Attitudinal polarization in an experimental survey.
Master's Thesis Graduate School of Communication Master's programme Communication Science University of Amsterdam
Eline Kimmel 11995211
Supervisor: Dr. Mr. J.H.P. van Spanje Word count: 7993 29th of June, 2018
in attitudinal polarization on the position issue of immigration. By situating these in a fabricated media message, the alleged polarizing nature of the current media environment is taken into account. This type of research is important as it provides insights into the hitherto
under-researched role the argumentation source plays in political persuasion. In addition, it can uncover how persuasion works for different groups. The research question of this study is: in a multiparty
system, to what extent does exposure to varying arguments from varying expected sources in a
news media message lead to attitudinal polarization? To answer this question, an experimental
survey was designed where respondents were randomly assigned to one of four versions of a fabricated news article on the topic of immigration (N = 240). The conditions varied in pro- and counter-attitudinal arguments in combination with an expected or unexpected source making these. Results showed that what mattered most for attitudinal polarization was the prior position of respondents. The pro-immigration group depolarized regardless of their condition. Contrarily, the anti-immigration group did not significantly alter their views save for the ones placed in the pro-attitudinal, unexpected source condition. This group did become more negative on
immigration, thereby showing attitudinal polarization. Future research should include the element of source expectedness and acknowledge prior positions need to be taken into account for persuasion attempts. Ideally, longitudinal studies are carried out to track the effects of arguments and their sources on attitudes over time. Only then can meaningful patterns when it comes to attitudinal polarization be established and, if desired, countered.
KEYWORDS: Attitudinal polarization, source expectedness, pro- and counter attitudinal arguments, media environment, persuasion
1
Introduction
Important aspects of a well-functioning democracy include the ability to engage in political dialogues and cooperate with others. It is precisely this ability that has been said to be under threat in many Western democracies. The causes of this erosion are multiple and interconnected, with authors mentioning the fragmentation of political systems (Pildes, 2014), the influence of tribal politics (Chua, 2018; Sullivan, 2017), elite polarization (Kimball & Gross, 2004) and political polarization (Körösényi, 2013). Lamenting the supposed decline of democracy is a time-honored tradition (see for instance Bever 2017). Nonetheless, there is an argument to be made that the problems are especially dire today. For instance, in the United States political
polarization, the predictable political positioning along rigid party lines, has become the norm. Scholars such as Rae (2007) have seen this result in congressional gridlock and an overall coarsening of the political discourse which inhibits political collaboration.
The contributing role the media play in this contemporary weakening of dialogue and cooperation in democratic political systems has not been ignored. Authors such as Bennett and Iyengar (2008), Gruzd and Roy (2014) and Prior (2007) point to the specific affordances of technologies such as broadcast (cable) television, the internet and social media. They note that the abilities of these platforms constrain rather than promote discussion and collaboration. Instead of audience members being subjected to a multiple viewpoints, the abundance of media options now results in so-called 'echo chambers'. 'Echo chambers' is a term to indicate that the self-selecting ability of citizens results in choices for content that amplifies and 'echoes back' their (political) opinions rather than challenges them (Colleoni, Rozza & Arvidsson, 2014). Political persuasion becomes more difficult and unlikely in this media environment. It gets easier for people to stick to their (political) dispositions due to their politically reinforcing media habits
2
(Vergeer, 2014). Moreover, these opinions can become more extreme due to the lack of
encountering counter-arguments. In such an environment, entrenchment and the expectedness of political opinions becomes more influential as shortcuts to base one's opinion on (Barber & Pope, n.d.).
The extremity of political opinions, the lack of dialogue and cooperation and the diminished likeliness of political persuasion have important implications for the workings of democracy. One could argue this is especially so for multiparty systems, where governing generally happens in a coalition. In other words, research focused on to what extent opinions become more extreme, a process which is called attitudinal polarization, in multiparty settings becomes all the more important. When connecting this topic to the role media play in
democracies, the question arises what the effects are of encountering reinforcing arguments (as is the norm in the current media environment) on attitudinal polarization. Moreover, what happens to attitudes when the opposite, counter-attitudinal information, is encountered instead? An additional question is if it matters if who is making the arguments is expected or unexpected. These two factors, the type of arguments and the expectedness of the source, are combined in the following research question of this thesis:
In a multiparty system, to what extent does exposure to varying arguments from varying expected
sources in a news media message lead to attitudinal polarization?
The main academic relevance of this research question lies in two different areas. Firstly, research on source expectedness has been scare in political communication, especially when connected to varying arguments and attitudinal polarization. This research therefore hopes to diminish this existing gap in the literature and provide important insights into what this
3
different types of arguments, a better understanding of the potentially polarizing effects of an opinion-reinforcing media environment can be achieved. Secondly, much of the empirical research on attitudinal polarization and persuasion has come from the United States. This
country, with its two-party system, may be especially susceptible to the effects of attitudinal and political polarization. Multiparty systems, with stronger traditions of cooperation and dialogue, might differ substantially in this regard. This begs the question how well results and information from an American context travels to multiparty systems from a comparative perspective. This is especially relevant as there are indications that parts of the electorate in European multiparty systems are becoming more extreme and consolidated in their political opinions (Munson & Resnick, 2010; Henley, Bengtsson & Barr, 2016). Therefore, this research wants to contribute these insights to the larger academic debate on political persuasion in this day and age.
For societal reasons this research is important because it can provide insight if attitudinal polarization is taking place. If so, this can be considered a warning sign for multiparty systems due to aforementioned detrimental effects on political dialogue and cooperation. Moreover, if combinations of factors are found to result in less attitudinal polarization, this can be of practical value for anyone who wishes to combat these effects. Lastly, this research acknowledges that persuasion does not work uniformly for everyone. Hereby insights can be gained for whom priorities should be set to combat attitudinal polarization if this is desired.
Theoretical Framework Attitudinal polarization
Following the influential study by DiMaggio, Evans and Bryson (1996), attitudinal polarization can be defined as "the extremity and distance between responses" of different
4
opinions (p. 693). Often, this concept has been used to refer to the process of this distance between diverging positions increasing over time (Hart & Nisbet, 2011). Attitudinal polarization is also known under the term 'opinion polarization' (Mouw & Sobel, 2001).
Most research on this topic stems from an American context, where the process has been linked to political polarization. This is the process by which Americans become more extreme and consolidated in their political opinions (Abramowitz & Saunders, 2005). The connection with attitudinal polarization is a logical one: political polarization does not happen without attitudinal polarization taking place firstly.
