Negativity and Attitudes –
Are we affected?
A study of negativity and candidate bias tested
across two media channels affecting the
attitudes and behavioral intentions
MASTER’S THESIS
Graduate School of Communication, University of Amsterdam RESEARCH MASTER IN COMMUNICATION SCIENCE
By Andreea Cristina Guguian, Student ID: 10602232 To be submitted by 29th of January, 2016
Word count: 7629
Supervision dr. M.H. van Egmond
This study investigated how potential voters are affected in their attitudes and behavioral intentions while processing information from media about a political candidate. It was hypothesized that citizens are prone to be influenced by negativity bias (attributing more weight to negative information than to the positive one), candidate bias (attributing more weight to a preferred candidate than to the opponent) and by the media channel (attributing more weight to offline media channels than to the online ones) in their attitude and behavioral intentions. Participants were exposed to either a newspaper or a blog article, either negatively or positively framed about one fictional candidate for whom they expressed their preference previously. The results did not support negativity bias solely, but they did support the negativity bias in conjunction with the candidate bias. The media channel had an influence on the behavioral intentions of participants, but not in the expected direction. Participants were more influenced by the online media channels than by the offline ones. Additionally, the findings regarding negativity and candidate bias suggest that participants engaged in counter arguing processes, rating more positively a preferred candidate after reading negative news about him.
1 | P a g e
Mass media negativity and the public responses
There is a growing body of research suggesting people are responding differently to negative versus positive information (e.g. Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; J T Cacioppo & Gardner, 1999; Rozin & Royzman, 2001). Previous studies have shown that negative content has a greater impact on how people process, are affected and recall the information than positive content (e.g. Meffert, Chung, Joiner, Waks, & Garst, 2006). The idea that negative content affects stronger than the positive one has been tested and supported among various areas, from psychology, economics to political behaviour, advertising and mass media. This asymmetry, also called negativity bias, has been explored in the field of psychology and political psychology regarding impression formation. The studies found evidence for the fact that negative information has a greater impact on impression formation than positive information (Singh & Boon Pei Teoh, 2000; Soroka & McAdams, 2015). Studies in the field of political communication show that negative information has more influence on political behaviour (e.g. Lau, 1982; Aragones, 1997; Lau & Rovner, 2009) and on impression formation regarding certain political parties or candidates (e. g. Klein & Ahluwalia, 2005; Lupfer, Weeks, & Dupuis, 2000; Pratto & John, 1991; Rozin & Royzman, 2001; Skowronski & Carlston, 1989). Negativity bias is also present in mass media. News tends to be more negative than positive (Lau, 1982). According to Soroka (2014), negative content has the power to be more attention-grabbing and more physiologically arousing. Therefore, at an individual level, journalists are more attracted to negative information and they regard it as being more important. In their study, Trussler & Soroka (2014) identify another reason for the highly prevalent negative content in the news, namely the readers’ demand. Lastly, mass media, often considered the “Fourth State”, has the duty of a watch-dog over the government and politicians and negative information is accounted as a more
2 | P a g e
critical indicator of the government and politicians’ performance than positive information (e.g. Shoemaker, 1991, 1996).
When it comes to the consequences of the highly negative news content, some scholars are concerned about the detrimental impact of negativity on democracy, such as increased cynicism, low political efficacy and depressed turnout (e. g. Cappella & Jamieson, 1996; Ansolabehere & Iyengar, 1996). Others (e. g. Soroka, 2014) consider there is no real evidence of the detrimental consequences of negativity. Martin (2008) finds evidence for the fact that media covering negative stories, serves as a sentinel for the people, arousing them to participate.
Though the nature and consequences of negativity can be examined further, this study aims to examine how negativity impacts the attitudes and behavioral intentions of citizens. Negativity bias in the news exists and, since the mass media is displaying this asymmetry in communicating rather negative content than positive, the current research aims to map the magnitude of public response towards it. Additionally, when it comes to defining negativity this study will use description given by Lengauer, Esser, & Berganza (2012) to project commonly used negative framing in the media onto participants.
Thus far, most of the literature in the area is focusing its attention on the firsts phases of the elaboration likelihood model of persuasion, namely the information processing (Meffert et al., 2006), information selection (Lyengar, Norpoth, & Hahn, 2004; Trussler & Soroka, 2014) or recall (Ito, Larsen, Smith, & Cacioppo, 1998; Pratto & John, 1991). Since we cannot always understand or identify which of the routes to persuasion are used and how different stimuli are processed, we are less likely to identify when negativity is more effective in changing the attitudes and, consequently the behavior. Therefore, this study aims to assess to
3 | P a g e
what extend the final attitudes and behavioral intentions of people are influenced by negativity.
Though negativity bias can have a great impact in various areas, when it comes to the political environment, mass media negativity can change attitudes towards the political candidates and voting behaviors, causing an impact on democracy. For this reason, a study mapping this impact and measuring its effects on the final attitudes and behavioral intentions of citizens and possible voters is valuable. According to Meffert et al. (2006) citizens are using and processing the information given by media according to their own motivations and preferences. Media effects on political attitudes and behavioral intention of voting are a mix of news stories and interpretations of citizens. Therefore, we aim to examine how citizens respond to negative versus positive news articles about potential candidates and what attitudes and behavioral intentions this type of content is depicting in citizens.
Garramone, G.M.Atkin (1990) found evidence supporting that citizens develop a more polarized attitude after the exposure to negative advertising. Likewise, Meffert et al. (2006) found that for a majority of their participants negative information lead to more positive evaluations of the candidate, rather than negative as commonly expected.
Therefore, we expect that:
H1: Negative news content will have a more positive effect on attitudes and behavioral intention of participants than positive news content.
Candidate preference
The political environment involves more than one candidate and the news media are presenting information about all these candidates simultaneously. Regardless of the individual preference, citizens may find information in the news about a perceived preferred candidate,
4 | P a g e
as well as about a perceived opponent candidate. In the field of political communication, some studies show that people pay more attention to the information about a preferred candidate (e. g. Donsbach, 1991; Redlawsk, 2001). Meffert et al. (2006) found proof of the candidate bias (attributing more attention to the preferred candidate) in their study. Participants selected, processed and recalled better the information about the candidate to whom they showed a strong initial preference. Therefore, we can expect citizens to pay more attention and to be more influenced by information about a candidate to whom they have made an initial commitment. When it comes to the interaction between the preference for a candidate and negativity, the same study found evidence supporting that voters with an initial candidate preference, after exposed to negative information about their preferred candidate, arrived at polarized evaluations in favor of their preferred candidate. The authors are stating: “Any raw information from the media is, if selected, likely to be used and transformed by voters according to their motivations and pre-existing preferences” (Meffert et al., 2006, p. 28). Thus, participants reading negative information about their preferred candidate rated him more favorably in their final evaluations. The current study aims to map how negativity bias affects the attitudes and the behavioral intention of citizens exposed to information about a preferred candidate compared to those exposed o information about an opponent. Therefore we expect negativity and candidate preference to shape the attitudes and behavioral intention of the participants as follows:
H2: The negative news content about the preferred candidate will have a more positive effect on the attitudes and the behavioral intentions of the participants than the positive news content about the preferred candidate, whereas the positive news content about the opponent candidate will have a more positive effect than negative news content about the opponent candidate.
