“Make (European) Politics Civil Again!”
The effects of political incivility in news media on citizens’ political engagement via affective and cognitive responses in the European Union context.
University of Amsterdam Graduate School of Communication
Research Master’s programme Communication Science
Research Master’s Thesis
Student Roos Thijssen (10048049) Supervisor Dr. Andreas R.T. Schuck Date of completion 3rd of February 2017
Abstract
Political incivility is a phenomenon surrounded by ambiguity. Prior research has found that exposure to uncivil negative media content can cause mobilizing and demobilizing effects on citizens’ political engagement. Yet, it is unclear if other negative features of media content (e.g., disagreement and conflict) yield different results and neither is it clear what the underlying affective and cognitive processes are. This study, based on an experiment (N= 392), provides empirical evidence that disagreement, conflict and incivility are three unique features of media content and can be theoretically - and empirically - distinguished based on specific conceptual content characteristics (i.e., tone, level of objectivity, level of interaction). The findings show that uncivil media content has the strongest demobilizing effects on
political engagement (i.e., political efficacy, cynicism, behavioral intention) compared to conflict and disagreement. These effects are mediated by both affective (i.e., emotions and arousal) and cognitive responses (i.e., message processing and recall). Positive emotions function as most prominent mediators. Incivility in political news coverage in particular decreases political efficacy via depressing citizens’ positive emotions. Overall, citizens exposed to uncivil news (vs. conflict vs. disagreement) – about free trade in the European Union (EU) – perceived less sense of competence and influence in EU politics, were more cynical towards EU politics and had less intentions to behave in EU politics, via different emotions and cognitions. This linear negative trend is in line with the current unstable
situation in Europe (e.g., concerning the rise of populism, Brexit, refugee crisis). The present study provides solution to the ambiguity surrounding political incivility: it is solely
detrimental. Thus, make (European) politics civil again!
Keywords: political incivility, conflict, disagreement, media effects, emotions, cognitive responses, political efficacy, political cynicism, social sharing, multiple mediation, experiment, European Union.
Introduction
Political incivility is a phenomenon surrounded by ambiguity; is it mobilizing or is it demobilizing citizens’ political engagement? This question is highly relevant these days because of the political turbulence worldwide. However, there is no academic consensus in how and why political incivility might influence the public as citizens show both negative and positive responses (e.g., Borah, 2012; Brooks & Geer, 2007; Emerson, Joosse, Dukes, Willis, & Cowgill, 2015; Fridkin & Kenney, 2011; Hwang, Pan, & Sun, 2008b; Maisel, 2012; Mutz & Reeves, 2005). Citizens are confronted daily with this “outrageous political discourse” (Sobieraj & Berry, 2011, p.21) by political elites who treat other ethnicities disrespectfully, publicly insult journalists and personally attack their political opponents (Mutz, 2015; Stryker, Conway, & Danielson, 2016; Sydnor, 2015). Political statements often go beyond facts, contain a lack of respect, have a hostile tone (Gervais, 2014) and are emotional by nature (Brader, 2005; Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings, 2011). Media apply the same strategy (i.e. the media logic) by focusing on drama, scandals and conflict and by providing (political) news with a negative tone to attract consumers (see e.g., Brants & van Praag, 2015; Roggeband & Vliegenthart, 2007). The negative strategy of media and politicians turns out to decrease citizens’ political engagement (e.g., Fridkin & Kenney, 2008; Kleinnijenhuis, van Hof, & Oegema, 2006; Maisel, 2012; Mutz, 2015; Sobieraj & Berry, 2011), but seems to trigger the public’s interest as well (e.g., Borah, 2012; Brooks & Geer, 2007; Daignault, Soroka, & Giasson, 2013; Fridkin & Kenney, 2011). Thus, is incivility a new effective ‘tool’ in politics? Or is it actually bad for citizens’ engagement? Previous research has suggested to distinguish different types of negative media content, since effects are both mobilizing and demobilizing (Brooks & Geer, 2007; Lau, Sigelman, & Rovner, 2007; Mutz, 2015; Mutz & Reeves, 2005; Stryker et al., 2016).
The present study tries to shed a new light on the effects of unique features of media content (i.e. disagreement, conflict and incivility) on political engagement. This study reports the findings of an experiment and incorporates both citizens’ affective and cognitive
responses. The contribution of this study is fourfold. First, by empirically establishing a theory-based index to distinguish three features of media content (i.e., disagreement, conflict and incivility). Second, by showing that uncivil media content has the strongest demobilizing effects on political engagement (i.e., political efficacy, cynicism and behavioral intention) compared to conflict and disagreement. Third, by identifying how incivility negatively affects political engagement via underlying mediating mechanisms of both affective (i.e., discrete emotions and arousal) and cognitive responses (i.e., message processing and recall). Positive emotions function as most prominent mediators. Fourth, by demonstrating that negative media content has a strong relation with political internal efficacy, which is not often tested currently. This study incorporates the European Union (EU) context (i.e., political discussion about free-trade agreement CETA), which makes the results relevant for European media professionals and politicians alike, particularly since various elections in Europe are coming up in 2017 (e.g., France, Germany, The Netherlands) in an ‘unstable’ Europe where populist parties are rising and Eurosceptic sentiment is increasing (e.g., Boomgaarden, Schuck, Elenbaas, & de Vreese, 2011; European Commission, 2016; Harteveld, van der Meer, & de Vries, 2013; Tupy, 2016), which, naturally, is also reflected in the style and content of public debate.
Disagreement vs. conflict vs. incivility
In recent years, the level of negativity in European media, public debates and political campaigns has increased (Lengauer, Esser, & Berganza, 2011; Schuck, Vliegenthart, & de Vreese, 2014). To understand the effects of negative news content, it is important to
distinguish different types of negativity (Brooks & Geer, 2007; Mutz, 2015; Mutz & Reeves, 2005; Richmond & McCrowskey, 2010; Stryker et al., 2016). Various scholars agree upon the starting position that disagreement is not the same – or has the same effects – as conflict, and conflict is not the same – or has the same effects – as incivility (Brooks & Geer, 2007; Fridkin & Kenney, 2011; Mutz, 2015; Richmond & McCroskey, 2010). However, distinguishing these concepts theoretically and empirically proved to be challenging in previous studies and thus we know little about the specific effects of these different types of negativity. Moreover, it is hard to systematically define what is perceived as appropriate or uncivil as it is very subjective (Coe, Kenski, & Rains, 2014). In this study, disagreement, conflict and incivility are seen as three unique content features of news media coverage.
Disagreement. Disagreement is inherent in human interaction and defined as the difference in opinion between communicators (McCroskey & Wheeless, 1976). People can strongly disagree on any subject, but will not get into animosity and derision (Brooks & Geer, 2007). Accordingly, disagreement is not necessarily negative or destructive, because two or more sides of an issue are presented, and it is particularly about exchanging substantive arguments without any kind of interpersonal conflict (see e.g., Richmond & McCroskey, 2010).
Conflict. When personal issues are involved, disagreement can escalate into conflict but can still remain civil (Gervais, 2014). As such, conflict is here defined as: the existence of negative affect between two or more actors, with explicit reference to each other, based on issue-focused arguments (see e.g., Richmond & McCroskey, 2010; Schuck, Lecheler, & Otto, 2016). Recent experimental research has further distinguished conflict in different types (i.e., disagreement, low-level conflict, high-level conflict and incivility) (Schuck et al., 2016). No significant differences were found for the low-level condition compared to disagreement, but most interestingly, the high-level conflict condition and incivility have yielded different
effects on cynicism and political participation. In the present study, conflict is therefore not further distinguished in low- or high-levels and incorporated as unique feature of media content next to disagreement and incivility.