Contrarily, in European studies the concept has mostly been used to refer to attitudinal changes on specific issues, such as climate change (McCright, Dunlap & Marquart-Pyatt, 2016), minority issues (Evans & Need, 2002) or income disparity (Pfeifer, 2009). The research
conducted in this thesis finds itself in line with the European focus by centering on measuring attitudinal polarization on one specific issue.
There are scholars who claim that attitudinal polarization is aggravated by the current media environment (see for instance Bennett & Iyengar, 2008). The theoretical framework now turns to this link.
The current media environment and attitudinal polarization
The ability of the media to 'narrowcast' has been seen as the main reason why the current media environment aggravates attitudinal, and by extension, political polarization. As McManus (1995) shows in his market-approach to news production, media channels are in the business of selling attention of audience members to advertisers. Therefore, they used to try and appeal to as broad an audience as possible to become more attractive to these advertisers. Croteau, Hoynes and Milan (2012) refer to 'catch-all media' to denote this past aim of capturing as large an
5
audience as possible. In order to achieve this aim, media channels tended to stay in the safe political middle as to not alienate any potential audience members.
However, with the advent of cable television and later the internet, the amount of media outlets on offer increased dramatically. This provided audience members with the possibility to self-select themselves into groups by seeking out content in line with their dispositions. Some media producers, aware of this tendency, began to 'narrowcast', or cater, to these specific groups. This means that rather than focusing on appealing to as many audience members as possible, it became feasible and profitable to focus on one specific sub-section of the audience (Liu, 2010; Overby & Barth, 2006). Often, these sub-groups have differentiated themselves on the basis of political orientation (Jones, 2002) which has implications for the type of content they encounter. Levendusky (2013) for instance found that media channels that narrowcast to a specific
politically partisan audience polarize these audience members' views further.
The prevalence & persuasiveness of pro-attitudinal arguments in the current media environment
In this present narrowcasting media environment, pro-attitudinal arguments are more prevalent and persuasive. Pro-attitudinal arguments are arguments which are in line with already-established beliefs on a topic. An economic incentive is the main reason why this type of
argumentation is more prevalent. Media channels have to retain the attention of the
(narrowcasted) audience to appeal to advertisers to make a profit. Research has long shown that people tend to dislike encountering views and opinions that go against their already-established beliefs (Harmon-Jones, 2002). These views are also called counter-attitudinal arguments as they go against currently-held opinions. What is more, audience members actively try to avoid such content if possible (Festinger, 1954). In other words, the repellant nature of having counter-attitudinal messages is to be avoided in media programming if an audience is to be retained as
6
shown by Resnick, Garrett, Kriplean, Munson and Stroud (2013). This is why echo chambers come about: people tend to self-select into (politically) homogenous groups when it comes to media content which is now made possible by the narrowcasting media environment (Vergeer, 2014). In addition to the general dislike of encountering non-congruent information, people also like being validated in their beliefs (Colleoni, Rozza & Arvidsson, 2014). Thus, pro-attitudinal arguments and viewpoints become the norm in media content while counter-attitudinal
information is to generally be avoided to maintain an audience.
Apart from being more prevalent, pro-attitudinal arguments are also more persuasive. Research has shown that persuasion efforts are often met with resistance which diminishes the effectiveness of persuasion attempts (Zuwerink & Devine, 1996). What is more, resistance to persuasion can even lead to more certainty about previously-held beliefs (Tormala & Petty, 2002). This resistance means that even when opposing viewpoints are encountered, audience members can employ defense mechanisms to downplay, counteract or adjust this information (see for instance Aronson, 1969 and Hall, 2006). Therefore, counter-attitudinal arguments run into more mental blockades than pro-attitudinal ones (Wheeler, Briñol & Hermann, 2007). Differently put, psychologically it is easier to give into predispositions and already-held beliefs which is why pro-attitudinal arguments are more persuasive (Vergeer, 2014).
Position issues and attitudinal polarization
In politics there are different types of issues. Valence issues are those issues that every voter and political party will generally agree upon (e.g. increasing prosperity) (Macdonald & Rabinowitz, 1998). When it comes to these valence issues, there is competition and disagreement about whom or which party is best at securing this goal. On the substance of the goal itself, however, everyone is in agreement.
7
In contrast, position issues are those political topics on which different political parties and leaders can hold very different opinions on a substantive level, as can voters (Stokes, 1963). Examples of position issues would be abortion (in a current US context), redistribution policies and immigration. These are the issues looked at with attitudinal polarization. As Edvardsson and Vegelius (1975) explain, inherent in looking at attitudinal polarization is the focus on the lack of consensus. In other words, the possibility of having very different opinions on a topic is
necessary. Therefore, for this thesis a position issue will be selected accordingly.
Variety of source making varying arguments
When it comes to the persuasiveness of different types of arguments in media messages, another relevant dimension is the specific source of the argumentation. For instance, people we know are deemed more persuasive than total strangers (Hermida, 2012). In politics too, the source of the argumentation matters: we tend to be more easily persuaded by political leaders with whom we have agreed previously than by those with whom we have not (Chaiken, 1980). For this particular research, the main relevant theoretical strand on political source is the one tied to expectedness and its accompanying persuasiveness.
This strand can be tied to the political source of the argumentation and the expectedness of the stance on a position issue. When a person takes a particular stance on a position issue, it is likely that this person is quite strongly against the opposite side of this issue (Enelow & Hinich, 1982). This has implications for the persuasiveness of the source making arguments. One of these implications is that when a source makes opposing arguments to a person's stance on a position issue, this source will probably be disliked by the person and internal resistance to these arguments commences. Nai (n.d.) explains that opposing arguments from an expected disliked source serves as a form of resistance which polarizes a person's views further. In other words,
8
when a party (leader) who is expected to make opposing arguments to a person's stance on a position issue, does exactly this, this person becomes more confident in his/her assertion that he/she is actually on the right side. Contrarily, if pro-attitudinal arguments are encountered from an expected source, people become more confident in their assertion that their stance on the position issue is the right one. Differently put, being in the choir that is preached to, is comfortable, easy and polarizing (Barber & Pope, n.d).
However, little is known about what happens to political attitudes when arguments from a source are encountered who was expected to be in opposition of a person's stance but is actually agreeing with that person's position. Likewise, what happens to attitudes when
counter-attitudinal arguments are encountered from a source that a person (wrongly) expected to be in line with his/her position and hence tended to like? Research from marketing suggests that this source unexpectedness can catch people off-guard, whereby they become more open to the persuasiveness of counter-arguments (Haynes, 1992; Shuper & Sorrentino, 2004). In short, source (un)expectedness is important when it comes to the persuasiveness of arguments.