5 | P a g e
Channel differences
The media channel is a very important determinant of the perception of people on the news. Almost half a century ago McLuhan (1964) stated “medium is the message”. Nevertheless, researchers are still struggling to determine and isolate the determinants and the factors accounted for channel differences. When we are talking about online media channels and the traditional media channels, clear differences can be observed. First, as Livingstone (1999) is pinpointing, the online media channels have the advantage to be interactive, with an unlimited range of content and with a global nature of communication and audience. Another important difference between online channels and the traditional ones is the “flow” of information. While traditional mass-media is characterized by a vertical-flow of information, where readers don’t have the opportunity to dialogize and actively participate, the new media gives citizens the opportunity to express views and communicate between each other and with their political leaders directly. Bentivegna (2002) summarizes several benefits of the online media channels over traditional channels, such as interactivity, co-presence of vertical and horizontal communication, promoting equality, limited intermediation by journalists and others. When discussing online media channels and the effects they have on voters, scholars suggest there should be made a clear distinction between the different online channels. Johnson & Kaye (2000) examined, among politically interested internet users, which of the online media channels are most believable, fair and accurate. By comparing political online issue oriented sources, candidate-focused Web sites, blogs, electronic mailing lists/bulletin boards and chat rooms/instant messaging, they discovered blogs, followed by issue-oriented sites as the most credible sources of online media channels. According to Meraz (2009), because of its decentralized citizen control as opposed to hierarchical, elite control, blogs are believed to be a vehicle of democracy. Therefore, taking into account the co-presence of vertical and horizontal flow of information, blogs are viewed as interactive. They are a form
6 | P a g e
of participation, conversation, collaboration and community (O’ Reilly, 2004). For these reasons, this study finds the most appropriate online political source of information in the form of a blog. As for the traditional media channel comparable to a blog, participants will be exposed to a newspaper article. A study conducted on the impact of media channel on candidate evaluation found evidence supporting the positive influence of online channels versus the offline ones (Gaddie, 2002). Participants rated candidates presenting their message over the internet as better than participants presenting their message on traditional media.
The above mentioned benefits of online media channels make us expect negativity to have a lower impact on peoples’ attitudes and behavioral intention when exposed to online media channels compared with when exposed to traditional media channels. The interactivity of the blogs and their role as a democratic vehicle may increase the political efficacy and may reduce the effects of negativity, compared with traditional media channels. Another important determinant of the channel differences is the perceived source credibility. According to Van Dalen & Van Aelst (2014), The Netherlands is considered one of the countries where media is more powerful than politicians. Thus, the trust in media in general and, in traditional media channels, specifically, is expected to be higher here. If, indeed, traditional media is more credible here, we may expect participants to take more seriously the negativity presented in newspapers and to be more influenced in their attitudes after reading a newspaper than after reading a blog.
Therefore, negativity is expected to have a higher impact in terms of attitudes and behavioral intention among the participants exposed to the offline media channel in comparison with those exposed to the online media channel. Nevertheless, as presented above, the stronger impact of negativity does not always lead to negative attitudes. Previous studies (Meffert et al., 2006) have found that negativity elicits positive attitudes when the news are about a preferred candidate. When taking into account negativity and candidate
7 | P a g e
preference in conjunction with online versus offline media channels, we expect the attitudes and behavioral intention of the participants to be shaped like this:
H3: The negative content about the preferred candidate presented in the offline media will have a more positive effect on the attitudes and behavioral intentions of the participants than the negative content presented about the preferred candidate presented in online media, and likewise the positive content about the preferred candidate presented in the offline media will have a more positive effect than the positive content presented about the preferred candidate presented in online media.
The existing literature studying the channel difference in conjunction with negativity and candidate bias is scarce. For this reason, a hypothesis regarding how the negativity bias works among different media channels for the perceived opponent candidate is impossible to be outlined. Therefore, for this hypothesis we will use an inductive reasoning while testing it.
The current study aims to test the above stated hypotheses by means of an experiment in which participants will be exposed to either positively or negatively framed news articles about a preferred or opponent candidate (based on a previously made choice) from a blog or from a newspaper, while their attitudes and behavioral intentions will be measured.
Method
Participants and design
The experimental design was a 2 (article valence) X 2 (media channel) X 2 (candidate preference congruence) between-subjects factorial design. The news article was either positively or negatively framed; the media channels were either offline (newspaper) or online (blog); and the candidate was represented either by the preferred or by the opponent candidate, according to a previously expressed preference. The experiment was run as a
8 | P a g e
computer-based experiment with the help of Qualtrics Survey Platform. Participants were presented different stimuli and their attitudes, behavioral intention and recall was measured afterwards. The procedures and materials are presented in the following section in the same order as in the experiment.
A convenience sample from the Dutch population of 157 participants was recruited and participated in this experiment (94 male and 63 female; mean age 28.50, ranging from 18 to 85; 120 - highly educated (scientific), 31 – medium educated (applied science) and 6 – lower educated (secondary school) and most of them with an above average subjective standard of living (1 – Arm to 9 – Rijk, M = 6.66, SD =.75.)). Participants were recruited online (51%) via social media channels and offline (49%) via direct contact, but the data collection was entirely made online. Participation was voluntary and it was not rewarded.
Predictors
News article valence
The news articles were fictional, but based on a news story present in the Dutch press few weeks before the experiment, in order to preserve the ecological validity and to help participants identify, to a certain degree, with the subject of the articles. The participants were announced they are going to read a news article that could be very well present in the Dutch news tomorrow. According to Lengauer, Esser, & Berganza (2012) the operational manifestation of negativity in the news has several forms and dimensions. Based on the description offered in his article, the negative articles were modeled by focus on incapability and misconduct; they deployed a negative tone towards the political candidate and they used negative words (‘negatief’, ‘vaag’, ‘onrealistisch’, ‘drammering’, ‘autoritair’, ‘maakt zorgen’). The positive stimuli did not make use of focus on incapability or misconduct, did not use a negative tone towards the candidate and used positive words (‘enthousiast’,
9 | P a g e
‘duidelijk’, ‘realistisch’, ‘zelfbewust’, ‘overtuigend’). The stimuli were pretested to verify whether they were perceived as negative or positive. After the first pretest, small adjustments have been made and the final version consisted of two negative articles and two positive articles, for each of the media channels (see Appendix 2). In the final analysis were included participants who read 79 positively framed news articles and 78 negatively framed news articles.
Figure 1. Example of article used as stimulus material (left – blog example, right – newspaper example).
