996; Jerit et al., 2006) and make the findings for the extreme media variables more robust. One expects those with more education and those who are older to know more about politics, but they do not subsume the impact or importance of extreme media consumption. Being Black or Latino is associated with statistically significant lower levels of political knowledge. Finally, on the interest in politics and media consumption variables, the results comport with previous research as well. Interest in politics is positive and significant, meaning that those who care about politics know more about it (average change in predicted probabilities 0.083 in Models 3 and 4), whereas more frequent newspaper readers and those who access the Internet often to obtain information are also correlated with increased levels of political knowledge.

The results reported in this section show that extreme television media consumption is correlated with higher levels of political knowledge. That media usage, in general, correlates with political knowledge is not a novel finding. The novelty here stems from the type of media—extreme media—that correlates with civics-based knowledge and suggests there may be more going on than recent reports have detailed (Beaujon, 2012). The results presented here—that extreme media can produce political knowledge on an issue rife with partisanship and ideological motivation, and that extreme media consumption is correlated with higher levels of general political knowledge—add an important piece to this growing area of literature.

I discuss these implications, limitations, and areas for future research in the conclusion.

Conclusion

The debate about the ways media generate and disseminate political information is an important one, and it is clear that there are normatively problematic issues with this type of media as people begin to self-select into an increasingly polarized media environment (see Berry & Sobieraj, 2013; Prior, 2007; Stroud, 2010). Incivility or outrage and partisan bias in media have documented detrimental effects on various political attitudes and behaviors (Arceneaux & Johnson, 2010; Dilliplane, 2011; Mutz, 2007; Mutz & Reeves, 2005). However, extreme media also serve a purpose to inform. To this point, news reports and scholars have found claims that extreme media can be normatively beneficial by increasing political knowledge dubious at best (Beaujon, 2012; Berry & Sobieraj, 2013; “Fox News Viewers Know Less,”2011; Rapoza, 2011; Sobieraj & Berry, 2011). Evidence presented here demonstrates that these media do perform an important democratic function: They can inform the public.

Using experiments for causality and cross-sectional data for external validity, I show extreme media cause policy-specific surveillance knowledge and are correlated with positive and significant increases in aggregate civics-based political knowledge. This is the first time this finding has been demonstrated with this medium in particular. In addition to the main finding, I show that there is slight polarization among participants in ideologically heterogeneous treatments, but the only statistically significant result is among liberals (i.e., they learn less from Beck than Olbermann), and that participants in extreme media treatments are more likely to report negative affect post-treatment. With regard to pre-test knowledge, I show that the effects of extreme media are generally similar to previous work on other types of media, and that participants who are less politically knowledgeable gain more from extreme media than PBS. Finally, when testing the interaction of affect and treatment on political knowledge, I find that Olbermann participants with negative affect are significantly more likely to have higher political knowledge scores. Beck and PBS are statistically insignificant. For external validity, I estimate NAES probit models showing that extreme media viewing is correlated with higher levels of general political knowledge as well. Importantly, the cross-sectional findings do not imply that extreme media are better at producing knowledge when compared with traditional, mainstream media, but that viewers of extreme media are as informed about general political knowledge as other media viewers. In other words, extreme media are not simply speaking to ignoramuses, or keeping the ill informed in that state.

The extreme or outrage media literature is growing exponentially (Berry & Sobieraj, 2013; Hansen & Kim, 2011; Levendusky, 2013; Sobieraj & Berry, 2011), and this study adds to it in an important way. Political knowledge is a fundamental trait for successful democratic citizenship with media the main conduit for knowledge in modern democracy (Delli Carpini & Keeter, 1996; Graber, 2001). As research in this area moves forward, coupling the effects of all extreme media—radio, television, blogs, and so on—into a unified theory will be paramount. As with any study, however, there are caveats.

More experiments need to be done to assess the differences in treatment effects found here. As stated earlier, this project is concerned with a holistic view of extreme media. Thus, more precise, shorter treatments could be done to isolate certain phrases or segments of these extreme programs that will isolate the effects on knowledge acquisition or misperceptions. Glenn Beck, for instance, spent hours of his shows devoted to President Obama’s father insisting that President Obama was not a “true” American, and priming viewers to believe the myth that President Obama is a Muslim (Leibovich, 2010). The knowledge studied here is in no way as ideologically polarizing as the Obama Muslim myth, so there is more work to do in this area.