The hypotheses
Based on the previous paragraphs, the following hypotheses can be formulated. First of all, pro-attitudinal arguments are both prevalent and persuasive in the contemporary media environment. This persuasiveness is amplified when the arguments are delivered by an expected source that is in alignment with one's views. When applied to a specific position issue, this will thus likely polarize attitudes:
1a: After participants are exposed to pro-attitudinal arguments from an expected political source, they will display higher levels of attitudinal polarization.
9
Less certain is what happens when pro-attitudinal arguments are encountered from an unexpected political source. Notwithstanding, the persuasiveness of the pro-attitudinal
information will likely still lead to more attitudinal polarization as it trumps the unexpectedness of the source:
1b: After participants are exposed to pro-attitudinal arguments from an unexpected political source, they will display higher levels of attitudinal polarization.
The following hypothesis on the difference in polarizing influences of source
expectedness versus unexpectedness should be treated as an exploratory one. There is not enough research on this topic in connection to political persuasion to be overly confident in stating which (an expected or an unexpected source) polarizes attitudes more. Nonetheless in this thesis the assumption is that the unexpectedness of the source will lead to a disruption in peoples' thought processes which could diminish persuasiveness. Therefore, the increase in attitudinal polarization when encountering pro-attitudinal arguments from an unexpected source will not be as high as with an expected source:
1c: After participants are exposed to pro-attitudinal arguments from an expected political source, they will display higher levels of attitudinal polarization than those who have been
exposed to pro-attitudinal arguments from an unexpected political source.
With counter-attitudinal arguments, the resistance that ensues can polarize peoples' attitudes further. This is especially the case when these arguments are delivered by an expected oppositional source, thereby increasing resistance:
2a: After participants are exposed to counter-attitudinal arguments from an expected political source, they will display higher levels of attitudinal polarization.
10
Less clear is what happens when these counter-attitudinal arguments are delivered by an unexpected political source. On the one hand, resistance to the information is to be expected but on the other hand the unexpectedness can disrupt this resistance. In addition, the unexpected source will likely be looked on more favorably than an expected source making these arguments. This seems to indicate people might come to doubt their previous position more or are at least more willing to entertain the other side to the position issue:
2b: After participants are exposed to counter-attitudinal arguments from an unexpected political source, they will display lower levels of attitudinal polarization.
Moderator: political sophistication
An important moderator in political communication is political sophistication. Political sophistication is a concept that has been defined in many different ways, although all have to do with individuals being attuned to politics in some manner. Luskin (1990) for instance outlines five strict criteria in his 'sophistication equation' namely, education, knowledge, interest,
intelligence and occupation (p. 335-336). Other scholars see this as too rigid and opt for broader definitions. Karp, Nai and Norris (2018) argue that involvement with politics and knowledge about political processes are enough to define sophistication. Huffmon (2003) too notes that crucial to any definition is a knowledge component and an interest one. Following these authors, in this thesis, political sophistication is defined by political knowledge in combination with political interest.
Political sophistication is an influence to be reckoned with when it comes to persuasion. This is because researchers have found that those higher in political sophistication are generally less easily persuaded (Lau & Redlawsk, 2001) and are better equipped to resist persuasion attempts (Petty & Cacioppo, 1979). Therefore, the following hypothesis can be formulated:
11 3: The higher the level of political sophistication, the smaller the effects of the exposure on attitudinal polarization.
Method Research design
An experimental design best serves the purpose of this thesis. This is because through this method, it can be asserted with more certainty that changes in attitudinal levels on immigration are the result of exposure to stimuli (Hayes, 2008). Hereby, a higher internal validity is ensured (Bernard, 2013). Moreover, an experiment can tightly control exposure to the varying levels of the variables (varying arguments and varying expected sources) this study is interested in. In addition, the data collected in an experimental survey are suitable for statistical testing, which is paramount in confirming or rejecting the hypotheses laid out in the theoretical framework.
The experiment has a 2x2 factorial design. The first factor of varying arguments is a between-subjects variable, with two levels (pro-immigration and anti-immigration). It is
important to note that for some respondents, pro-immigration arguments are pro-attitudinal while for others this is counter-attitudinal. It thus depends on the specific respondent if they saw pro- or counter-attitudinal arguments. The second factor of varying expected sources is also a between-subjects variable, with two levels (expected and unexpected political source). Of interest to this research is the separate influence of combinations of these factors. This is why it is important to have a between-subjects design. Only then can each respective influence in isolation of each other be captured. Exposing respondents to multiple treatments might confound these results (Charness, Gneezy & Kuhn, 2012). As this research can be classified as an effects
12 Table 1: conditions of the research design
study, the pre-test on measuring the attitudes on the policy topic of immigration serves as the control group for each condition. Hence, no separate control group is included. The experimental design of this thesis can be visualized in the following table:
Varying arguments
Varying expected sources Pro-immigration Anti-immigration
Expected political source Condition 1 Condition 3 Unexpected political source Condition 2 Condition 4
The experiment has been supplied to participants via an internet survey. This method has been chosen as it is cost and time effective. In addition, it places the lightest burden on partakers. This is because the survey can easily be filled in from any location with an internet connection which, in the age of smartphones, can be anywhere. Moreover, the comfort of respondents being able to choose the time and location makes for a less artificial testing environment which
improves the external validity of the research (Babbie, 2011). The survey was accessible from the eighteenth of May until the fourth of June. The survey utilized an opt-in design where
respondents decided themselves if they wanted to participate by taking the survey via a universal link.
Selection of multiparty system
The Netherlands has been chosen as the context in which the experiment is set. This has been done for a number of reasons. Firstly, the Netherlands has always had a multiparty political system which is the system this research is interested in. Secondly, the Netherlands serves as a least-likely case for attitudinal polarization to flourish due to its long-standing tradition of
13
coalition building and dialogue (Keune, 2016). Thirdly, it is a novel and very different context from the normally used American one, which can provide new theoretical insights. Lastly, from a practical viewpoint, the Netherlands is a context in which enough data can be gathered due to the personal background of the researcher.
Selection of respondents and their characteristics
Respondents have been selected through convenience and snowball sampling. Extra consideration has been given to include respondents with varying levels of political
sophistication. This was done to go beyond the rather standard university-students convenience sample that is often disproportionally politically sophisticated. These respondents were then randomly assigned to one of the four conditions as outlined in Table 1.