Media channel
Participants were instructed before reading each article that they are going to read a blog, respectively newspaper article. The writing style was inspired by two articles that covered the same story in Dutch press, one article from a newspaper and one article from a blog. Essentially the newspaper article used a more detached tone, was impartial, used quotes and used a formal writing style. On the contrary, the blog article was opinionated, it generalized opinions without giving quotes, it used an informal writing style and it was engaging. The news story was kept consistent across the two media channels and the two valences, offering
10 | P a g e
the same information, but using different writing style (see Appendix 2). The newspaper and blog articles were designed to look similar to a newspaper (see Figure 1), respectively a blog. In the final analysis were included participants who read 76 blog articles and 81 newspaper articles.
Candidate
The stimuli presented had as main character of the story one of two fictional candidates. Participants read in the first part of the experiment the standpoints of both of the fictional candidates and were asked to show their preference for one of the two candidates, by this their preferred and opponent candidate being established. In the final analysis 74 participants reading articles about their preferred candidate and 83 participants reading articles about their opponent candidate were included.
Materials and procedures
Introduction and informed consent
Participants were announced they will participate in a study about Political Communication and the general ASCOR guidelines were followed for writing this section (see Appendix 4).
Presentation of the candidates’ standpoints
Participants were asked to read the political standpoints of two fictional candidates, the leaders of two political parties. They were instructed to make a choice for one of the two candidates while reading the standpoints. The standpoints referred to different areas, such as safety & security, economy & work, health system, environment and foreign affairs. Although it was not mentioned, Thomas Dijkstra leaned more towards a right-conservative ideology and Pieter de Groot more to a left-progressive ideology (see Appendix 1). Without being able to return to previously read standpoints, participants were asked to show their preference for one
11 | P a g e
of the two candidates. This question determined the preferred and the opponent candidate for each of the participants. The order in which the candidates’ standpoints were presented was randomized.
Distraction material
Right after the presentation of the standpoints and after determining the preference for a certain candidate, a distraction material was included in order to make the manipulation less obvious. A short text was presented to the participants and three recall questions about the text were asked (see Appendix 4).
Presentation of the news articles
Participants were instructed they will read a blog/newspaper article that resembles an article that could be presented in the Dutch press. They received in evenly randomized order, depending on their preference for one candidate, one news article to read and the stimuli were equally distributed among groups (see Table 1).
POSITIVE NEGATIVE
Thomas Pieter Thomas Pieter
NEWSPAPER 26 20 22 13
BLOG 17 16 16 27
Table 1. Distribution of the stimuli.
Outcomes
Attitude I towards the article
The first scale measuring the attitude towards the article was constructed after a scale of Cacioppo, Petty, & Morris (1983) meant to assess the message evaluation. It contained 7 items (‘Duidelijk’, ‘Leuk’,’Overtuigend’, ‘Goed Geschreven’,’Geloofwaardig’,
12 | P a g e
‘Waarschijnlijk in de nederlandse politiek’, ‘Betrouwbaar’) measured on a Likert scale from 1 (helemaal niet) to 9 (heel erg). An example of questions is: “Hoe duidelijk vind je dat de informatie gecommuniceerd wordt in het artikel?” A principal component analysis (PCA) showed the 7 items formed a two-dimensional scale: two components had an eigenvalue above 1 and there was no clear point of inflexion in the scree plot. All items correlated positively with the first component, the item "waarschijnlijk in de Nederlandse politiek" having the weakest association (factor loading = .29). An orthogonal (Varimax) rotation revealed the loading of component "waarschijnlijk in de Nederlandse politiek" of -.006 (see Appendix 3), whereas the other components’ loadings were above .69. Therefore we decided to exclude the problematic item, also taking into account the fact that it might not measure the same construct of how good the article is. The six remaining items formed a reliable scale together (N = 157, M = 4.28, SD = 1.52, α = .884, KMO = .876).
Attitude II towards the article
The second outcome measuring the attitude towards the article referred to the quality of the source of the article and used a scale adapted after the scale of Gierl, Stich, & Strohmayr (1997). It contained 6 items measured on a semantic-differential scale (‘Competent’,’Ervaren’,’ Professioneel’, ‘Deskundig’, ‘Betrouwbaar’, ‘Eerlijk’) from 1 (negative) to 9 (positive). A principal component analysis (PCA) showed the 6 items formed a single uni-dimensional scale: only one component had an eigenvalue above 1 (eigenvalue = 4.11) and there was a clear point of inflexion after this component in the scree plot. All items correlated positively and the item "Deskundig" had the strongest association (factor loading = .91, see Appendix 3). The six items formed a reliable scale together (N = 157, M = 4.46, SD = 1.49, α = .907, KMO = .854).
13 | P a g e
Attitude I towards the candidate
The first outcome measuring the attitude towards the candidate included the sympathy expressed for the candidate presented in the article with a semantic-differential scale from 1 (‘Zeer onsympathiek’) to 9 (‘Zeer sympathiek’) (N = 157, M = 4.59 and SD = 1.64).
Attitude II towards the candidate
The second outcome measuring the attitude towards the candidate used a scale adapted after the scale of Meirick & Nisbett (2011). It contained 8 items measured on a Likert scale (‘Realistisch’, ‘Betrouwbaar’,’ Fatsoenlijk’, ‘Rechtvaardig’, ‘Toegewijd’, ‘Direct’, ‘Besluitvaardig’, ‘Leuk’) with 1 (‘Helemaal niet’), 2 (‘Nauwelijks’), 3 (‘Redelijk’), 4 (‘Behoorlijk’), 5 (‘Heel erg’), 6 (‘Kan niet zeggen’) as answering options. The value 6 was declared missing. A principal component analysis (PCA) showed the 8 items formed a two-dimensional scale: two components had an eigenvalue above 1 and there was no clear point of inflexion in the scree plot. All items correlated positively with the first component, the items “Direct” and “Besluitvaardig” having the weakest association. An orthogonal (Varimax) rotation revealed the loading of the problematic components of -.03 and .24 (see Appendix 3). Therefore we decided to exclude these two items, also taking into account the fact that probably the article did not revealed enough information so that participants could answer rightfully to these questions. The six remaining items formed a reliable scale together (N = 136, M = 3.08, SD = .75, α = .871, KMO = .829).
Attitude III towards the candidate
The third outcome measuring the attitude towards the candidate used a self-developed scale. It contained 7 items measured on a semantic-differential scale (‘Verstandig’,’Goed’, ‘Aangenaam’,’Juist’, ‘Positief’,’Gewenst’, ‘Acceptabel’) from 1 (negative) to 9 (positive). A principal component analysis (PCA) showed the 7 items formed a single uni-dimensional
14 | P a g e
scale: only one component had an eigenvalue above 1 (eigenvalue = 5.46) and there was a clear point of inflexion after this component in the scree plot. All items correlated positively and the item "Aangenaam" had the strongest association (factor loading = .92, see Appendix 3). The six items formed a reliable scale together (N = 157, M = 4.85, SD = 1.37, α = .952, KMO = .933).
Voting likelihood
Voting likelihood for the candidate presented in the article was assessed by a 9 point scale (1 (Ja) to 9 (Nee)) (N = 157, M = 5.93 and SD = 2.34).