Incivility. Incivility is often associated with disagreement and conflict. Scholars propose that incivility and impoliteness are most likely to arise from disagreement and conflict over substantive issues (Sydnor, 2015). Incivility goes beyond negativity, deals with violating social norms and is perceived as impolite and disrespectful (e.g., Gervais, 2014, 2016). Political scientists refer to incivility as the style rather than the substance of political discourse (Mutz, 2015). Research has shown that the type of media has effect on people’s perception of uncivil language: text in online sources is perceived more civil compared to the experience of incivility in television news programs or talk radio (Papacharissi, 2004; Sydnor, 2015). In the present study, the focus is on incivility in textual political discourse (i.e., news articles) and the differentiation of uncivil negative claims (i.e., incivility) from civil negative claims (i.e., disagreement and conflict) (see e.g., incivility index for content analysis
developed by Gervais, 2014). An example of extreme uncivil negative political discourse is outrage: a strategy to purposefully invoke ‘visceral responses’ (e.g., anger, fear, moral indignation) about opponents through hyperboles1, misleading or inaccurate information (Sobieraj & Berry, 2011). Outrage is by definition uncivil, but not all incivility is outrage. As such, incivility is operationalized as: the outrageous interaction between two or more actors, in which people purposefully discredit the other viewpoint on a personal level, go beyond facts, contain a lack of respect and often involving a hostile tone and hyperboles in their negative claims (see e.g., Brooks & Geer, 2007; Gervais, 2014, 2016; Mutz & Reeves, 2005).
Summarizing, the present study focuses on disagreement, conflict and incivility as
1 Hyperboles are similar to exaggerations, e.g., “Candidate A lacks many of the qualities necessary to be
successful in office, and her policies are not impressive,” is negative but civil. A hyperbolic version might be: “Candidate A is the worst candidate in the history of the office and has the dumbest policies I have ever seen.” (Gervais, 2016, p. 5).
three unique features of media content which are incorporated in written political discourse (i.e., speech, claims and statements, political discussions, news articles). Gervais (2014, p.569) distinguished four dimensions to gauge levels of incivility in political discourse. With this study, I aim to extend this index with a theory-based model to distinguish the unique media content features: disagreement, conflict and incivility (see Appendix A).
Different effects on political engagement
In politics, engagement can be defined as generating for instance political interest, intentions to vote, political trust and efficacy (Brooks & Geer, 2007). Disagreement, conflict and incivility are often used in the media as tools to accomplish these political ends.
However, previous research has yielded conflicting results with regards to the effects of these three media content features on citizens’ political engagement.
Negative effects: The first studies on the effects of political conflict and incivility have focused on numerous expected negative effects. This is plausible because conflict is perceived as an undesirable state of affairs for most people and negative news reminds voters of
political imperfection (Brooks & Geer, 2007; Ulbig & Funk, 1999). In the everyday political discourse, incivility became a way to express dissatisfaction and disagreement with the establishment and status quo (Mutz, 2015; Sydnor, 2015). Among researchers this provoked concerns that incivility might harm democracy. Prior studies have shown for instance that exposure to uncivil messages decreases voter turnout (Sobieraj & Berry, 2011), trust in government (Fridkin & Kenney, 2008; Maisel, 2012; Mutz, 2015; Mutz & Reeves, 2005), legitimacy of political candidates (Sobieraj & Berry, 2011) and credibility of (online) media (Thorson, Vraga, & Ekdale, 2010). Furthermore, political incivility increases extreme elite partisan polarization (Stryker, Danielson, & Conway, 2015; Wolf, Strachan, & Shea, 2012), political cynicism (Schuck et al., 2016) and individuals using ‘like-minded’ uncivil media
seem to mimic this in how negative they talk about politics (Gervais, 2014). This is supported by empirical findings that people engaging in aggressive online discussions (i.e., flaming), easier accept this way of communication and in turn are more likely to do it themselves (Hmielowski, Hutchens, & Cicchirillo, 2014). Thus, exposure to severe types of negative content might have detrimental effects on political engagement and democratic processes.
Positive effects: Simultaneously, several scholars pointed out that exposure to uncivil and conflictive messages do not only have harmful consequences. Prior research has shown that negative content – not uncivil per se – compared to positive or mixed content was perceived as more attractive, powerful, sensational and interesting and therefore generated more attention to the message and a higher level of physiological activation (Daignault et al., 2013; Fridkin & Kenney, 2011). In particular, conflict in news can be expected to increase political engagement as a tool of highlighting opinion differences between political actors (Schuck et al., 2014). Previous research has shown that uncivil negative messages lead to increased credibility of political news (Borah, 2012) and generate slightly more interest in politics than either civil negative or positive messages (Brooks & Geer, 2007). Furthermore, issue-based uncivil messages generate strong interest in voting, but when it gets more negative and personal it might turn people off (Brooks & Geer, 2007). Recent research has referred to the existence of a turning point of incivility leading to mobilization of political engagement (Schuck et al., 2016). It has identified that disagreement and low conflict are least mobilizing and that high conflict and incivility increase political participation, but that incivility is less mobilizing than civil conflict. The results are slightly in line with prior findings that relevant negative messages are the most powerful in increasing political
engagement (Fridkin & Kenney, 2011). Thus, although uncivil media content – compared to civil negative and neutral content - in general is perceived as less fair, less important and less informative, it does not automatically lead to detrimental effects.
Different effects on specific types of political engagement.
Because of the existence of several types of political engagement (Brooks & Geer, 2007), in this study – focusing on different features of negative media content – three specific types are further highlighted.
First, negative media content is often related to political cynicism (e.g., Borah, 2012; Fridkin & Kenney, 2008; Maisel, 2012; Mutz, 2015; Schuck et al., 2016). Political cynicism is defined as: “Mistrust generalized from particular leaders or political groups to the political process as a whole – a process perceived to corrupt the persons who participate in it and that draws corrupt persons as participants.” (Cappella & Jamieson, 1997, p. 166). A common concern is that the focus of media on conflict, negativity and strategic games in politics, leads to political cynicism (i.e., the ‘spiral of cynicism’) (Cappella & Jamieson, 1997; Jackson, 2010; Patterson, 1993). Recent research has shown that when conflict in news media content gets too extreme and turns uncivil it increases cynicism (Schuck et al., 2016).
Second, another concept related to political engagement and cynicism is political efficacy: a combination of one’s sense of competence and influence in politics (i.e., internal efficacy) and one’s assessment of the responsiveness of and general trust in the political system (i.e., external efficacy) (Campbell & Miller, 1954; Caprara, Vecchione, Capanna, & Mebane, 2009; Hoffman & Thomson, 2009; Morrell, 2003; Niemi, Craig, & Mattei, 1991; Valentino, Gregorowicz, & Groenendyk, 2009). External efficacy can be explained as ‘generalized cynicism’ as well, since both concepts comprise the general trust towards the political system (e.g., Caprara et al., 2009; Cappella & Jamieson, 1997; Niemi et al., 1991).