This project differs and from similar studies in that the main dependent variable here is knowledge generated from extreme media. Prior work shows that extreme media hosts can polarize viewers’ attitudes (Levendusky, 2013), and Jerit and Barabas (2012) demonstrate that partisanship can affect the amount of knowledge citizens acquire from their mainstream news sources. More work can be done to see what types of political knowledge are more likely to be polarized. The effects here are suggestive, but they do not reach the level of polarization found in previous research. Because of the self-selection rife in today’s segmented media market (Stroud, 2010), getting more and better empirical results on polarization is key for the future of this area of research.

The purpose of this project is simply to establish a baseline that knowledge can be generated through extreme media content, and to establish the theoretical link that affect plays a role in producing knowledge through extreme media outlets. More investigation into the divergent effects among those who are more and less politically knowledgeable will help elucidate the full effect of these media. Fundamentally, the point remains that for too long the popular and academic accounts have painted with too broad a brush in describing the problems associated with the media labeled “extreme” in this project. That these media may produce effects like polarization is certainly an issue, but that makes them no different from their mainstream counterparts (see Jerit & Barabas, 2012). They can also produce value such as political knowledge as a positive externality, which means there still may be some normative value in them yet.

Footnotes

Acknowledgements

A previous version of this paper was presented at the 2014 annual meeting of the Midwest Political Science Association. I would like to thank Sean Richey, Jason Reifler, Sarah Gershon, Brian Gaines, and the anonymous reviewers for their very helpful comments on this manuscript. Any errors are mine alone.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

J. Benjamin Taylor is an assistant professor of political science at UNCW. He studies political behavior, the effect of political communication and media on attitudes, and Latino politics.

References

Abramowitz

A. I.

(2010). The disappearing center: Engaged citizens, polarization, and American democracy. New Haven, CT: Yale University Press.

Althaus

S. L.

(1998). Information effects in collective preferences. The American Political Science Review, 92, 545-558.

Arceneaux

Johnson

(2010, September). Does media choice minimize videomalaise? Selective exposure and the effect of talk shows on political trust. Paper presented at the American Political Science Association Annual Meeting, Washington, DC.

Barker

D. C.

(2002). Rushed to judgment: Talk radio, persuasion, and American political behavior. New York, NY: Columbia University Press.

Bartels

L. M.

(1996). Uninformed voters: Informational effects in presidential elections. American Journal of Political Science, 40, 194-230.

Bartels

L. M.

(2005). Homer gets a tax cut: Inequality and public policy in the American mind. Perspectives on Politics, 3, 15-31.

Bartels

L. M.

(2007). Homer gets a warm hug: A note on ignorance and extenuation. Perspectives on Politics, 5, 785-790.

Baum

M. A.

(2003a). Soft news and political knowledge: Evidence of absence or absence of evidence? Political Communication, 20, 173-190.

Baum

M. A.

(2003b). Soft news goes to war: Public opinion and American foreign policy in the new media age. Princeton, NJ: Princeton University Press.

10.

Baum

M. A.

Jamison

A. S.

(2006). The Oprah effect: How soft news helps inattentive citizens vote consistently. The Journal of Politics, 68, 946-959.

11.

Beaujon

(2012, May 23). Survey: NPR’s listeners best-informed, Fox viewers worst-informed. Retrieved from http://www.poynter.org/latest-news/mediawire/174826/survey-nprs-listeners-best-informed-fox-news-viewers-worst-informed/

12.

Berinsky

A. J.

Huber

G. A.

Lenz

G. S.

(2012). Evaluating online labor markets for experimental Research: Amazon.com’s Mechanical Turk. Political Analysis, 20, 351-368. Retrieved from http://doi.org/10.1093/pan/mpr057

13.

Berry

J. M.

Sobieraj

(2013). The outrage industry: Political opinion media and the new incivility. New York, NY: Oxford University Press.

14.

Borooah

V. K.

(2002). Logit and probit: Ordered and multinomial models. Thousand Oaks, CA: Sage.

15.

Brader

(2006). Campaigning for hearts and minds: How emotional appeals in political ads work. Chicago, IL: University of Chicago Press.

16.

Brader

Valentino

N. A.

Suhay

(2008). What triggers public opposition to immigration? Anxiety, group cues, and immigration threat. American Journal of Political Science, 52, 959-978.

17.

Chefets

(2010, November 8). What rush wants, now and in 2012. Newsweek, 156(19), 34-36.

18.

Citrin

Green

D. P.

Muste

Wong

(1997). Public opinion toward immigration reform: The role of economic motivations. The Journal of Politics, 59, 858-881.

19.

Damasio

A. R.

(1994). Descartes’ error: Emotion, reason, and the human brain. New York, NY: Quill.

20.

DellaVigna

Kaplan

(2007). The Fox News effect: Media bias and voting. The Quarterly Journal of Economics, 122, 1187-1234.