Recall
The recall of the information presented in the article was assessed by three questions regarding the content of the article: 1. “Wat was de naam van het opiniepeilingbureau dat in het artikel wordt genoemd?”; 2. “Wat was de naam van de andere kandidaat dat in het artikel wordt genoemd?”; 3. “Toen men het had over de partij die vooraan staat in de peilingen, hoeveel kamerzetels had deze partij?”. Each of the questions had three answering possibilities: 1. “Deze informatie is aangeboden, namelijk” (open answer box); 2. “Deze informatie is aangeboden, maar ik herinner me het niet”; 3. “Deze informatie was niet aangeboden”. The wrong answers filled in the open answer box of the first answering option were recoded as the second option of answering. The responses of the three questions were merged and recoded. The participants answering all questions correctly received the highest value (7) and the participants answering all questions wrongly received the lowest value (1) (N = 157, M = 4.66 and SD = 1.37).
15 | P a g e
Control measures
In order to ensure we avoid making false assumptions regarding the impact of negativity, candidate preference and media channel on the attitudes and behavioral intentions of participants, we included several control measures in the questionnaire. Based on a review of the literature most relevant and commonly used control measure in political communication, besides the demographic measure, are the political orientation of participants, the level of interest in politics and media usage. Because this study involved two media channels, source credibility was also included as a control measure.
Political interest and orientation
Political interest was assessed by two questions: voting intention, namely how likely is that participants will vote in the next elections (“Hoe groot is de kans dat jij bij de volgende verkiezingen zal gaan stemmen?” – measured 1 (‘Heel klein’) to 9 (‘Heel groot’)) (N = 157, M = 7.43 and SD = 2.39) and how much interest they have in politics (“In hoeverre heb je belangstelling voor politiek? Is dat...” – measured 1 (‘Helemaal geen’), 2 (‘Niet zo veel’), 3 (‘Tamelijk veel’), 4 (‘Veel’)) (N = 157, M = 2.73 and SD = .78). The political orientation was measured by two dimensions: Left – Right (1 (‘Links’) to 9 (‘Rechts’)) (N = 157, M = 4.05 and SD = 1.89) and Conservative – Progressive (1 (‘Conservatief’) to 9 (‘Progressief’)) (N = 157, M = 6.79 and SD = 1.75). Considering the unfamiliarity expressed by some participants regarding the measure conservative – progressive, this measure will be excluded from the analyses.
Media usage
Media usage was assessed by a six point measurement: 1 (‘Nooit’), 2 (‘Een keer per maand’), 3 (‘Meerdere keren per maand’), 4 (‘Een keer per week’), 5 (‘Meerdere keren per week’), 6 (‘Eenmaal per dag’). Each of the following media channel usage was assessed: how often
16 | P a g e
participants watch the news on TV (N = 157, M = 3.89 and SD = 1.76), how often they read the newspaper (N = 157, M = 3.83 and SD = 1.72) and how often they read political blogs (N = 157, M = 2.87 and SD = 1.71).
Source credibility
The credibility of several news sources (TV, newspaper, free newspaper, online newspaper, renowned blogs and radio) from The Netherlands was assessed by an 11 point scale: 0 (‘Helemaal niet geloofwaardig’) to 10 (‘Heel erg geloofwaardig’) (N = 157, M = 5.41 and SD = 1.75).
Demographics
The demographic measures included the gender (N = 157, 59,9% male), age (N = 157, M = 28,50, SD = 10.15), level of education (N = 157, M = 6.66, SD = .75) and subjective standard of living (N = 157, M = 6.22, SD = 1.62).
Manipulation check
Right before the end, participants were asked if they read a positively or a negatively framed news article and whether the news article was from a blog or from a newspaper.
End message
The participants were thanked again for their contribution and were reminded about the contact details of the ethical commission for complaints. An open answer box was made available for their remarks and observations. The contact details of the researcher were mentioned one more time, should they need a summary of the results or should they have questions about the study. Lastly, participants were kindly asked to avoid disclosing any information from the questionnaire, to prevent priming other possible participants.
17 | P a g e
Findings
A total of 53 participants failed to correctly answer one of the questions that were verifying the manipulation (which media channel was used in the stimulus material and what was the valence of the content). Among these 53 participants, 11 participants failed to answer correctly both of the questions. A new variable was created, describing how the manipulation worked (0 – participants who failed in correctly answering both questions and 1 – participants who answered correctly both questions). This variable was included in the tested models, along with the other control measures discussed in the previous section were included in the models.
The results of the OLS regression are reported below per each dependent variable tested, in the same order as presented in the method section, including several models which test the hypotheses as follows: (1) model 1 is testing how the valence of the news article affects the attitudes and behavioral intentions (H1), (2) model 2 is testing how the interaction between valence of the article and preference for a candidate is influencing the attitudes and behavioral intentions (H2), and finally model 3 is testing how the interactions between the valence of the article, preference for a candidate and media channel is influencing the attitudes and behavioral intentions of participants (H3).
Attitude towards the article (N1 = 157; N2 = 157)
For the first measure of attitude towards the article (attitude I towards the candidate), the variance was explained by the third model as follows: Adj. R2 = .189, F (20, 136) = 2.81, p < .001, while the first model explained 10.3% of the variance (F (14, 142) = 2.28, p < .01) and the second only 9.8% (F (16, 140) = 2.05, p < .05). Nonetheless, the coefficients (see Table 2) of the predictors for each of the models were not significant. The valence of the article had no significant effect, neither the preference for a candidate, nor the media channel, thus rejecting
18 | P a g e
all the three hypotheses for the first measure of attitude towards the article. In the analyses examining the effects of the second measure of attitude towards the article (attitude II towards the candidate), the third model explained the variance as follows: Adj. R2 = .137, F (20, 136) = 2.23, p < .01, and the second model explained 11.3% of the variance (F (16, 140) = 2.24, p < .01), while the first model explained 10.8% of the variance F (14, 142) = 2.34, p < .01. Again, none of the coefficients (see Table 2) were significant in the models, thus all the three hypotheses being rejected as well for the second measure of the attitude towards the article. Moreover, the direction of the effect of valence for both outcome measures was opposite than predicted, but the direction of interaction between valence and preference was always as hypothesized. For the first measures of attitude towards the article, the control measures age, low education and media usage: blog were significant in the first two models, while in the third age and source credibility were significant. As for the second measure of attitude towards the article the age and political orientation were significant in all three models, and media usage: blog only in the first two models.