Internal efficacy meanwhile is, on a conceptual basis, less related to cynicism and more to people’s political competence level. Some research has demonstrated that negative
campaigning (Lau et al., 2007) and uncivil debate in political blogs decrease political efficacy and political trust (Borah, 2012). Other research did not show evidence of a relationship
between incivility and either political trust or efficacy (Brooks & Geer, 2007). The direct effect of incivility on political efficacy was recently assessed in an experiment where uncivil blog commentary was compared with civil blog commentary in either a value frame or a strategy frame (Borah, 2013). External efficacy was lowest when participants were exposed to an uncivil blog in a strategic news frame. No significant results were found with regard to effects on internal efficacy. However, citizens are expected to be more engaged in politics when they believe their political actions make sense and would be therefore less intimidated by challenges, conflicts or disagreements in political discourse (Valentino et al., 2009). Previous research has indeed identified that other forms of critical political expressions, such as ‘satires’ – defined as “political commentary as a socially acceptable outlet for criticism of elites and the politically powerful” (Glazier, 2014, p.867) – have both negative effects on political cynicism (i.e., external efficacy) and positive effects on internal political efficacy (Baumgartner & Morris, 2006). The findings have indicated that ‘The Daily Show’, an
American satire, generates cynicism but simultaneously makes young viewers more confident about their own ability to understand politics. Complementary research, on the effect of use of satire by teachers, has found that internal efficacy levels were higher but not significant for participants exposed to satire in class (Glazier, 2014). In the present study, the focus is on internal political efficacy, expecting that uncivil content, without substantial political information, subsequently lowers attention to the message and lowers political knowledge, leading to a decrease of internal efficacy (Pedersen, 2012).
Third, regarding the effects of different features of negative media content on political engagement, citizens may – in the presence of political conflict and arousing content –
become more motivated to vote in elections, to seek more information or to discuss politics with others, since they might feel there is something at stake (i.e., political behavioral intentions) (Chen & Berger, 2013; Kahn & Kenney, 1999; Schuck et al., 2014; Sobieraj &
Berry, 2011; Wolak & Marcus, 2007). In line with these mobilizing effects of negative media content on behavioral intentions, other findings have indicated that controversial content increases the likelihood that something will be discussed with others (Chen & Berger, 2013). Other scholars have underlined this need of people to share uncivil and emotionally charged information with friends and family (i.e., social sharing) (Mutz, 2015; Nabi & Green, 2015).
Thus, in line with the effects outlined above on political engagement, it is expected that uncivil media content compared to more civil media content, will lead to more cynicism, to less political efficacy and to more behavioral intentions in politics. This leads to the following hypotheses.
H1a: Incivility in news media content has stronger positive effects on citizen’s political cynicism (i.e., leading to more cynical attitudes) than disagreement and/or conflict. H1b: Incivility in news media content has stronger negative effects on citizen’s political efficacy (i.e., leading to less internal efficacy) than disagreement and/or conflict.
H1c: Incivility in news media content has stronger positive effects on citizen’s political behavioral intentions (i.e., leading to more behavioral intentions) than disagreement and/or conflict.
Furthermore, it is important to understand the reason why such effects on political engagement occur. In the following part the underlying mediating role of affective and cognitive responses – that might play a role in this process – are discussed.
Mediating role of affective and cognitive responses
As described above, media content - including disagreement, conflict and/or incivility – trigger both different cognitive (e.g., attention, recall) and emotional responses (e.g., anger, aversion). Communication research is predominantly focused on the cognitive effects on
people’s political attitudes and behavior since it is easier to disclose these effects (e.g., Lecheler & de Vreese, 2012; Slothuus, 2008). However, recent studies show that affective responses influence the effects of news frames on political opinions and attitudes (Gross, 2008; Lecheler, Schuck, & de Vreese, 2013; Lecheler, Bos, & Vliegenthart, 2015). Scholars therefore unanimously recommend that future research should simultaneously examine affective and cognitive responses and their interaction (Daignault et al., 2013; Gervais, 2014; Gross, 2008; Lecheler et al., 2013; Lecheler et al., 2015; Namkoong, Fung, & Scheufele, 2012; Schuck & Feinholdt, 2015).
In this perspective it is relevant to refer to the Elaboration Likelihood Model of persuasion (ELM) developed by Richard E. Petty and John Cacioppo in 1981. This model states that people use two different routes to elaborate content and adjust their attitudes. First the peripheral route, where people use superficial attributes (emotional cues) of the media content to accept or reject the message leading to affective responses, and second the central route, where people analyze the arguments presented leading to cognitive responses. A key prediction of the ELM is that attitudes, which are changed via the central route to persuasion, will have different effects from attitudes changed via the peripheral route (Petty & Cacioppo, 1981). It is unclear, in relation to the effects of media content, to what extent these two routes occur simultaneously (e.g., Gross, 2008; Holm, 2012; Kühne, 2014; Lecheler et al., 2013). Many theorists prefer to approach emotion and cognition as two independent but interacting systems (Lecheler et al., 2013; Scherer, 2005). In this research, affective responses are seen as a complementary system to cognitive responses, which - together - are even more likely to reveal powerful effects of (negative) media content on political engagement (see Figure 1).
Affective responses
Emotions immediately follow after exposure to a stimulus, such as media content, and therefore are considered as internal mental states (evaluative reactions) instead of traits (personal characteristics) (Nabi, 2003). In general, emotions are short-lived, intense and often elicited by personal relevant events (Feinholdt, 2016; Lecheler et al., 2013; Nabi, 2003; Scherer, 2005).
A distinction should be made between different types of emotions: utilitarian emotions and aesthetic emotions. Utilitarian emotions - also referred to as discrete emotions (Brader & Marcus, 2013) - are elicited when assessing the importance and consequences of certain events (e.g., anger, fear, enthusiasm, hope, disgust) (Lecheler et al., 2015; Scherer, 2005). Aesthetic emotions - also referred to as dimensional emotions (Brader & Marcus, 2013) - do not motivate practical behaviors and are produced by the appreciation of the intrinsic qualities of particular experiences (e.g., awed, admiration, bliss, and solemnity) (Scherer, 2005).
In this research the focus is on discrete emotions because these are easier to isolate when the relation between media content and subsequent affective and cognitive responses is specified (Nabi, 2010; Schuck & Feinholdt, 2015). Discrete emotions can be roughly
distinguished into positive (e.g., contentment, compassion, enthusiasm, hope) and negative emotions (e.g., anger, fear, sadness, disgust). With respect to incivility in particular, research has identified that the most common affective responses of individuals exposed to negative and uncivil content are anger and aversion (Brooks & Geer, 2007; Feinholdt, Schuck, Lecheler, & de Vreese, 2016; Gervais, 2015, 2016; Mutz, 2015; Phillips & Smith, 2004; Sobieraj & Berry, 2011), but incivility also increases positive affections for instance enthusiasm and amusement (Mutz, 2015; Mutz & Reeves, 2005; Feinholdt, 2016).