There were four criteria respondents had to satisfy to be eligible for participation: they had to be eighteen or older, they had to give their explicit consent to participate, they had to have the Dutch nationality and the Netherlands had to be the primary country of residence. This was done to ensure the study captures attitudes of Dutch respondents who are situated in the daily Dutch political context. Dutch was deliberatively chosen as the language of the survey to ensure respondents from all (educational) backgrounds could participate.
Selection of position issue
In a Dutch context, the position issue of 'immigration' is an appropriate one for the following reasons. First of all, immigration has been a long-standing polarizing issue for many people (van der Veer, 2006). This means many neutral opinions are relatively unlikely, thus attitudinal polarization can be observed. Secondly, the issue has also polarized political parties (Akkerman & Hagelund, 2007). This indicates necessary opposing positions from political sources can be identified. Moreover, some Dutch political parties and by extension their leaders
14
are associated with a specific position on immigration making the expectedness of their stance on immigration high. In short, both criteria for having varying expected sources and varying
arguments on a topic on which people are likely to have a certain opinion are satisfied by selecting immigration as the position issue in this experimental design.
Operationalizations of variables
The four conditions were operationalized in a media message that takes the form of a fictional newspaper article. This has as a benefit that all elements can be completely controlled. Moreover, the articles had to be fictional as there are simply no real life articles where political sources take a position opposite of what is expected. This research utilized four versions of the same article, one for each condition as seen in Table 1.
Varying arguments have been narrowed down by limiting it to pro- and anti-immigration arguments. For those in favor of immigration in a Dutch context, pro-immigration arguments are pro-attitudinal. For those against immigration in a Dutch context, pro-immigration arguments are counter-attitudinal. Of course for anti-immigration arguments, this situation is reversed. All versions of the article included:
The stance of the source in the headline
A short introduction clarifying the position of the source in question
One statistic that backs up the position of the source
A total of three arguments made by the source in question
A closing sentence on the (un)expectedness of this politician's position
When it comes to varying expected sources, two politicians are needed who are well-known (or there can be no expectedness) and are on opposite sides of the immigration issue. Importantly, the unexpected position of the source also had to be at least plausible. This is why
15
two sources that played a prominent role in the political past of the Netherlands have been selected. This distance adds to the plausibility that the encountered article relays previously unknown information about these past figures in the unexpected conditions. For these reasons, Femke Halsema, former leader of the Green Left party, and Pim Fortuyn, former leader of his own right wing party (LPF), have been selected. Not only are these politicians representative of the left-right divide that generally exists on this issue in the Netherlands (van Heerden, de Lange, van der Burg & Fennema, 2014), they are commonly associated with a certain expected position on immigration. An expected position for Pim Fortuyn would thus be to voice opposition to immigration. An unexpected position would be for him to draw attention to the more beneficial sides of immigration. For Halsema, this is (logically) reversed.
To make the article resemble a news item better, the publication of a (non-existent) biography on respectively Halsema and Fortuyn was conjectured to be the reason for the article. In this biography, either unexpected or unexpected argumentation for Fortuyn's or Halsema's position on immigration came to the foreground which constituted the content of the article. All other elements were held constant.
The dependent variable, attitudinal polarization, has been operationalized with the aid of batteries of questions relating to immigration attitudes from respected publications 'The
European Social Survey' (2016; 2015), 'the NEtherlands Longitudinal Lifecourse Study '(de Graaf, Kalmijn, Kraaykamp & Monden, 2011) and the 'Friedrich-Ebert Stiftung ' (Zick, Küpper & Hövermann, 2011). These questions were measured on a seven-point scale. In order to minimize the risk of respondents becoming sensitized to the research goal, a different battery of questions was used in the post-test than in the pre-test. Attitudinal polarization has taken place when the score on the post-test is more extreme in the direction than the pre-test indicated (either
16
pro- or anti-immigration). If the score moved towards the middle position, depolarization took place.
Political sophistication in this thesis consisted of interest in and knowledge of politics. The operationalization therefore included questions on self-reported interest in politics. To measure political knowledge, an updated version of the battery of questions proposed by
Hendriks Vettehen, Hagemann and van Snippenburg (2004) was used. This battery included both questions on knowledge of the Dutch political structure as well as questions on current political affairs.
Buffer questions on the issue of the environment were added to ensure respondents were less aware of the research goal. In addition, they were exposed to a 'buffer' article on this topic. This article was an adapted version of a news message produced by CBS (2018). The four versions of the article, as well as the buffer article, can be found in Appendix A. The full list of the survey questions and their order can be found in Appendix B. The figure below summarizes the research in a model:
17 1a 3 2b 2a 1b
Figure 1: the research model
Results Randomization and manipulation checks
Randomization checks were run to see if the distribution of respondents differed
significantly among important characteristics across the four conditions. This is why for age and political orientation, two one-way ANOVAs were run and chi-square tests were conducted for gender and education level. All these tests yielded non-significant results; distributions of these
Pro-attitudinal + Expected source (condition 1) Political sophistication (moderator) Attitudinal polarization Counter-attitudinal + Unexpected source (condition 4) Counter-attitudinal + Expected source (condition 3) Pro-attitudinal + Unexpected source (condition 2) Media message on position issue 1c
18
characteristics did not differ substantially across conditions. Therefore, the randomization checks for this experiment were all passed.
Secondly, manipulation checks were carried out for the specific type of varying
argumentation (pro- or anti-immigration) and varying expected source (expected or unexpected). The data showed that those exposed to pro-immigration conditions (conditions 3 and 4), agreed to a much larger extent that they had seen arguments in favor of immigration (M = 5.68, SD = 1.32) than those who had seen anti-immigration arguments (M = 2.36, SD = 1.41). An
independent-samples t-test, revealed this difference was statistically significant, t(244) = -19.10,
p <.001, 95% CI [-3.67, -2.98].
Moreover, the data showed that those who had been exposed to an unexpected source (conditions 2 and 4) were higher in agreement that they did not expect this position (M = 4.89,
SD = 1.55) than those who had encountered an expected source (M = 2.38, SD = 1.50). This
difference was statistically significant as shown, by the independentsamples ttest, t(242) = -12.91, p <.001, 95% CI [-2.90, -2.13]. This result is backed up by the second check on the unexpectedness of the source as those in the unexpected conditions also agreed more strongly that the source they encountered normally held a very different position (M = 5.03, SD = 1.52) than those exposed to an expected source (M = 2.49, SD = 1.53). Again, this difference was statistically significant as demonstrated by an independent-samples t-test, t(242) = -13.02, p
<.001, 95% CI [-2.93, -2.16]. In short, the experiment carried out in this thesis passed both the
randomization and the manipulation checks.