21.

Delli Carpini

M. X.

Keeter

(1996). What Americans know about politics and why it matters. New Haven, CT: Yale University Press.

22.

Dilliplane

(2011). All the news you want to hear: The impact of partisan news exposure on political participation. Public Opinion Quarterly, 75, 287-316.

23.

Dilliplane

(2014). Activation, conversion, or reinforcement? The impact of partisan news exposure on vote choice. American Journal of Political Science, 58, 79-94. Retrieved from http://doi.org/10.1111/ajps.12046

24.

Domke

McCoy

Torres

(1999). News media, racial perceptions, and political cognition. Communication Research, 26, 570-607.

25.

Druckman

J. N.

Kam

C. D.

(2011). Students as experimental participants: A defense of the “narrow data base.” In Druckman

J. N.

Green

D. P.

Kuklinski

J. H.

Lupia

(Eds.), Cambridge handbook of experimental political science (pp. 41-57). New York, NY: Cambridge University Press.

26.

Dunaway

Branton

R. P.

Abrajano

M. A.

(2010). Agenda setting, public opinion, and the issue of immigration reform. Social Science Quarterly, 91, 359-378.

27.

Eysenck

M. W.

(1992). Anxiety: The cognitive perspective. East Sussex, UK: Psychology Press.

28.

Forgette

Morris

J. S.

(2006). High-conflict television news and public opinion. Political Research Quarterly, 59, 447-456.

29.

Fox News viewers know less than people who don’t watch any news: Study. (2011, November 21). The Post Huffington. Retrieved from http://www.huffingtonpost.com/2011/11/21/fox-news-viewers-less-informed-people-fairleigh-dickinson_n_1106305.html

30.

Gilens

(2001). Political ignorance and collective policy preferences. The American Political Science Review, 95, 379-396.

31.

Graber

(1988). Processing the news: How people tame the information tide. New York, NY: Longman Press.

32.

Graber

(2001). Processing politics: Learning from television in the internet age. Chicago, IL: University of Chicago Press.

33.

Graber

(2004). Mediated politics and citizenship in the twenty-first century. Annual Review of Psychology, 55, 545-571.

34.

Greene

W. H.

(2003). Econometric analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

35.

Groseclose

Milyo

(2005). A measure of media bias. The Quarterly Journal of Economics, 120, 1191-1237.

36.

Hansen

G. J.

Kim

(2011). Is the media biased against me? A meta-analysis of the hostile media effect research. Communication Research Reports, 28, 169-179.

37.

Hood

M. V.

III Morris

I. L.

(1997). ¿Amigo o enemigo? Context, attitudes, and Anglo public opinion toward immigration. Social Science Quarterly, 78, 309-323.

38.

Hood

M. V.

III Morris

I. L.

(1998). Give us your tired and your poor, . . . but make sure they have a green card: The effects of documented and undocumented migrant context on Anglo opinion toward immigration. Political Behavior, 20, 1-15.

39.

Huddy

Feldman

Cassese

(2007). The affect effect: Dynamics of emotion in political thinking and behavior ( Neuman

W. R.

Ed.). Chicago, IL: University of Chicago Press.

40.

Isbell

L. M.

Lair

E. C.

Rovenpor

D. R.

(2013). Affect-as-information about processing styles: A cognitive malleability approach. Social & Personality Psychology Compass, 7, 93-114. Retrieved from http://doi.org/10.1111/spc3.12010

41.

Jamieson

K. H.

Capella

J. N.

(2010). Echo chamber: Rush Limbaugh and the conservative media establishment. New York, NY: Oxford University Press.

42.

Jerit

Barabas

(2012). Partisan perceptual bias and the information environment. The Journal of Politics, 74, 672-684. Retrieved from http://doi.org/10.1017/S0022381612000187

43.

Jerit

Barabas

Bolsen

(2006). Citizens, knowledge, and the information environment. American Journal of Political Science, 50, 266-282.

44.

Kuklinski

J. H.

Quirk

P. J.

Jerit

Schwieder

Rich

R. F.

(2000). Misinformation and the currency of democratic citizenship. Journal of Politics, 62, 790-816.

45.

Lazarus

R. S.

Lazarus

B. N.

(1994). Passion and reason: Making sense of our emotions. Oxford, UK: Oxford University Press.

46.

Leibovich

(2010, September 29). Being Glenn Beck. The New York Times, p. MM34.

47.

Levendusky

M. S.

(2013). Why do partisan media polarize viewers? American Journal of Political Science, 57, 611-623. Retrieved from http://doi.org/10.1111/ajps.12008

48.