Attitude towards the candidate (N1 = 157, N2 = 136, N3 = 157)
In the analyses including first measure assessing the attitude towards the candidate (attitude I towards the candidate), the variance was explained by the third model as follows: Adj. R2 = .160, F (20, 136) = 2.48, p < .01, while the second model explained 13% of the variance (F (16, 140) = 2.45, p < .01). The first model was not a good fit for the data (F (14, 142) = 1.35, p = .18). For the second model, the preference for a candidate had a significant effect on the attitude towards the candidate (b* = .38, t = 3.49, p < .01), but the interaction which could have supported the second hypothesis was not significant (see Table 3). Thus, the candidate bias was supported in this model, participants rating more positively the candidate when they
Table 2 Effects of the valence, preference and media channel on the attitudes towards the article
ATTITUDE I TOWARDS THE ARTICLE (N = 157) ATTITUDE II TOWARDS THE ARTICLE (N = 157) Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 B S.E. B S.E. B S.E. B S.E. B S.E. B S.E. Constant 4.366** .785 4.360** .792 4.063** .773 5.546** .766 5.557** .769 5.292** .780 Control Measures
Male .303 .255 .315 .257 .159 .248 .169 .249 .190 .249 .092 .250
Age -.037** .013 -.037** .013 -.036** .013 -.035** .013 -.034** .013 -.034** .013
Education (High vs. Med.) .030 .309 .031 .310 .095 .295 -.320 .302 -.322 .301 -.267 .298
Education (Low vs. Med.) 1.403* .671 1.441* .675 1.259 .646 1.083 .655 1.149 .655 1.080 .652
Political Interest -.375 .295 -.401 .300 -.410 .285 -.395 .288 -.426 .291 -.428 .287 Voting Intention .031 .056 .029 .056 -.010 .055 .010 .055 .007 .055 -.021 .055 Political Orientation -.105 .066 -.115 .067 -.084 .064 -.168* .064 -.185** .065 -.164* .065 Source Credibility .107 .079 .108 .079 .176* .077 .109 .077 .109 .077 .145 .078 Media Usage: TV -.007 .073 -.002 .074 -.021 .071 .014 .071 .023 .071 .019 .072 Media Usage: NP -.034 .076 -.043 .077 -.042 .074 -.032 .074 -.046 .075 -.042 .075
Media usage: Blog .171* .081 .181* .082 .136 .081 .167* .079 .181* .079 .147 .082
Standard of living .038 .081 .044 .083 .030 .079 .017 .079 .029 .080 .020 .079 Manipulation check -.242 .261 -.249 .262 -.300 .257 -.472 .255 -.481 .254 -.508 .259 Predictors Valence .391 .239 .585 .331 .555 .490 .167 .233 .440 .321 .620 .495 Preference .044 .346 -.197 .438 -.008 .336 .018 .442 Media Channel .761 .456 .697 .461 Interactions Valence X Pref. -.392 .483 -.112 .664 -.544 .469 -.711 .670 Valence X Media C. -.138 .648 -.448 .654 Pref. X Media C. 1.034 .665 .253 .671
Val. X Pref. X Med. C. -.843 .921 .150 .930
Adj. R2 .103 .098 .189 .108 .113 .137
Adj. R2 Control Measures .093 .111
19 | P a g e
read about their preferred candidate than when they read about their opponent. The third model did not reveal any significant results either, rejecting the third hypothesis (see Table 3). For the second measure of attitude towards the candidate (attitude II towards the candidate), the variance was explained by the third model as follows: Adj. R2 = .313, F (20, 115) = 4.07, p < .001, the second model explaining 29.2% of the variance (F (16, 119) = 4.47, p < .001), while the first model explained 20.7% of the variance (F (14, 121) = 3.51, p < .001). The valence of the article had no significant effect on the attitude towards the candidate, rejecting the first hypothesis (see Table 3). In the second model both the preference for a candidate (b* = .43, t = 3.87, p < .001) and the interaction between valence and preference for a candidate (b* = -.27, t = -2.03, p < .05) had a significant effect on the attitude towards the candidate. Participants rated more positively the candidate when they read about their preferred candidate than when they read about their opponent, this confirming the candidate bias. The interaction confirmed the negativity bias for the preferred candidate and the candidate bias, overall. Articles about the preferred candidate were rated overall more positively than articles about the opponent. Reading negatively framed articles about the preferred candidate lead participants to rate even more positively the candidate than after reading positively framed articles about him. The negativity bias was not confirmed for the opponent. These findings support the second hypothesis. The third model did not reveal any significant effects on the attitude towards the candidate, thuis rejecting the third hypothesis (see Table 3). In the analyses for the third measure of the attitude towards the candidate (attitude III towards the candidate) the third model explained 31.5% of the variance (F (20, 136) = 4.58, p < .001), the second model explained 27.8% of the variance (F (16, 140) = 4.74, p < .001) and the first model explained 16.8% of the variance in the attitude towards the candidate (F (14, 142) = 3.256, p < .001). The first hypothesis was not supported by the model (see Table 4). In the second model, the preference for the candidate had again a significant effect (b* = .40,
Table 3 Effects of the valence, preference and media channel on the attitudes towards the candidate
ATTITUDE I TOWARDS THE CANDIDATE (N = 157) ATTITUDE II TOWARDS THE CANDIDATE (N = 136) Model 2 Model 3 Model 1 Model 2 Model 3 B S.E. B S.E. B S.E. B S.E. B S.E. Constant 4.618** .835 4.798** .845 3.885** .366 3.621** .352 3.734** .357 Control Measures
Male -.042 .271 -.081 .271 -.164 .122 -.211 .116 -.205 .116
Age -.034* .014 -.033* .014 -.029** .006 -.023** .006 -.025** .006
Education (High vs. Med.) -.089 .327 -.082 .323 -.020 .143 .001 .135 -.003 .134
Education (Low vs. Med.) .860 .711 .816 .705 .306 .303 .221 .287 .207 .285
Political Interest -.255 .316 -.270 .311 -.222 .140 -.303* .135 -.307* .134 Voting Intention -.054 .059 -.065 .060 -.051* .026 -.056* .024 -.062* .024 Political Orientation .084 .070 .102 .070 .011 .032 .013 .030 .019 .030 Source Credibility .089 .083 .141 .084 .055 .037 .065 .035 .082* .035 Media Usage: TV .066 .078 .033 .078 .056 .036 .057 .034 .044 .034 Media Usage: NP .024 .081 .021 .081 -.021 .036 -.015 .034 -.017 .034
Media usage: Blog .039 .086 .060 .088 -.022 .039 -.011 .037 -.003 .038
Standard of living -.097 .087 -.109 .086 .010 .038 -.020 .037 -.019 .037 Manipulation check .313 .276 .159 .280 .144 .123 .122 .116 .064 .118 Predictors Valence .299 .349 -.101 .535 .039 .114 .210 .149 .000 .237 Preference 1.275** .365 .723 .479 .630** .163 .379 .217 Media Channel -.393 .499 -.146 .215 Interactions Valence X Pref. -.407 .509 -.173 .725 -.450* .222 -.262 .325 Valence X Media C. .726 .708 .358 .308 Pref. X Media C. 1.417 .727 .564 .315
Val. X Pref. X Med. C. -.679 1.006 -.375 .436
∆ R2 .130 .160 .207 .292 .313
∆ R2 Control Measures .035 .213
20 | P a g e
t = 4.01, p < .001) on the attitude towards the candidate, this confirming the candidate bias. Participants rated more positively the candidate when they read about their preferred candidate than when they read about their opponent. But the interaction between preference and valence (H2) had no significant effect (see Table 4), thus rejecting the hypothesis. In third model, the preference for the candidate had also a significant effect (b* = .28, t = 2.16, p < .05) on the attitude towards the candidate, but no other coefficients were significant, again rejecting the third hypothesis. In this model as well participants rated more positively the candidate when they read about their preferred candidate than when they read about their opponent. Age had a significant effect on all three measures in all models, while political interest had a significant effect on the second and third measure for the second and third model. Voting intention had a significant effect on the second and third measure of attitude towards the candidate for all the models, while source credibility had a significant effect only on the second measure of attitude in the third model.