Factor and reliability analysis of batteries of questions
The dependent variable of attitudinal polarization has been measured with a pre- and post-test on immigration attitudes. To test whether the separate items of the pre- and post-tests
19
could be justifiably combined into one scale for each test, two factor analyses were run. In the pre-test, item five was reversed (high scores indicated favorable views on immigration instead of the opposite) so this item was reversely coded. The rotated principal component analysis
demonstrated that the five pre-test immigration items can form a single scale, with an Eigenvalue of 3.20. This one factor extraction was also visible from the scree plot. The factor explains 64.0% of the total variance. All items correlated positively. The reliability of this newly created scale was high (Cronbach's alpha = .86) and deleting items did not raise this reliability, so none were. The new scale was named 'pre-test immigration total' and ranged from 1 to 7, with low scores (close to 1) indicating strong pro-immigration attitudes and high scores (close to 7) indicating strong anti-immigration attitudes (M = 3.22, SD = 1.41).
In the post-test, items three and four were reversed, so these were reversely coded before the factor analysis. The rotated principal component analysis showed that the five pre-test immigration items form one scale, with an Eigenvalue of 2.68. The scree plot supported this finding. The factor explains 53.7% of the total variance. All items correlated positively. The reliability of this newly created scale was acceptable (Cronbach's alpha = .77) and deleting items did not raise this reliability substantially. Therefore, none were. The new scale was named
'post-test immigration total' and ranged from 1 to 7. Low scores (close to 1) indicate strong
pro-immigration attitudes and high scores (close to 7) strong anti-pro-immigration ones (M = 3.91, SD = 1.14). The extraction of one factor for the pre-test and one for the post-test, whose items came from three different studies, inspires confidence in the construct validity of measuring
immigration attitudes in this thesis. In other words, that it are immigration attitudes that is being measured and not some other construct (Andrews, 1984).
20
Political interest was measured with two items. A factor analysis, in the form of a rotated principal component analysis, revealed that these could be combined into one scale, with an Eigenvalue of 1.86. The total variance explained by the factor was 93.0%. The reliability of this scale was very strong (Cronbach's alpha = .92) and no items were deleted. The name of this new scale was 'total political interest' and ranged from 1 to 7 with low scores indicating low political interest and high score indicating a high political interest (M = 4.62, SD = 1.65).
Computing new and recoding existing variables
Based on the five items measuring immigration attitudes in the pre-test, a variable of the total level of immigration attitudes was made. The mean value of this total scale (3.5) was used as the cut-off point to create a dummy variable to indicate if someone was favorable or
unfavorable towards immigration. This dummy variable (0 = pro-immigration, 1 = anti-immigration) was subsequently used to compute the dummy variable if someone had been exposed to a counter- or pro-attitudinal condition (0 = counter-attitudinal, 1 = pro-attitudinal).
The variable of political sophistication was computed by combining the variables for
political knowledge and political interest. Political knowledge (M = 2.85, SD = 1.22) was
computed by adding the five items measuring knowledge. This means respondents could score between a zero (no question answered correctly) and a five (all questions answered correctly).
Political interest, computed by combining self-reported interest in politics and the extent to
which the respondent followed political news, ranged from one (very low in political interest) to a seven (very high in political interest). Although the scaling for political interest and knowledge were not completely equal, they were added together nonetheless as it can be argued
self-reported interest outweighs the importance of factual political knowledge. The mean of this newly created variable (labeled 'total political sophistication') (M = 7.47, SD = 2.54) was used
21
to create a dummy variable of political sophistication. Respondents scoring below the mean resulted in a 0 on the dummy variable, (low in political sophistication) and above the mean a 1 (high in political sophistication).
For the hypothesis on the moderator (H3), the scores of the pre- and post-tests were combined to have one measure for attitudinal polarization (M = -0.70, SD = 1.03) to be used in a regression analysis. This variable was constructed by calculating the differences between pre- and post-test scores differently depending on the immigration attitude group (pro- and counter-immigration). This was done to ensure the scores on this new dependent variable (attitudinal
polarization score) correctly reflected either polarization or depolarization. For example, a
pro-immigration respondent scoring a 2 on the pre-test and a 3 on the post-test, scores a -1 on this variable, thereby denoting depolarization. Contrarily, an anti-immigration respondent scoring a 4 in the pre-test and a 5 in the post-test has a positive score of 1, denoting polarization. Thus, negative scores indicate depolarization and positive ones polarization.
Descriptives of the sample
Overall, 270 people started the survey. Of these respondents, 240 finished the entire questionnaire, which brings the finalization rate of this survey to 88.9% (N = 240). For a convenience sample, important characteristics were relatively well distributed. The division between men and women was almost perfect with 49.5% of respondents reported being male and 50.5% female. The average age of the respondents was 44 years of age, with 46 being the
median. The ages of 23 and 24 were overrepresented which makes the age distribution skew to the left. Nonetheless, apart from these two ages the distribution was relatively normal. When it comes to the level of education, most people reported having completed some sort of academic
22
education (34.8%) (WO), followed by those with a higher education level (30.4%) (HBO) and those with a practical education (20.2%) (MBO).
Testing the hypotheses: pro-attitudinal exposure
In order to test hypothesis 1a (pro-attitudinal, expected source) two pro-attitudinal groups were made. The first group consisted of those who were pro-immigration and had been exposed to a pro-immigration article from Femke Halsema as the expected source and the second of those who were anti-immigration and had been exposed to an anti-immigration article from Pim
Fortuyn as the expected source. Two paired-samples t-tests were run for these groups on their pre- and post-tests to see if these respondents had become more extreme (either more positive or more negative) in their respective attitudes on immigration. These tests revealed that the anti-immigration group had not become significantly more negative (p = .185). The pro-anti-immigration group did show a significant difference in their pre- (M = 2.16, SD = 0.65) and post-test (M = 3.25, SD = 0.74) but in the opposite direction than was expected. This group had become
significantly less positive in their attitudes on immigration as shown by the paired-samples t-test,
t(90) = -13.74, p <.001, 95% CI [-1.32, -0.86]. Thus, hypothesis 1a has to be rejected: those
exposed to pro-attitudinal arguments from an expected source did not show higher levels of attitudinal polarization; in fact for pro-immigration respondents, depolarization took place.