Another affective response strongly related to the experience of discrete emotions is arousal. Arousal is a physical mechanism to prepare for responding to a certain stimuli and
can be highly varying in intensity, but it is nonspecific with respect to the kind of emotion (negative or positive) that is being experienced (Mutz, 2015). Prior research has shown for instance that positive and negative emotions with high levels of arousal (e.g., enthusiasm, fear and anger) can work as mobilizers for political engagement (Brader & Marcus, 2013; Marcus et al., 2000; Valentino et al., 2011). Research have recently stressed that higher levels of arousal enhance attention and recall, and therefore lead to more political engagement (Berger, 2011; Chen & Berger, 2013). Besides, information that induces high-arousal leads to a higher intention to share that information, whether interpersonally or via the Internet (Mutz, 2015). This is especially the case when arousal is elicited by violation of social norms (incivility), but not automatically by disagreement or conflict, as it remains civil (Mutz, 2015). In this sense, the expectation is that uncivil media content stronger relates to affection (i.e., discrete emotions, arousal) compared to disagreement and conflict. Leading to the following
hypothesis.
H2a: Incivility in news media content has stronger effects on affective responses (i.e., discrete emotions, arousal) than disagreement and/or conflict.
Cognitive responses
Like affective responses, cognitive responses are immediately following after exposure to a stimulus – for instance news frames – as well (Daignault et al., 2013). This ‘black box’ between exposure and effect is often experienced as complex to reveal. Certain theories explain the mediating role of cognitive responses on the basis of consecutive steps: 1) accessibility change (i.e., making considerations more salient and likely to be used when forming opinions), 2) belief importance change (i.e., altering the weight/importance of certain considerations) or 3) belief content change (i.e., offering new considerations and beliefs to the individual) (e.g., Chong & Druckman, 2007; Lecheler, de Vreese, & Slothuus, 2009; Nelson,
Clawson, & Oxley, 1997; Nelson, Lecheler, Schuck, & de Vreese, 2012; Nelson, Lecheler, Schuck, & de Vreese, 2015; Schuck & Feinholdt, 2015; Slothuus, 2008).
In the current study cognitive responses are defined in terms of message processing (e.g., attention, learning, motivation) and recall. Recall can be explained as ‘reflective
information processing’ (Shah, Kwak, Schmierbach, & Zubric, 2004). It is about what people learn and/or remember about the substantive content of the stimuli – after being exposed to the stimuli (Brooks & Geer, 2007; Mutz, 2015). Message processing and recall are expected to vary depending on people’s competence level and on the substance of the message.
Several studies show that uncivil messages are perceived as more attractive, powerful, sensational and interesting and therefore generating more attention and are more likely to be remembered (Daignault et al., 2013; Fridkin & Kenney, 2011; Mutz & Reeves, 2005). On the other hand, experimental research has revealed that uncivil messages were perceived as less informative than civil negative and positive messages, although they contained exactly the same substantive information as civil messages. No significant evidence was found that positive, civil negative and uncivil negative messages are recalled at different levels (Brooks & Geer, 2007). Incivility only slightly generated more attempts to recall information about a political campaign; but this not appeared to affect political learning (Brooks & Geer, 2007).
With regards to most recent findings, it is expected that incivility will decrease attention for the message and decrease learning effects about the substantive information in the message. This leads to the following hypothesis.
H2b: Incivility in news media content has stronger negative effects on cognitive responses (i.e., message processing, recall) than disagreement and/or conflict.
Finally, the present study aims to contribute to the research gap regarding parallel testing of affective and cognitive responses. So far, studies have investigated emotions and
cognitions as outcome variables, as specific conditions and as mediators, but mostly in isolation and not simultaneously (de Vreese & Lecheler, 2012; Feinholdt et al., 2016). In this study, emotions and cognitions are seen as mediators through which the effects of different features of media content on political engagement unfold, at the same time. Therefore, I will test the indirect effects of both affective and cognitive responses, in the relation between different features of media content and political engagement.
RQ1: To what extent do affective and cognitive responses mediate the effect of incivility (compared to conflict and disagreement) on political engagement (i.e., political efficacy, cynicism and behavioral intention)?
* FIGURE 1 ABOUT HERE *
Method
To test the hypotheses an online experiment was set up. Experimental designs are relevant in examining the causal effects of subtle variation in levels of incivility on political engagement while keeping the substantive context of media content constant (e.g., Brooks & Geer, 2007; Mutz & Reeves, 2005;).
Procedure
First, a pilot-study (N=20) was conducted to test the function of the experimental stimulus material, the length of the questionnaire and to see if the central relations in the theoretical model (see Figure 1) could be confirmed with the measures in the expected direction. Based on the outcomes the questionnaire was adjusted and finalized.2
2 Main decisions based on pilot-study: to shorten the questionnaire – participants commented on the length of the
questionnaire – e.g., by cutting unreliable scales (e.g., conflict avoidance, political values), reduce number of items for reliable scales, and deleting measures for belief importance change and belief content change – since participants had difficulties with answering. Political external efficacy was replaced by a political cynicism measures in the pre- and post-test.
The experiment was conducted between the 30th of November and 23rd of December 2016. Participants were randomly assigned to one of the three conditions: a news article that included either disagreement or conflict or incivility. First, participants had to provide informed consent and then were directed to the questionnaire (average duration in minutes: Mdn=32.5). The questionnaire consisted of a pre-test (i.e., demographics, personality traits, tolerance scales, need for affect & cognition, interests, political trust, internal efficacy and political knowledge), thereafter participants were exposed to one of the three experimental news articles. Afterwards, participants completed the post-test questionnaire (i.e.,
manipulation checks, cognitive and emotional mediators and dependent variables). The questionnaire ended with a debriefing in which the research goal was explained and further information about the study was provided. See Appendix B for the complete Dutch
questionnaire.
Sample
The online experiment was conducted based on a convenience sample of Dutch adults (18 years and older) and recruited through three different channels: first via online and personal messages, second via the student database of the University of Amsterdam and third via the members of a sports club ‘US Volleybal Amsterdam’. The participants recruited via the first route did not get any compensation for participation. The participants entering via the student database received ‘research credits’ needed to complete the university degree
requirements and the participants via the sports club did get a small financial compensation (5 euros). In total 392 participants3 completed the full questionnaire and were randomly assigned to the three different experimental conditions: 127 in the disagreement condition, 133 in the
3 536 participants started the questionnaire, 392 were included into the analysis. Regarding the feedback after the
debriefing I assume that 144 participants prematurely dropped out due to the length and complexity of the questionnaire.
conflict condition and 132 in the incivility condition. Most participants had Dutch nationality (390) and two participants had Belgian nationality, 58.7% of the participants were females with age ranges from 18 to 82 (M=39.4, SD=17.9) and most participants completed higher education levels (24.2% pre-university education, 26.5% Bachelors’ degree and 45.2% Masters’ degree). The randomization was successful; there were no significant differences between the three conditions with respect to gender (F(2,389)=0.75, p=.474), age
(F(2,389)=0.11, p=.897), education (F(2,389)=1.31, p=.272) and political orientation (F(2,389)=1.78, p=.170).
Design
The design was a single-factor, post-test only, between-subjects experimental design with random assignment to one of the three conditions. All participants were exposed to a news article about an issue linked to the Comprehensive Economic Trade Agreement (CETA). It comprised a debate between EU politicians about this free trade agreement between the European Union (EU) and Canada. See Appendix C for the final Dutch versions of the stimuli.