Voting Likelihood (N = 157)
The third model explained 18.8% of the variance in the voting likelihood for the candidate (F (20, 136) = 2.80, p < .001), the second model explaining 16.9% of the variance F (16, 140) = 2.99, p < .001, while the first model was not a good fit for the data (F (14, 142) = .50, p = .92). In the second model, the preference for a candidate had a significant effect on the voting likelihood (b* = -.52, t = -4.82, p < .001). Participants were more likely to vote for the candidate when they read about the preferred candidate than when they read about the opponent. The interaction between the valence of the article and the preference for the candidate had no significant effect on the voting likelihood (see Table 4), thus rejecting the second hypothesis. The third model revealed as well a significant effect of the preference for a
Table 4 Effects of the valence, preference and media channel on the attitudes towards the article and voting likelihood
Note: **p < 0.01, *p < 0.05.Cells contain OLS unstandardized (B) regression coefficients and standard errors (SE). Valence: Positive (1) vs. Negative (0); Preference: Preferred (1) vs. Opponent (0); Media C.: Newspaper (1) vs. Blog (0).
ATTITUDE III TOWARDS THE CANDIDATE (N = 157) VOTING LIKELIHOOD (N = 157) Model 1 Model 2 Model 3 Model 2 Model 3 B S.E. B S.E. B S.E. B S.E. B S.E. Constant 7.046** .681 6.745** .639 6.812** .641 5.165** 1.167 4.635** 1.187 Control Measures
Male -.255 .222 -.327 .207 -.397 .205 .268 .378 .289 .380
Age -.041** .011 -.034** .011 -.035** .010 -.009 .019 -.006 .019
Education (High vs. Med.) -.112 .268 -.062 .250 -.045 .245 -.117 .457 -.073 .454
Education (Low vs. Med.) .410 .583 .255 .545 .190 .535 -.120 .994 -.071 .992
Political Interest -.389 .256 -.493* .242 -.514* .236 .304 .441 .356 .437 Voting Intention -.136** .049 -.140** .045 -.157** .045 .075 .083 .073 .084 Political Orientation -.041 .057 -.022 .054 .002 .053 -.086 .098 -.111 .099 Source Credibility .072 .068 .079 .064 .126 .064 .041 .116 -.007 .118 Media Usage: TV .121 .063 .096 .059 .067 .059 .273* .108 .334** .110 Media Usage: NP -.046 .066 -.032 .062 -.026 .061 -.076 .113 -.089 .113
Media usage: Blog -.030 .070 -.029 .066 -.031 .067 .078 .120 .039 .124
Standard of living -.044 .070 -.103 .067 -.110 .065 .085 .122 .085 .121 Manipulation check .256 .227 .226 .212 .112 .213 -.157 .386 .044 .394 Predictors Valence .149 .207 .276 .267 -.183 .406 .032 .487 1.445 .752 Preference 1.119** .279 .787* .363 -2.459** .509 -1.769* .673 Media Channel -.059 .378 1.089 .701 Interactions Valence X Pref. -.379 .390 -.009 .550 .492 .711 -.732 1.019 Valence X Media C. .725 .537 -2.449* .996 Pref. X Media C. .946 .551 -1.512 1.021
Val. X Pref. X Med. C. -.785 .763 2.146 1.414
Adj. R2 .168 .278 .315 .169 .188
21 | P a g e
candidate and the voting likelihood for him (b* = -.37, t = -2.62, p < .05). Participants were more likely to vote for the candidate when they read about the preferred candidate than when they read about the opponent. Also the interaction between the valence of the article and the media channel had a significant effect on the voting likelihood (b* = -.47, t = -2.45, p < .05). Reading positively framed content in a newspaper made the participants more willing to vote for the candidate than after reading the same content in a blog, while reading negatively framed news content in a blog made the participants more likely to vote for the candidate than after reading the same content in a newspaper. The third hypothesis is only partly confirmed by this result. For the positive news content the relation is in the expected direction, but for negatively framed news content, the expectation was that the newspaper will elicit a higher likelihood for voting the candidate than the blog. This unexpected result confirms the difficulty of stating a correct hypothesis when more factors are interacting. Further, no other significant effects were revealed by this model (see Table 4). Media usage: TV was the only control measure with a significant effect on both models.
Recall (N = 157)
The variance in recall of the information was explained only by the first (Adj. R2 = .065, F (14, 142) = 1.77, p < .05) and the second model (Adj. R2 = .069, F (16, 140) = 1.71, p < .05), but in both cases a larger amount of variance was explained by the control measures (Adj. R2 = .071). The third model (F (20, 136) = 1.54, p = .07) proved to be a bad fit for the data. None of the effects were significant, thus rejecting the three stated hypotheses (see Table 5).
Conclusion and discussion
This study aimed to examine how negativity bias, candidate bias and media channels are influencing the citizens in their attitudes towards a political candidate and towards the
Table 5 Effects of the valence, preference and media channel on the recall of information (N = 157) Model 1 Model 2 B S.E. B S.E. Constant 3.445** .721 3.496** .724 Control Measures Gender .209 .235 .233 .235 Age -.012 .012 -.013 .012
Education (High vs. Medium) .264 .284 .255 .284
Education (Low vs. Medium) .081 .617 .146 .617
Political Interest .331 .271 .326 .274 Voting Intention .076 .051 .075 .051 Political Orientation -.094 .060 -.107 .061 Source Credibility .024 .072 .022 .072 Media Usage: TV .054 .067 .064 .067 Media Usage: NP .049 .070 .038 .070
Media usage: Blog -.037 .074 -.028 .075
Standard of living .050 .075 .066 .076 Manipulation check .309 .240 .308 .240 Predictors Valence -.082 .219 .073 .302 Preference -.170 .316 Media Channel Interactions Valence X Pref. -.290 .442 Valence X Media C. Pref. X Media C. Val. X Pref. X Med. C.
Adj. R2 .065 .069
Adj. R2 Control Measures .071
Note: **p < 0.01, *p < 0.05.Cells contain OLS unstandardized (B) regression coefficients and standard errors (SE). Valence: Positive (1) vs. Negative (0); Preference: Preferred (1) vs. Opponent (0); Media C.: Newspaper (1) vs. Blog (0).