Hypothesis 1b (pro-attitudinal, unexpected source) was tested in a similar manner with the construction of the two groups. The difference was that these two groups now consisted of those who had seen a pro-attitudinal message from an unexpected source. Two paired-samples t-tests were run for both these groups. These t-tests revealed that the (pro-attitudinal)
anti-immigration group exposed to Halsema had indeed become more negative when comparing their pre- (M = 4.67, SD = 0.98) and their post-test scores (M = 5.20, SD = 0.73) and that this
23
difference was significant, t(25) = -2.68, p <.005, 95% CI [-0.94, -0.12]. Interestingly, a paired-samples t-test showed that those who were immigration and had been exposed to a pro-attitudinal message from unexpected source Fortuyn also became significantly more negative in their attitudes towards immigration based on the difference between their pre- (M = 2.47, SD = 0.61) and post-test scores (M = 3.45, SD = 0.97), t(38) = -7.46, p <.001, 95% CI [-1.25, -0.71]. Therefore hypothesis 1b is partially confirmed: those who were anti-immigration and exposed to a attitudinal message from Halsema did display attitudinal polarization. However, those pro-immigration and exposed to the same type of condition also became more negative in their attitudes towards immigration, thereby showing depolarization which was not expected.
To test hypothesis 1c (pro-attitudinal hypothesis comparing expected and unexpected
sources) the measure on the difference between the pro- and pre-test on immigration attitudes
was used. This was used as the dependent variable in an independent-samples t-test with those who were exposed to expected sources and those to unexpected sources serving as the two groups. Results showed that the differences between these groups were not statistically
significant (p = .719). Therefore hypothesis 1c has to be rejected: those who are exposed to pro-attitudinal arguments from an expected source do not display higher levels of pro-attitudinal
polarization than those who were exposed to pro-attitudinal arguments from an unexpected source.
Testing the hypotheses: counter-attitudinal exposure
Hypothesis 2a (counter-attitudinal, expected source) was tested similarly to 1a, only now selecting those who had seen a counter-attitudinal message. Two paired-samples t-tests were run on pre- and post-test scores to see if these respondents had become more extreme (either more positive or more negative) in their attitudes on immigration. These tests revealed that the
anti-24
immigration group had not become significantly more negative when comparing pre- and post-test scores (p = .341). The pro-immigration immigration group did show a significant difference in pre- (M = 2.32, SD = 0.70) and post-test results (M = 3.42, SD = 0.99), but in the opposite way than was expected, t(42) = -10.00, p <.001, 95% CI [-1.32, -0.88]. This means that hypothesis 2a needs to be rejected: those exposed to counter-attitudinal arguments from an expected source did not display attitudinal polarization. Contrarily, the pro-immigration respondents displayed depolarization by moving more towards the middle.
To test hypothesis 2b (counter-attitudinal, unexpected source) the two groups as in 2a were again used. The difference was that these two groups now consisted of those who had seen a counter-attitudinal message from an unexpected source. Two paired-samples t-tests were run for both these groups. These tests revealed that the anti-immigration group had indeed become less negative about immigration, when comparing their pre- and post-results. However, this difference was not statistically significant (p = .129). The pro-immigration group did show a significant difference in their pre- (M = 2.33, SD = 0.78) and post-scores (M = 3.32, SD = 1.07), in the expected direction, t(34) = -9.94, p <.001, 95% CI [-1.71, -1.13]. Hereby they showed lower levels of attitudinal polarization. Therefore hypothesis 2b is partially confirmed: those who were exposed to a counter-attitudinal message from an unexpected source did display lower levels attitudinal polarization but this was only true for the pro-immigration group.
Testing the hypotheses: the moderator1
1 In an earlier analysis, 'partisan strength' was included as another potential moderator. Not only was the theoretical
foundation for its inclusion not as solid as desired, the findings were insignificant. Therefore, this moderator was not enclosed in this final work.
25 0 5 10 15 20 25 30 -3,4 -3,2 -3 -2,8 -2,6 -2,4 -2,2 -2 -1,8 -1,6 -1,4 -1,2 -1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2 Num b er of c ases Polarization scores
(de)polarization scores per attitude group
Pro-immigration group Anti-immigration group
For testing hypothesis 3 (the higher the level of political sophistication, the smaller the
effects of the exposure on attitudinal polarization), the dependent variable of attitudinal polarization score is used. This variable is important as it illustrates the very different patterns
for the pro-immigration (n = 82) and anti-immigration group (n = 165) (see Figure 2), something to which the results of the hypotheses testing above already alluded to. The two different patterns that emerge, is that the pro-immigration group overall collectively depolarized quite strongly while the anti-immigration group did not change their attitudes too much or showed modest polarization or depolarization. For similar descriptive graphs for these polarization score differences for varying arguments (pro- and counter-attitudinal) and varying expected source (expected and unexpected), see the figures in Appendix C.
26
A linear regression analysis was run with this attitudinal polarization score as the dependent variable and political sophistication included as a moderator. Results showed that neither the model with the moderator (p = .082) nor the F-change for inclusion of the moderator was significant (p = .221). Therefore, hypothesis 3 needs to be rejected; being high in political sophistication does not lead to smaller effects of the exposure on attitudinal polarization. In other words, the hypothesized interaction yielded no significant effect. For a summary of the results of the hypotheses testing, see Appendix D.
Conclusion Answer to the research question
When it comes to the research question of this thesis on what leads to attitudinal
polarization in terms of varying arguments and expected sources, the most important answer that can be formulated is that the prior position of respondents matters most of all. Those who were pro-immigration and those who were anti-immigration displayed starkly different patterns in their (lack of) attitudinal polarization on the topic of immigration. In general, the
pro-immigration group became significantly less positive (or more neutral) about pro-immigration regardless of the condition they were in (thereby displaying depolarization). This can be seen from the opposite significant results from testing hypotheses 1a, 1b, 2a, 2c and from Figure 2. In contrast, the anti-immigration group did not significantly change their attitudes (as shown by the non-significant hypotheses 1a, 2a and 2b), except when they were exposed to pro-attitudinal arguments from unexpected source Halsema. This did cause their attitudes to polarize, as indicated by the significant result of hypothesis 1b. The research also showed that expectedness of source did not matter in the manner which was conjectured, as unexpected sources did not
27
lead to lower levels of attitudinal polarization than with expected sources as stated in hypothesis 1c. This is supported by figure C2 in Appendix C which also shows a lack of notable differences in polarization scores per source. In addition, political sophistication did not matter in the
strength of the effect of the exposure on attitudinal polarization as seen from the rejection of hypothesis 3. This means that being knowledgeable about or interested in politics made no difference in the strength of the effect of reading the newspaper article when it comes to polarizing immigration attitudes.