Stimulus material
The experimental stimulus material consisted of a constructed newspaper article based on a recent discussion between the EU member states about CETA - a proposed free trade agreement with Canada. The political discussion took place at the end of October 2016 and beginning of November 2016 and was a salient topic in (inter)national media. To keep the manipulated news articles as realistic as possible the articles included a core of factual information collected during that period from various news media (i.e., newspapers,
the authentic character and external validity of the manipulated articles. First a basic reference news article was constructed with an overall negative tone with respect to the negative
discussion concerning CETA. The negative tone was held constant across conditions. Thereof, three unique news articles were built to distinguish disagreement, conflict, or incivility, based on a theoretical distinction (see Appendix A for theoretical differences between the three concepts and their respective operationalization). These differences were added through sentences, claims, statements, quotes and lay-out, keeping the factual
information core constant and only differing with regard to the characteristics of incivility, conflict and disagreement. All versions of the news article were constructed so that they were in accordance with Dutch newspaper writing style. The articles contained a title, a lead and two subheadings introducing larger paragraphs (+/- 600-800 words).
The political discussion in the disagreement condition focused on opposing issue standpoints and the exchange of substantive arguments, with no direct reference to interpersonal conflict. The conflict condition proposed the responses of actors on issue-focused arguments with an explicit reference to each other. And in the incivility condition the discussion was not about substantive issue-focused arguments, but about attacking the other viewpoint on a personal level with disrespectful claims. The combination of the identical section of the reference news article and sections with alternative statements was common in previous experimental studies (e.g., Price, Tewksbury, & Powers, 1997). See Appendix C for the manipulated news articles.
The news issue in all conditions was the same (i.e., CETA) the free-trade agreement between the EU and Canada. It was deliberately decided to choose an issue that is salient in
the media, but not of extremely high personal importance4. The results of the pilot study (n=20) were helpful and guided the decision which topic to select. Participants ranked
‘international trade agreements’ (M=4.10, SD=1.53) as least important compared to education (M=5.90, SD=0.91), care (M=5.40, SD=1.19), counterterrorism (M=5.15, SD=1.42) and climate agreements (M=4.90, SD=1.53)5. Therefore CETA was chosen as main issue in the experimental stimuli. In the main study the issue ‘international trade agreements’ was ranked as personally least important as well (M=4.62, SD=1.38), compared to education (M=6.26, SD=0.95) and climate agreements (M=5.33, SD=1.40).
Manipulation checks
The manipulation checks assessed participants’ perceptions of the article’s level of incivility. Participants were asked on a seven-point semantic differential scale to what extent they had perceived the political discussion in the article as friendly vs. hostile, civilized vs. crude, emotionless vs. emotional, non-confronting vs. confronting, respectful vs. disrespectful and entertaining vs. tragic6. The assumption was that participants’ responses are influenced by the different features of media content (disagreement vs. conflict vs. incivility), thus
participants in the incivility condition were expected to have stronger perceptions of the article’s level of incivility (i.e., higher scores on hostile, crude, emotional, confronting, disrespectful and tragic), compared to the disagreement and conflict condition. Conflict was expected to lead to higher levels of these perceptions compared to disagreement.
4 The salience of issues is reflected by personal importance, high interest of elites and the ongoing discussion on
the issue (Lecheler et al., 2009). Previous research argues that the effects on highly salient issues tend to be smaller, because people already have strong opinions and values on the issue (Nelson et al., 2015; Lecheler et al., 2009). To avoid this in the present study, the CETA-issue was chosen because it was frequently covered in media, but not perceived as personal relevant.
5 Issues are based on the most important political Dutch themes as indicated by Ipsos Politieke Barometer,
September 2016.
6 Various items are based on significant manipulation checks created by Sydnor (2015) to measure ‘perceptions
of incivility’ in news in terms of positive and negative perceptions and based on the specific characteristics of incivility (see Appendix A).
Measures
Control variables.
Single item questions were used to measure participant’s age, gender, education and nationality (see descriptives above). Political orientation was measured on an eleven-point scale from far left (0) to far right (10). Furthermore, other individual predispositions were measured on a seven-point Likert scale in the pre-test, for instance: ‘Political interest’ asking how interested participants were in national and EU politics; ‘Issue interest’ for interests in international free-trade; and ‘Individual issue importance’ asking how important participants perceived international free-trade agreements personally. ‘EU attitudes’ were measured with two standard questions of the Eurobarometer covering general EU membership support and perceived benefits of a country’s EU membership (Boomgaarden et al., 2011). The pre-test ended with four questions to measure participants’ national and EU-specific political knowledge. See Table 1 for all descriptives of these variables.
* TABLE 1 ABOUT HERE *
Mediators.
Affective responses: arousal & discrete verbal emotions.
Arousal was measured with one of the standard dimensions of the Self-Assessment Manikin (SAM) scale developed by Lang (1980). It is a non-verbal pictorial assessment technique to directly measure participants’ level of arousal after exposure to a stimulus. Participants had to indicate which picture out of five – displaying increasing levels of arousal – best identified their feeling when reading the news article on an unlabeled nine-point scale. Overall, participants were not very aroused by the article (M=3.81, SD=2.11).
Discrete emotions were measured with explicit verbal self-reports using a sliding scale (0-100% i.e., higher values indicating stronger positive or negative emotions), which is
supposed to generate more reliable and valid measures (Brader & Marcus, 2013). Based on prior research (Gross, 2008; Lecheler et al., 2013; Lecheler et al., 2015; Nabi, 2003; Quirin & Bode, 2014; Quirin, Kazén, & Kuhl, 2009; Sydnor, 2015) ten different emotions were chosen, five adjectives referred to positive affect (i.e., happy, energetic, satisfied, hopeful and
enthusiasm) and five to negative affect (i.e., helpless, frustrated, aversion, fear and anger). A principal component analysis (PCA) showed that the ten items form two uni-dimensional scales (eigenvalue >1). The five positive emotions (M=31.96, SD=17.74) and five negative emotions (M=40.87, SD=21.28) correlated positively and internal consistency was good (Cronbach’s alpha=.88 for positive emotions, Cronbach’s alpha=.86 for negative emotions) and thus two scales for positive emotions and negative emotions were built and used in the analysis.
Cognitive responses: message processing & recall.
Message processing was measured with eight items about the way participants read and perceived the article on a seven-point Likert scale (higher values indicating stronger message processing). Two items were included for each of the following cognitive
dimensions: attention, motivation, liking and learning (Busselle & Bilandzic, 2009; Hwang, Borah, Namkoong, & Veenstra, 2008a; Wolak & Marcus, 2007). Examples include: “I had a hard time keeping my mind on the news article.” (attention); “I wanted to know how the story ended.” (motivation); “I was fascinated by the article from beginning to end.” (liking); and “I felt more open to the arguments on both sides of the issue.” (learning). A PCA-analysis showed that seven items formed a single uni-dimensional scale (eigenvalue >1), one item for
learning (“I have learned new things about the issue.”) was excluded from further analysis.7 The scale was reliable (Cronbach’s alpha= .87; M=2.52, SD=1.30).
Recall was measured to identify if participants learned and remembered more or less of substantial information regarding the issue after being exposed to the different articles. The measure was based on ‘message recall’ (see e.g., Brooks & Geer, 2007), ‘cognition recall’ and ‘open-ended recall’ (see e.g., Mutz, 2015). Four questions – three multiple choice and one open question – were asked after the stimulus about factual information presented in all three news articles about the CETA-agreement. The open question asked participants: “What does the abbreviation CETA stand for?”. Correct answers on the open question were manually coded (0=not correct – 1=correct). The level of recall was measured by the sum of the correct answers on all four questions (0=no correct answer – 4=all correct answers). Thus, the more correct answers, the better recall participants showed (M=1.85, SD=1.07).