22 | P a g e
news content, as well as in their voting behavior and recall of the information. Most studies earlier examined these biases at the incipient level of persuasion, namely at the information processing stage. Nevertheless, although negativity bias might prove very strong at these incipient levels, because of the different routes and factors of persuasion, models cannot predict with certainty how the final attitudes of citizens are shaped. By assessing the final attitudes and the behavioral intention of the citizens, and taking into account a series of factors that may influence their decision, this study offers valuable insight into how people are affected by their own biases across different media channel.
Negativity bias was not present in any of the models examined and therefore did not influenced any of the attitudes towards the article, the candidate or the voting intention. The discrepancy in the results might be explained by the fact that this study examined final attitudes and behavioral intentions. Despite the fact that many studies investigated and found proof of the negativity bias at the information processing level, this study found evidence for the fact that in the end citizens are not affected by this solely in their attitudes formation and behavioral intentions. If we consider the detrimental consequences of negativity discussed previously and the media asymmetry in communicating rather negatively, than positively, this finding comes as alleviation.
Notwithstanding, when we take into account the interaction between the valence of the article and the preference displayed for a candidate, matters are not so comforting. Though negativity bias doesn’t solely affect the attitudes and behavioral intentions, when analyzed in synergy with the preference for a candidate, it has the potential to be accounted for a strengthening in attitudes towards the candidate. Moreover, the negativity bias is only present when we are talking about the preferred candidate, and not about the opponent. A broader
23 | P a g e
analysis should be conducted in this area, but this result might be considered as a strong argument on how people attain polarized opinions. Supposedly, even after being exposed to counter information about their preferred candidate, citizens are reinforcing their views and opinions and they are reaching polarization. Considering the fact that the study focused on fictional candidates and participants had no prior attitudes towards them, this finding is even more striking.
Concerning the influence of the media channel in this already complicated relation between biases and attitudes, a statistically significant three way interaction supporting the third hypothesis fully was not found. Regardless, the interaction between valence and media channel was found to be responsible for a change in the behavioral intention of voting for the candidate. This interaction effect can be observed in the same direction for all the dependent variables measuring the attitude towards the candidate and the voting intention, though the results are not always statistically significant. While reading positive information in the newspaper elicits more favorable attitudes than reading negative information, the contrary happens for the blog. Albeit the considerations that the online media channels are raising the political efficacy and are more suitable to suppress a negativity bias, this study doesn’t find evidence in this sense. The online media channels are more prone to reveal a negativity bias, than are the offline ones. The results for the influence of media channels over the attitudes and behavioral intentions are different than expected. The reasons behind these unexpected results must be explored in further studies. Though a clear explanation of why channel differences impact the attitudes in the way revealed by this study cannot be momentarily given, the implication of these preliminary results can be distressing. Taking into account the high popularity of online media channels and citizens consuming more and more online news, the negativity bias in this context can have detrimental consequences.
24 | P a g e
As suggestions for future studies, channel differences have to be considered further; studies involving various media channels exploring how the negativity and candidate bias work would bring an important added value to the field. The attitudes towards the article moderating the relation between - the valence of the article, the preference for a candidate and the media channel - and attitudes towards the candidate might be explored in a new study. Other moderating factors between the predictors and outcomes might be taken into account, such as source credibility or political orientation. One of the major limitations of the current study might be recognized in the inappropriate sampling method. A future study might consider evaluating a representative sample of the Dutch population. The writing style of the blog article was inspired by the most read blog in The Netherlands, but nonetheless this style is quite uncommon for a regular Dutch news outlet. The manipulation did not work on a considerable part of participants, especially for the valence of the article. Keeping the same story line and framing it positively or negatively proved to be challenging. Future studies might consider improving the stimulus material.
Conjointly not as obvious and as straight forward as expected, negativity does affect the attitudes. The synergies between negativity bias, candidate bias and media channels are revealing interesting effects on the attitudes and behavioral intentions. As presented previously, people are often prone to biases, even on subconscious level; therefore we can expect negativity to thrive in the future and to inflict its effects on citizens. Disseminating the results of the current study, as well as similar studies, might make citizens aware of their own biases and might convince them considering these when making important decisions.
Reference list
Ansolabehere, S., & Iyengar, S. (1996). Going Negative: How Political Advertisements Shrink and Polarize the Electorate (Review). The Public Opinion Quarterly, 60, 456– 461.
Aragones, E. (1997). Negativity Effect and the Emergence of Ideologies. Journal of Theoretical Politics. http://doi.org/10.1177/0951692897009002003
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad Is Stronger Than Good. Review of General Psychology, 5(4), 323–370. http://doi.org/10.1037//1089-2680.5.4.323
Bentivegna, S. (2002). Politics and new media. In Handbook of new media (Vol. 48, pp. 50– 61).
Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50, 191– 214. http://doi.org/10.1146/annurev.psych.50.1.191
Cacioppo, J. T., Petty, R. E., & Morris, K. J. (1983). Effects of need for cognition on message evaluation, recall, and persuasion. Journal of Personality and Social Psychology, 45(4), 805–818. http://doi.org/10.1037/0022-3514.45.4.805
Cappella, J. N., & Jamieson, K. H. (1996). News Frames, Political Cynicism, and Media Cynicism. The ANNALS of the American Academy of Political and Social Science, 546(1), 71–84. http://doi.org/10.1177/0002716296546001007
Donsbach, W. (1991). Exposure to Political Content in Newspapers: The Impact of Cognitive Dissonance on Readers’ Selectivity. European Journal of Communication, 6(2), 155– 186. http://doi.org/10.1177/0267323191006002003
Gaddie, K. C. (2002). 'through which channel, with what effect': the impact of channel on voter learning and candidate evaluation from political messages. The University of Oklahoma.
Garramone, G.M.Atkin, C. K. (1990). Effects of Negative Political Advertising on the Political Process. Journal of Broadcasting & Electronic Media, 34(3), 299–311. http://doi.org/10.1080/08838159009386744
Ito, T. a, Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998). Negative information weighs more heavily on the brain: the negativity bias in evaluative categorizations. Journal of Personality and Social Psychology, 75(4), 887–900. http://doi.org/10.1037/0022-3514.75.4.887
Johnson, T. J., & Kaye, B. K. (2000). Using is believing: The influence of reliance on the credibility of online political information among politically interested Internet users. Journalism & Mass Communication Quarterly, 77(4), 865–879.
http://doi.org/10.1177/107769900007700409
Klein, J. G., & Ahluwalia, R. (2005). Negativity in the Evaluation of Political Candidates. Journal of Marketing, 69(1), 131–142. http://doi.org/10.1509/jmkg.69.1.131.55509 Lau, R. R. (1982). Negativity in political perception. Political Behavior, 4(4), 353–377.
http://doi.org/10.1007/BF00986969
Lau, R. R., & Rovner, I. B. (2009). Negative Campaigning. Annual Review of Political Science, 12(1), 285–306. http://doi.org/10.1146/annurev.polisci.10.071905.101448
Lengauer, G., Esser, F., & Berganza, R. (2012). Negativity in political news: A review of concepts, operationalizations and key findings. Journalism, 13, 179–202.
http://doi.org/10.1177/1464884911427800
Livingstone, S. (1999). New Media, New Audiences? New Media & Society, 1, 59–66. http://doi.org/10.1177/1461444899001001010
Lupfer, M. B., Weeks, M., & Dupuis, S. (2000). How Pervasive is the Negativity Bias in Judgments Based on Character Appraisal? Personality and Social Psychology Bulletin, 26, 1353–1366. http://doi.org/10.1177/0146167200263004
Lyengar, S., Norpoth, H., & Hahn, K. S. (2004). Consumer Demand for Election News: The Horserace Sells. Journal of Politics, 66(1), 157–175. http://doi.org/10.1046/j.1468-2508.2004.00146.x
Martin, P. S. (2008). The mass media as sentinel: Why bad news about issues is good news for participation. Political Communication, 25(2), 180-193.