Explaining the results - limitations and implications
Before the implications of the findings are discussed it is necessary to acknowledge some of the limitations of this study that could have played a role in how these findings came about. First of all, a reason as to why the pro-immigration group became more neutral in their attitudes on immigration could have to do with social desirability. This is the effect that people are afraid to answer honestly on sensitive topics, such immigration, and instead opt for those answers that they deem more socially acceptable (Nederhof, 1985). Since the pre-test was asked in the beginning of the survey to perhaps trepid respondents, more people might have been assigned to the pro-immigration group than their true views might have warranted. In the post-test, after having been subjected to many more questions, respondents could have been less on their guard and therefore answered questions more truthfully.
A second, methodological, concern is that the groups for anti-immigration (n = 107) and pro-immigration respondents (n = 184) were relatively unequal. In other words, more people were assigned to the pro-immigration group (perhaps due to social desirability). When different conditions were compared in the t-tests, the samples for the anti-immigration groups became
28
relatively small. This resulted in relatively unequal group comparisons which might have influenced the trustworthiness of the results.
Thirdly, connecting to theory, this research was carried out at a single point in time while scholars such as Abramowitz and Saunders (2005) and Prior (2007) all note that polarization is a process that happens over time. In other words, the anti-immigration group only showing
polarization in one instance does not mean that in the real world, after lengthier exposure to media messages, their attitudes do not polarize at all.
Nonetheless, the consistency of the pattern of the pro-immigration group can indicate that when citizens are proponents of an issue, they are more open to entertaining other arguments and to moderate their opinions. Perhaps people feel that when they are proponents on an issue, they need to listen to other (even pro-attitudinal) arguments in order to formulate a defense or response by which their position (inadvertently) becomes more moderate. Contrarily, when someone is against an issue this antagonistic demeanor forecloses entertaining other thoughts. This would mean it is not as much varying arguments that need to be considered but rather the prior positions of people and the framing that is then needed when persuasion is the goal. From a practitioner's point of view, it would then be easier to frame people on an issue as being
proponents of an issue (e.g. anti-immigration framed as pro-security), if persuasion or moderation of views is the goal.
When it comes to source expectedness, this research failed to uncover a singular
consistent pattern. Nonetheless, the anti-immigration group did polarize when their position was echoed by Femke Halsema from GroenLinks, an unexpected source. Perhaps this
unexpectedness emboldened them more than an expected source could by decreasing the effects of social desirability. In other words, the feeling to have to respond in a more socially desirable
29
way on immigration was diminished when a figure such as Halsema endorsed their position. This result of the unexpected source emboldening rather than moderating peoples' position goes against the earlier findings from marketing research and is a powerful indication source unexpectedness in political persuasion should be researched in its own right.
Generalizability and future research
The sample of this research was relatively good, considering the method of snowball and convenience sampling. Nonetheless, all respondents were contacted through the personal
network of one researcher so this begs the question of the quality of the external validity, or how well the results generalize to other contexts. Moreover, the cross-sectional method of testing can mean the reliability of this research is dubious as the consistency of respondents' answers over lengthier amounts of time is not observable. Therefore, this research would benefit from replication in other contexts where immigration can be considered a position issue, for which multiple European countries could qualify. For more future research, ideally, a longitudinal panel study would be carried out on a wide variety of policy topics that include the prior stance of respondents, their media habits and the sources that their encountered argumentation normally comes from. With this research serving as a useful indication, only with such a multifaceted approach can political persuasion attempts be accurately captured and judgments made on their polarizing effects. Which, in the current media environment and with the aforementioned decline of political cooperation in many Western democracies (Council for European Studies, 2017), should not be an afterthought. Instead, this approach should be taken to heart and researched accordingly.
30
References
Abramowitz, A., & Saunders, K. (2005). Why can't we all just get along? The reality of a polarized America. The Forum, 3(2), 1-22. http://dx.doi.org/10.2202/1540-8884.1076 Akkerman, T., & Hagelund, A. (2007). 'Women and children first!' Anti-immigration parties and
gender in Norway and the Netherlands. Patterns of Prejudice, 41(2), 197-214. http://dx.doi.org/10.1080/00313220701265569
Andrews, F. M. (1984). Construct validity and error components of survey measures: A
structural modeling approach. The Public Opinion Quarterly, 48(2), 409-442. Retrieved from http://www.jstor.org/stable/2749034
Aronson, E. (1969). A theory of cognitive dissonance: A current perspective. Advances in
Experimental Social Psychology, 4, 1-34.
http://dx.doi.org/10.1016/S0065-2601(08)60075-1
Babbie, E. (2011). Introduction to social science research. Belmont, CA: Wadsworth.
Barber, M., & Pope, J. C. (n.d.). Does party trump ideology? Disentangling party and ideology in America [forthcoming]. Retrieved from
https://www.dropbox.com/s/ofh5bzwnt4ixwdj/Does_Party_Trump_Ideology%3F%20 Non-Anonnomyzed.pdf?dl=0
Bennett, W. L., & Iyengar, S. (2008). A new era of minimal effects? The changing foundation of political communication. Journal of Communication, 58, 707-731. http://dx.doi.org/10.1111/j.1460-2466.2008.00410.x
Bernard, H. R. (2013). Research design: Experiments and experimental thinking. In Social
research methods: Qualitative and quantiative approaches (pp. 90-126). Thousand Oaks:
31
Bever, L. (2017, February 18). Memo to Donald Trump: Thomas Jefferson invented hating on the media. The Washington Post. Retrieved from
https://www.washingtonpost.com/news/the-fix/wp/2017/02/17/trumps-war-with-the-
media-isnt-new-thomas-jefferson-railed-about-newspaper-lies-too/?utm_term=.90a32a66d610
CBS. (2018, May 7). Rioolwaterzuivering steeds effectiever en duurzamer. CBS. Retrieved from https://www.cbs.nl/nl-nl/nieuws/2018/19/rioolwaterzuivering-steeds-effectiever-en-duurzamer
Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39(5), 752-766. http://dx.doi.org/10.1037/0022-3514.39.5.752
Charness, G., Gneezy, U., & Kuhn, M. A. (2012). Experimental methods: Between-subject and within-subject design. Journal of Economic Behavior & Organization, 81(1), 1-8. http://dx.doi.org/10.1016/j.jebo.2011.08.009
Chua, A. (2018, February 20). The destructive dynamics of political tribalism. New York Times. Retrieved from https://www.nytimes.com/2018/02/20/opinion/destructive- political-tribalism.html
Colleoni, E., Rozza, A., & Arvidsson, 2014. Echo chamber or public sphere? Predicting
political orientation and measuring political homophily in Twitter using big data. Social
Science Computer Review, 64(2), 317-332. http://dx.doi.org/10.1111/jcom.12084
Council for European Studies (2017, April 1st). Long-term changes and the crisis of democracy in Western society. Europe Now. Retrieved from
32
https://www.europenowjournal.org/2017/01/04/long-term-changes-and-the-crisis-of-democracy-in-western-society/
Croteau, D., Hoynes, W., & Milan, S. (2012) Media/Society. Thousand Oakes, CA : SAGE Publications Inc
DiMaggio, P., Evans, J., & Bryson, B. (1996). Have American's social attitudes become more polarized? American Journal of Sociology, 103(3), 690-755.