Dependent variables.8 EU political efficacy.
Political efficacy was measured with two items (‘understand’ and ‘informed’) of a standard internal efficacy scale (Niemi et al., 1991). Internal efficacy was defined as one’s sense of competence and influence in EU politics. The items were adjusted to the context of European Union politics and measured on a seven-point Likert scale (i.e., higher values indicating stronger perceptions of efficacy). Both items correlated strongly (r=.81, p<.001) and were combined for further analysis.
EU political cynicism.9
7 Item was included to make a subscale ‘learning’ based on research of Hwang et al. (2008a) examining how
blogger’s commentary affected participants’ thoughts.
8 Items to measure: EU-specific political efficacy, EU political cynicism and behavioral intentions, are based on
Five statements for general political cynicism that are frequently used (e.g., Mutz & Reeves, 2005) were adjusted to specifically refer to EU politics. Participants were asked five questions to indicate how cynical they felt about European politics and EU politicians on a seven-point Likert scale (i.e., higher values indicating stronger cynicism). All five items created a reliable scale for EU-specific political cynicism (Cronbach’s alpha=.74; M=3.94, SD=0.88).
Political behavior: behavioral intention vs. actual behavior.
Behavioral intentions were measured with a battery of seven items about the article topic CETA, asking participants how probable it was for them in the future to for instance search for information, talk about the consequences of CETA with others, pay more attention to the topic, or convince others about their opinion about CETA (see Appendix B for full questionnaire, exact question wording and references to measures). A PCA-analysis showed that these four items formed a uni-dimensional scale (eigenvalue >1), that was reliable (Cronbach’s alpha=.86). Overall, participants scored not very high on the final seven-point scale (1=not probable at all – 7=completely probable) for behavioral intentions (M=2.96, SD=1.41).
Researchers have stressed that sharing uncivil information with friends and family is increasingly important (Mutz, 2015) and that people have a need to share emotionally charged information (Nabi & Green, 2015). Therefore, three items10 to measure the intentions to share information with family and friends formed a second behavioral intentions measure (i.e.,
9 Literature strongly relates the concept of cynicism to external efficacy and less to internal efficacy. Bivariate
correlations showed that in this research, cynicism and internal efficacy are indeed weakly correlated (r=-.17,
p<.01).
10 The items for ‘social sharing’ - based on recommendations of Mutz, 2015, covered participants’ intentions to
forward the article – containing characteristics disagreement, conflict or incivility – to friends, to share the article on social media and to talk about the consequences of the article’s issue with others. The last item was incorporated in the general scale for behavioral intentions as well.
social sharing). This second scale was reasonably reliable, Cronbach’s alpha=.66 (M=2.09, SD=1.02) and deemed sufficient to be used in the further analysis. The overall scale for behavioral intentions and the second scale for social sharing were moderately correlated (r=.648, p<.001).
Additionally, actual behavior was measured with one item, to see if participants not only intended to exert a certain political behavior, but also engaged in actual, real political behavior. For this purpose participants were asked if they wanted to receive additional information about CETA after finishing the questionnaire. This question was included before the debriefing and thus indicated a form of actual behavior; that was going beyond mere behavioral intentions by committing to a particular action which had to be considered to be ‘real’ at the moment of answering the question. Most participants (81.6%) were not willing to receive extra information about the issues presented in the news article (e.g., CETA,
international free-trade).
Analysis
The direct effects of disagreement, conflict and incivility on the dependent variables in this study were measured with one-way ANOVA’s and independent samples t-tests, using mean comparisons. The experimental conditions were dummy-coded as follows for the different analyses discussed below: disagreement = 0 and incivility = 1; conflict = 0 and incivility = 1; disagreement = 0 and conflict = 1.
For further analysis I calculated multiple mediation models as specified by Preacher and Hayes (2004, 2008) using bootstrapping. Multiple mediation analysis is perceived as a “convenient, precise, and parsimonious” (Preacher & Hayes, 2008, pp. 886-887) method to include and test competing theories, as potential mediators selected on the basis of theory, within a single model. It is important to minimize collinearity due to mediators that correlate
too strong (Preacher & Hayes, 2008). Bivariate correlations showed very weak correlations for the affective responses: negative emotions x positive emotions (r=-.05, p=.319), positive emotions x arousal (r=.02 , p=.717), moderate association for negative emotions x arousal (r=.47, p<.001), and a weak correlation for the cognitive responses: message processing vs. recall (r=.24, p<.001). Thus, the mediators represented sufficiently unique constructs. The indirect effects analyses were tested with PROCESS (Hayes, 2013) using 5000 bootstrap samples and 95% bias-corrected and accelerated bootstrap-confidence intervals (BCa-CI) for significance testing.
Results
This research was set up to examine the effects of uncivil media content. In the
analysis I tested the effect of incivility (vs. conflict vs. disagreement) on political engagement, and the mediating role of cognitive and affective responses (see Figure 1 for the theoretical model). I assumed that incivility – compared to disagreement and conflict – will lead to more political cynicism (H1a), less political efficacy (H1b) and more intentions to behave in politics (H1c) (i.e., specific types of political engagement), and that incivility – compared to disagreement and conflict – stronger affects negative affective (H2a) and cognitive responses (H2b). First, the differences between the three experimental conditions were checked.
Manipulation checks
The experimental treatment in this study is a valid test to expose participants to three substantially distinct versions of a political discussion in the media (disagreement, conflict, incivility). Manipulation checks confirmed that all three versions were perceived consistently different in levels of civility in the expected direction (see Appendix D). A one-way ANOVA revealed that the experimental conditions differ significantly for the perceptions ‘hostile’ (F(2, 389)=87.11, p<.001), ‘crude’ (F(2, 389)=169.40, p<.001) and ‘disrespectful’ (F(2,
389)=113.93, p<.001). Independent-samples t-tests confirmed significant differences for the perception ‘hostile’ between incivility (M=5.46, SD=1.10), conflict (M=4.32, SD=1.07) and disagreement (M=3.92, SD=1.12) as well as for ‘crude’ between incivility (M=5.45,
SD=1.13), conflict (M=3.36, SD=1.18) and disagreement (M=2.93, SD=1.25) and for ‘disrespectful’ between incivility (M=5.27, SD=1.22), conflict (M=3.61, SD=1.11) and disagreement (M=3.32, SD=1.02) (see t-values in Appendix E). A similar pattern was found for two other perceptions: ‘emotional’ and ‘confronting’11. Concluding, the manipulation checks show a significant distinction in the perception of the three different conditions: disagreement, conflict and incivility. Thus, manipulation was successful.
Main effects on political engagement
First, a one-way ANOVA revealed no significant main effects for the experimental conditions in relation with EU political cynicism (F(2,389)=0.63, p=.531). H1a was not supported.
Second, main effects were found for the experimental conditions in relation with participants’ belief in one’s own ability (efficacy) regarding EU politics12. Independent-samples t-tests revealed significant differences between incivility (M=3.39, SD=1.39) and disagreement (M=3.74, SD=1.49) for EU political efficacy (t(257)=-2.00, p=.047) and
differences approaching significance between incivility and conflict (M=3.71, SD=1.61) in the same direction (t(257)=-1.75, p=.081, two-tailed test). Thus, incivility decreases participants’ political efficacy stronger compared to conflict and disagreement, confirming H1b.