Long

J. S.

Freese

(2006). Regression models for categorical dependent variables using stata (2nd ed.). College Station, TX: Stata Press.

49.

Lupia

McCubbins

M. D.

(1998). The democratic dilemma: Can citizens learn what they need to know? Cambridge, UK: Cambridge University Press.

50.

Luskin

R. C.

(1990). Explaining political sophistication. Political Behavior, 12, 331-361.

51.

Marcus

G. E.

(2002). The sentimental citizen: Emotion in democratic politics. University Park: Pennsylvania State University Press.

52.

Marcus

G. E.

Mackuen

M. B.

(1993). Anxiety, enthusiasm, and the vote: The emotional underpinnings of learning and involvement during presidential campaigns. The American Political Science Review, 87, 672-685. Retrieved from http://doi.org/10.2307/2938743

53.

Marcus

G. E.

Sullivan

J. L.

Theiss-Morse

Stevens

(2005). The emotional foundation of political cognition: The impact of extrinsic anxiety on the formation of political tolerance judgments. Political Psychology, 26, 949-963. Retrieved from http://doi.org/10.1111/j.1467-9221.2005.00452.x

54.

Miller

J. M.

Krosnick

J. A.

(2004). Threat as a motivator of political activism: A field experiment. Political Psychology, 25, 507-523.

55.

Mutz

D. C.

(2007). Effects of “in-your-face” television discourse on perspectives of a legitimate opposition. The American Political Science Review, 101, 621-635.

56.

Mutz

D. C.

Reeves

(2005). The new videomalaise: Effects of televised incivility on political trust. The American Political Science Review, 99, 1-15.

57.

Nyhan

Reifler

(2010). When corrections fail: The persistence of political misperceptions. Political Behavior, 32, 303-330.

58.

Paletz

D. L.

(2002). The media in American politics. New York, NY: Longman.

59.

Parker

M. T.

Isbell

L. M.

(2010). How I vote depends on how I feel: The differential impact of anger and fear on political information processing. Psychological Science, 21, 548-550. Retrieved from http://doi.org/10.1177/0956797610364006

60.

Prior

(2003). Any good news in soft news? The impact of soft news preference on political knowledge. Political Communication, 20, 149-171.

61.

Prior

(2007). Post-broadcast democracy: How media choice increases inequality in political involvement and polarizes elections. New York, NY: Cambridge University Press.

62.

Prior

(2013). Media and political polarization. Annual Review of Political Science, 16, 101-127. Retrieved from http://doi.org/10.1146/annurev-polisci-100711-135242

63.

Rapoza

(2011, November). Fox News viewers uninformed, NPR listeners not, poll suggests. Forbes. Retrieved from http://www.forbes.com/sites/kenrapoza/2011/11/21/fox-news-viewers-uninformed-npr-listeners-not-poll-suggests/

64.

Richey

(2012). Campaign advertising and the stimulation and activation of the authoritarian personality. Political Communication, 29, 24-43.

65.

Sides

Citrin

(2007, April). How large the huddled masses? The causes and consequences of public misperceptions about immigration populations. Annual Meeting of the Midwest Political Science Association, Chicago, IL.

66.

Sobieraj

Berry

J. M.

(2011). From incivility to outrage: Political discourse in blogs, talk radio, and cable news. Political Communication, 28, 19-41.

67.

Stroud

N. J.

(2008). Media use and political predispositions: Revisiting the concept of selective exposure. Political Behavior, 30, 341-366.

68.

Stroud

N. J.

(2010). Polarization and partisan selective exposure. Journal of Communication, 60, 556-576.

69.

Taber

C. S.

Lodge

(2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50, 755-769.

70.

Tolbert

Hero

(2001). Dealing with diversity: Racial/ethnic context and social policy change. Political Research Quarterly, 54, 376-383.

71.

Valentino

N. A.

Hutchings

V. L.

Banks

A. J.

Davis

A. K.

(2008). Is a worried citizen a good citizen? Emotions, political information seeking, and learning via the internet. Political Psychology, 29, 247-273. Retrieved from http://doi.org/10.1111/j.1467-9221.2008.00625.x

72.

Verba

Schlozman

K. L.

Brady

H. E.

(1995). Voice and equality: Civic voluntarism in American politics. Cambridge, MA: Harvard University Press.

73.

Visser

P. S.

Bizer

G. Y.

Krosnick

J. A.

(2006). Exploring the latent structure of strength-related attitude attributes. In Zanna

(Ed.), Advances in experimental social psychology (Vol. 38, pp. 1-67). San Diego, CA: Academic Press.