McLuhan, M. (1964). The Medium is the Message. In Understanding Media: The Extensions of Man (pp. 7–21). http://doi.org/10.1038/2251082a0
Meffert, M. F., Chung, S., Joiner, A. J., Waks, L., & Garst, J. (2006). The Effects of
Negativity and Motivated Information Processing During a Political Campaign. Journal of Communication, 56(1), 27–51. http://doi.org/10.1111/j.1460-2466.2006.00003.x Meirick, P. C., & Nisbett, G. S. (2011). I Approve This Message: Effects of Sponsorship, Ad
Tone, and Reactance in 2008 Presidential Advertising. Mass Communication and Society, 14(March 2015), 666–689. http://doi.org/10.1080/15205436.2010.530381 Meraz, S. (2009). Is There an Elite Hold? Traditional Media to Social Media Agenda Setting
Influence in Blog Networks. Journal of Computer-Mediated Communication, 14(3), 682–707. http://doi.org/10.1111/j.1083-6101.2009.01458.x
O’ Reilly, T. (2004). The Architecture of Participation.
Pratto, F., & John, O. P. (1991). Automatic vigilance: the attention-grabbing power of negative social information. Journal of Personality and Social Psychology, 61(3), 380– 391. http://doi.org/10.1037/0022-3514.78.6.1024
Redlawsk, D. P. (2001). You Must Remember This: A Test of the On-Line Model of Voting. The Journal of Politics, 63(1), 29–58. http://doi.org/10.1111/0022-3816.00058
Rozin, P., & Royzman, E. B. (2001). Negativity Bias, Negativity Dominance, and Contagion. Personality and Social Psychology Review, 5(4), 296–320.
http://doi.org/10.1207/S15327957PSPR0504_2
Shoemaker, P. J. (1991). Gatekeeping. Communication concepts, 3.
Shoemaker, P. J. (1996). Media Gatekeeping. In An Integrated Approach To Communication Theory and Research (p. 591).
Singh, R., & Boon Pei Teoh, J. (2000). Impression formation from intellectual and social traits: evidence for behavioural adaptation and cognitive processing. The British Journal of Social Psychology / the British Psychological Society, 39 Pt 4, 537–554.
http://doi.org/10.1348/014466600164624
Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in impression formation: A review of explanations. Psychological Bulletin, 105(1), 131–142.
http://doi.org/10.1037/0033-2909.105.1.131
Soroka, S., & McAdams, S. (2015). News, Politics, and Negativity. Political Communication, 32(1), 1–22. http://doi.org/10.1080/10584609.2014.881942
Soroka, S. N. (2014). Negativity in democratic politics: Causes and consequences. Cambridge University Press.
Trussler, M., & Soroka, S. (2014). Consumer Demand for Cynical and Negative News Frames. The International Journal of Press/Politics, 19(3), 360–379.
http://doi.org/10.1177/1940161214524832
Van Dalen, A., & Van Aelst, P. (2014). The Media as Political Agenda-Setters: Journalists’ Perceptions of Media Power in Eight West European Countries. West European Politics, 37(1), 42–64. http://doi.org/10.1080/01402382.2013.814967
APPENDIX 1 Stimulus materials – Candidates’ statements
Thomas Dijkstra
Thomas Dijkstra is van mening dat veiligheid het meest belangrijk is voor onze samenleving. Hij vindt dat iedereen zich veilig moet kunnen voelen. Zonder veiligheid heb je niet de kans om je talenten te ontdekken en te ontwikkelen. Daarom is Thomas van mening dat de overheid hoge prioriteit moet geven aan het garanderen van veiligheid. Thomas zal alle maatregelen nemen die nodig zijn om de veiligheid van de bevolking te garanderen. Thomas gelooft dat het een overheidstaak is zuinig te zijn, en geld alleen uit te geven aan zaken die echt nodig zijn voor het land. Op deze manier is een duurzame toekomst voor het land mogelijk. Wat betreft de stijgende zorgkosten ziet Thomas dat het systeem van de gezondheidszorg zwaar onder druk staat. Hij wil er alles aan doen om te voorkomen dat het systeem onder deze druk bezwijkt. Daarom neemt hij maatregelen om de kosten beheersbaar te houden, zodat goede zorg voor iedereen beschikbaar blijft. Ook in de toekomst. Wat betreft het milieu gelooft Thomas niet dat het zin heeft om extra milieuregels op te leggen of met subsidies te strooien. Bedrijven zijn zelf al bezig met milieubewust ondernemen. De grote stappen waar de natuur écht iets aan heeft komen van de mensen en bedrijven zelf.
Pieter de Groot
Pieter de Groot is van mening dat te veel bezuinigingen zal leiden tot de ineenstorting van de gezondheidzorg. De reeds toegepaste bezuinigingen hebben enorme gevolgen. Kwetsbare mensen worden hard getroffen. Pieter stelt dat de werkloosheid hoog blijft, vooral onder jongeren, jongeren met een migrantenachtergrond en ouderen. Hij gelooft dat de overheid moet investeren in nieuwe bedrijvigheid en meer banen, in plaats van de economie kapot te bezuinigen. Er zijn voldoende werkplekken als deze eerlijk worden verdeeld onder de bevolking: een vierdaagse werkweek moet de norm worden. Met een duurzame economie kan het land sterker worden. Met betrekking tot het milieu is Pieter van mening dat de overheid verantwoordelijk is voor de milieuwetgeving. Dit kan niet worden overgelaten aan de bedrijven zelf. Strenge regelgeving en handhaving vanuit de overheid moet de bevolking beschermen. Klimaatverandering is geen mythe meer. De regering moet hoge prioriteit geven aan groene energie en investeren in het milieu. Pieter gelooft dat veiligheid belangrijk is, maar is van mening dat de NAVO zich niet moet ontwikkelen richting een agressieve interventie macht. Dit draagt niet bij aan de veiligheid van het land. Het is de verantwoordelijkheid van de Verenigde Naties te bemiddelen bij internationale conflicten.