11 A combined reliable scale of all five manipulation check items, i.e. hostile, crude, disrespectful, emotional and
confronting (Cronbach’s alpha =.86), showed that the three experimental conditions differed significantly based on all perceptions (F(2,389)=173.15, p<.001)). This further confirmed that the manipulation worked
successfully. ‘Entertaining vs. Tragic’ was originally part of the manipulation check, based on recommendations of Sydnor (2015), and showed significant differences between incivility and the other two conditions but not between conflict compared to disagreement.
12 One-way ANOVA showed tentative supporting evidence, approaching conventional levels of significance, in
Third, a one-way ANOVA indicated that exposure to incivility vs. conflict and disagreement significantly decreases behavioral intentions (F(2,389)=3.69, p=.026). T-tests revealed that mean differences between conditions were significant for incivility (M=2.69, SD=1.38) compared to disagreement (M=3.11, SD=1.41) (t(257)=-2.45, p=.015) and incivility compared to conflict (M=3.08, SD=1.43) (t(263)=-2.26, p=.025). These findings were based on the overall behavioral intention scale. Furthermore, selecting the specific items that cover people’s intentions to share political information with relatives (i.e., subscale social sharing), t-tests showed here as well that incivility (M=1.92, SD=0.95) has lead to less social sharing intentions compared to disagreement (M=2.19, SD=1.04) (t(257)=-2.18, p=.030) and conflict (M=2.16, SD=1.04) (t(263)=-1.97, p=.050). These findings present tentative support for H1c that incivility indeed had stronger effects on behavioral intentions compared to disagreement and conflict, but in the unexpected negative direction.
Main effects on affective responses
Furthermore, the findings showed, as assumed, significant main effects of the experimental conditions on affective responses. First, F-tests revealed that all conditions differed significantly from each other with regard to positive emotions in response to the stimuli (F(2, 389)=9.48, p<.001), and the incivility condition from the other two conditions with regard to negative emotions (F(2, 389)=16.50, p<.001). Mean comparisons (see Table 2) showed the significant differences in the expected direction, which meant that incivility most strongly decreased positive emotions: incivility (M=27.48, SD=17.56) vs. disagreement (M=36.87, SD=17.84) (t(257)=-4.27, p<.001), incivility vs. conflict (M=31.72, SD=16.69) (t(263)=-2.02, p=.045) and conflict vs. disagreement (t(258)=-2.40, p=.017). And incivility most strongly increased negative emotions: incivility (M=49.20, SD=19.50) vs. disagreement (M=36.24, SD=21.16) (t(257)=5.13, p<.001) and incivility vs. conflict (M=37.03, SD=20.76)
(t(263)=4.92, p<.001) differed significantly. Thus, incivility has the strongest effects – compared to disagreement and conflict – on positive and negative emotions in the expected direction, providing support for H2a.
Second, a main effect was found on arousal (F(2, 389)=20.89, p<.001). Although the mean scores in Table 2 differed between the groups in the expected direction, the post-hoc test revealed that the differences between the groups were only significantly different for incivility (M=4.72, SD=2.08) compared to disagreement (M=3.24, SD=1.93) (t(257)=5.91, p<.001) and incivility compared to conflict (M=3.44, SD=2.01) (t(263)=5.11, p<.001). Thus, incivility has strongest effects on affective responses (i.e., emotions and arousal) compared to disagreement and conflict in the expected directions, confirming H2a.
Main effects on cognitive responses
Besides affective responses, the analysis also showed that incivility has the assumed main effects on cognitive responses. As posited in H2b exposure to incivility leaded to the lowest levels of cognitive message processing (F(2, 389)=6.64, p<.001) and recall (F(2, 389)=3.15, p=.044). T-tests revealed significant differences for message processing between incivility (M=3.20, SD=1.30) vs. disagreement (M=3.75, SD=1.35) (t(257)=-3.47, p<.001) and incivility vs. conflict (M=3.62, SD=1.29) (t(263)=-2.80, p=.005). Significant differences for recall only occurred between incivility (M=1.67, SD=1.01) and disagreement (M=1.99, SD=1.13) (t(257)=-2.45, p=.015), providing more support for H2b.
Concluding, compared to conflict and disagreement, exposure to incivility causes strongest effects on political efficacy, affective and cognitive responses in the expected direction and strongest effects on behavioral intention in the unexpected direction. This confirms the hypotheses: H1b, H1c (partly), H2a and H2b.
Parallel multiple mediation analysis
In the next part, three different multiple mediator models were tested, with the
PROCESS tool by Hayes (2013) using bootstrapping, to analyze the indirect effects of uncivil media content on: (A) EU political efficacy, (B) EU cynicism, and (C) behavioral intention via affective and cognitive responses (see Appendix F).
MODEL A – EU political efficacy.
The multiple mediation analysis revealed significant indirect effects of incivility vs. disagreement, incivility vs. conflict and conflict vs. disagreement via the various affective and cognitive responses on EU political efficacy (see Table 3). The results showed that the three different conditions had significant effects on the mediators in the expected direction (see above): negative emotions and arousal were increased by incivility; and positive emotions, message processing and recall were decreased compared to the other conditions.
Figure 2 demonstrates the results of the multiple mediation analysis for incivility compared to disagreement. Exposure to incivility had a negative effect on positive emotions (b=-9.387, SE=2.20, p<.001)13, which was related positively to EU political efficacy (b=.011, SE=.005, p=.024). Simultaneously, incivility led to lower levels of cognitive message
processing (b=-.549, SE=.158, p<.001) and recall (b=-0.326, SE=.133, p=.015), which were both also positively related to EU political efficacy (b=.346, SE=.070, p<.001; and b=.313, SE=.078, p<.001 respectively).14 The positive relations between the mediators and the dependent variable mean that incivility could lead to lower levels of EU political efficacy, because it depresses positive emotions, message processing and recall.
13 Mediators are measured on different scales, so b (unstandardized coefficient) varies. See also Figures 2-4. 14 The significant direct effect of incivility on EU political efficacy (b=-.358, SE=.179, p=.047), becomes
Multiple mediation analysis also revealed indirect effects for incivility vs. conflict via positive emotions and recall, and for conflict vs. disagreement via positive emotions (see Table 3). Thus, of the selected affective and cognitive responses, positive emotions turn out to be a significant mediator of the relationship between all experimental conditions and EU political efficacy. Concerning the indirect effects, incivility had a stronger negative effect via positive emotions on EU political efficacy compared to both disagreement (PE=-.102,
BootSE=.051, BCa CI .2199, -.0176]) and conflict (PE=-.068, BootSE=.044, BCa CI [-.1863, -.0060]), than conflict has compared to disagreement (PE=-.057, BootSE=.036, BCa CI [-.1536, -.0041]. The negative unstandardized coefficients (PE) suggested that the effect of the respective conditions reported above (incivility compared to conflict and disagreement; and conflict compared to disagreement) on EU political efficacy was negative, was expected.
* TABLE 3 & FIGURE 2 ABOUT HERE *
MODEL B – EU political cynicism.
Research has often focused on the negative effects of incivility on political trust
instead of efficacy (e.g., Fridkin & Kenney, 2008; Mutz, 2015). However, this study shows no negative relation between the experimental conditions and EU political cynicism in terms of main effects. Nevertheless, a multiple mediator analysis revealed significant indirect effects15 for incivility compared to disagreement on EU political cynicism via both positive emotions (PE=.053, BootSE=.033; BCa CI [.0012, .1373]) and cognitive message processing (PE=.040, BootSE=.027; BCa CI [.0001, .1096]) in the expected direction; the positive effects suggested increasing EU political cynicism (see Table 4).
15 Direct effects for both positive emotions and message processing are approaching conventional levels of
Figure 3 illustrates the results of the multiple mediation analysis for incivility
compared to disagreement. Concerning the direct effects, exposure to incivility had a negative effect on positive emotions (b=-9.387, SE=2.20, p<.001), which had a negative relation with EU political cynicism (b=-.006, SE=.003, p=.065). Via the cognitive route, incivility had a negative effect on message processing (b=-.549, SE=.158, p<.001), which was negatively related to political cynicism as well (b=-.073, SE=.044, p=.099). Thus, incivility indirectly leads to more EU political cynicism, because it depresses positive emotions and message processing.16
* TABLE 4 & FIGURE 3 ABOUT HERE *
MODEL C – Political behavioral intention.
As discussed above, incivility decreases people’s behavioral intentions. A multiple mediation analysis revealed significant indirect effects of the experimental conditions through affective and cognitive responses on behavioral intentions (see Table 5). Figure 4 illustrates the results for incivility compared to disagreement. Incivility decreased positive emotions (b=-9.387, SE=2.20, p<.001), which were positively related to behavioral intentions (b=.014, SE=.005, p=.003), and increased negative emotions (b=12.962, SE=2.527, p<.001), which were also positively related to behavioral intentions (b=.008, SE=.004, p=.054).
Simultaneously via the cognitive route, incivility decreased message processing (b=-.549, SE=.158, p<.001), which was positively related to behavioral intention (b=.404, SE=.066, p<.001).17 The positive relations on the b-path mean that incivility could lead to lower
16 The insignificant direct effect of incivility on EU political cynicism (b=-.096, SE=.101, p=.347), becomes
significant when accounting for the mediators (b=-.249, SE=.114, p=.030).
17 The significant direct effect of incivility on behavioral intention (b=-.425, SE=.173, p=.015), becomes
behavioral intentions, because it depresses positive emotions and message processing, and strengthens negative emotions.18
Multiple mediation analysis also revealed indirect effects for incivility vs. conflict via positive emotions and message processing, and for conflict vs. disagreement via positive emotions (see Table 5). Positive emotions again function as the most prominent mediator, comparing all three experimental conditions. The significant indirect effects showed that incivility had a stronger negative effect via positive emotions on behavioral intention compared to both disagreement (PE=-.129, BootSE=.055, BCa CI [-.2615, -.0435]) and conflict (PE=-.085, BootSE=.049, BCa CI [-.2025, -.0108]), than conflict had compared to disagreement (PE=-.054, BootSE=.036, BCa CI [-.1573, -.0045]. The negative coefficients suggested that the effect of the respective conditions reported above (incivility compared to conflict and disagreement; and conflict compared to disagreement) on behavioral intentions was negative, as expected. A multiple mediation analysis for social sharing – subscale of behavioral intention – has yielded the same significant indirect patterns for all experimental conditions via the same mediators (see Table 6 for the indirect effect values).
With regards to the research question (RQ1), the findings show that both affective and cognitive responses simultaneously mediate the effect of the experimental conditions on political engagement. Incivility – compared to conflict and disagreement – has the strongest effects.
* TABLE 5, TABLE 6 & FIGURE 4 ABOUT HERE *
18 The indirect effect of incivility, via negative emotions, on behavioral intention is approaching conventional
levels of significance in a two-tailed test (p=.054). The positive direct effect of negative emotions on behavioral intention could be interpreted as a tentative mobilizing effect since incivility increases negative emotions. Replication studies should further investigate the demobilizing or mobilizing effects of negative emotions.
Discussion
The present study provides empirical evidence that political incivility is detrimental. Uncivil media content has solely negative effects on political engagement: it reduces citizen’s sense of competence and influence in EU politics (i.e., political efficacy) and it reduces intentions to search for information about political news issues, to share the information with others and talk about politics (i.e., behavioral intentions). Furthermore, incivility decreases people’s positive emotions (i.e., affective responses) and their attention on and learning from political news (i.e., cognitive responses), which in the end leads to less political engagement. It should be noted that conflict in media content, compared to mere disagreement,
demobilizes engagement in European politics as well, but uncivil media content is the biggest malefactor. With regards to previous political communication research which states that uncivil content can both mobilize and demobilize citizens’ engagement (e.g., Brooks & Geer, 2007; Daignault et al., 2013; Fridkin & Kenney, 2011; Schuck et al., 2016; Sobieraj & Berry, 2011; Mutz, 2015), the current findings provide evidence for a somewhat consistent, linear overall trend: all main findings point into the same direction that political incivility in media is detrimental, and more so than for disagreement and conflict. This reinforces prior findings on the demobilizing effects and confirms the concerns of researchers that incivility harms democratic processes.
Furthermore, this study provides valid answers to the important why question – as stressed by previous studies: why incivility leads to such detrimental effects on political engagement. The current study provides evidence that both affective and cognitive responses mediate effects simultaneously. With regards to the Elaboration Likelihood Model (see Theoretical framework), it can be assumed that the uncivil news article – emphasizing subjective claims instead of factual information – is more extensively processed via the peripheral route than via the central route. In line with this, the findings presented here are
plausible showing that incivility decreases cognitive responses (i.e., message processing and recall) and increases affective responses (i.e., arousal and negative emotions). With reference to the uncertainty among scholars which route is most important (e.g., Gross, 2008; Holm, 2012; Kühne, 2014; Lecheler et al., 2013), we can indeed see affective responses as a complementary system to cognitive responses in explaining the effects of media content in general, and of political incivility in particular, on political engagement. In line with this reasoning, this research sheds a new light on the power of positive emotions in political research. Positive emotions function as the most prominent mediator in the current study. So far, positive emotions are often disregarded in political communication research, which is surprising given the fact that they are very powerful as such and also when compared to negative emotions in political contexts (Lecheler et al., 2015). This stresses the need for future studies to incorporate positive emotions in their study designs.
Another relevant contribution is that (uncivil) negative media content in particular decreases citizens’ political internal efficacy (i.e., leading to less sense of competence and influence in politics), which is not often tested. This is in line with Pedersen’s (2012) study on the effects of game frames in political news coverage (i.e., the media focus on political
strategy instead of substantial issues) on political efficacy, instead of political cynicism. He argues that, although several studies have confirmed the ‘spiral of cynicism’, in which game framing leads to more cynicism, it might not be the best or theoretically most relevant concept to measure when analyzing the impact of game frames. And indeed, despite the conceptual similarity between cynicism and external efficacy, his study provides evidence that a game frame decreases internal efficacy more than external efficacy (Pedersen, 2012). The findings presented here supplement this work by incorporating both separate concepts (i.e., internal efficacy and cynicism), and demonstrating that uncivil media content had likewise more effects on EU-specific internal political efficacy (i.e., decreasing people’s feelings of