http://dx.doi.org/10.1086/230995
Edvardsson, B., & Vegelius, J. (1975). Measures of attitudinal polarization. Scandinavian
Journal of Psychology, 16(1), 1-20.
http://dx.doi.org/10.1111/j.1467-9450.1975.tb00191.x
Enelow, J. M., & Hinich, M. J. (1982). Nonspatial candidate characteristics and electoral competition. The Journal of Politics, 44(1), 115-130. Retrieved from
http://www.jstor.org/stable/2130286
European Social Survey. (2015). Round 7 Module on Attitudes towards Immigration and their
Antecedents - Question Design Final Module in Template. London: Centre for
Comparative Social Surveys, City University London. Retrieved from
https://www.europeansocialsurvey.org/docs/round7/questionnaire/ESS7_immigration_fin al_module_template.pdf
European Social Survey. (2016). Attitudes Towards Immigration and their Antecedents - Topline
Results from Round 7 of the European Social Survey. London: Centre for Comparative
Social Surveys, City University London. Retrieved from
https://www.europeansocialsurvey.org/docs/findings/ESS7_toplines_issue_7_immigratio n.pdf
33
Evans, G., & Need, A. (2002). Explaining ethnic polarization over attitudes towards minority rights in Eastern Europe: A multilevel analysis. Social Science Research, 31(4), 653-680. http://dx.doi.org/10.1016/S0049-089X(02)00018-2
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117-140. http://dx.doi.org/10.1177/001872675400700202
de Graaf, P. M., Kalmijn, M., Kraaykamp, G., & Monden, C. W. S. (2011). Design and content
of the NEtherlands Longitudinal Lifecourse Study (NELLS). Research report. Tilburg
University & Radboud University Nijmegen, Netherlands. Retrieved from http://matthijskalmijn.nl/onewebmedia/NELLSWave1Codebook_v_1_2.pdf Gruzd, A., & Roy, J. (2014). Investigating political polarization on Twitter: A Canadian
perspective. Policy & Internet, 6(1), 28–45. http://dx.doi.org/10.1002/1944-2866.POI354 Hall, S. (2006). Encoding/decoding. In M. G. Durham, & D. M. Kellner (Eds), Media and
cultural studies (pp. 163-173), Oxford: Blackwell Publishing Ltd.
Harmon-Jones, E. (2002). A Cognitive Dissonance Theory Perspective on Persuasion. In: Dillard, J. P., & Pfau, M. (Eds.), Persuasion handbook: Developments in theory and
practice (pp. 99-116). Newcastle upon Tyne: SAGE Publications.
Hart, P. S., & Nisbet, E. C. (2011). Boomerang effects in science communication. How motivated reasoning and identity cues amplify opinion polarization about climate mitigation policies. Communication Research, 39(6), 701-723.
http://dx.doi.org/10.1177/0093650211416646
Hayes, D. (2008). Does the messenger matter? Candidate-media agenda convergence and its effects on voter issue salience. Political Research Quarterly, 61(1), 134-146.
34
Haynes, P. (1992). The impact of bias and unexpectedness on message response. International
Journal of Consumer Studies, 16(2), 167-177.
http://dx.doi.org/10.1111/j.14706431.1992.tb00510.x
van Heerden, S., de Lange, S. L., van der Brug, W., & Fennema, M. (2014). The immigration and integration debate in the Netherlands: Discursive and programmatic reactions to the rise of anti-immigration parties. Journal of Ethnic and Migration Studies, 40(1), 119-136. http://dx.doi.org/10.1080/1369183X.2013.830881
Hendriks Vettehen, P. G. J., Hagemann, C. P. M., & van Snippenburg, L. B. (2004). Political knowledge and media use in the Netherlands. European Sociological Review, 20(5), 415-424. http://dx.doi.org/10.1093/esr/jch035
Henley, J., Bengtsson, H. & Barr, C. (2016, May 25). Across Europe, distrust of mainstream political parties is on the rise. The Guardian. Retrieved from
https://www.theguardian.com/world/2016/may/25/across-europe-distrust-of-mainstream-political-parties-is-on-the-rise
Hermida, A. (2012). Social journalism: Exploring how social media is shaping journalism. In E. Siapera & A. Vegli (Eds), The Handbook of global online journalism (pp. 309-328). Hoboken: John Wiley & Sons.
Huffmon, S. H. (2003). Revisiting the role of information format in candidate evaluation: An 'update' model of evaluation. The Journal of Political Science, 31, 35-69. Retrieved from http://faculty.winthrop.edu/huffmons/Huffmon%20Role%20of%20Information.htm Jones, D. A. (2002). The polarizing effect of new media messages. International Journal of
35
Karp, J. A., Nai, A., & Norris, P. (2018). Dial 'F' for fraud: Explaining citizens suspicions about elections. Electoral Studies, 53, 11-19. http://dx.doi.org/10.1016/j.electstud.2018.01.010 Keune, M. (2016) Inleiding: De legitimiteit van het poldermodel in de eenentwintigste eeuw.
In: M. Keune, (Ed.): Nog steeds een mirakel? De legitimiteit van het Poldermodel in de
eenentwintigste eeuw, Amsterdam: Amsterdam University Press.
Kimball, D. C., & Gross, C. A. (2004, Oct). The growing polarization of American voters. Paper presented at The State of the Parties: 2004 and Beyond conference, Akron, OH. Retrieved from
http://www.uakron.edu/bliss/docs/state-of-the-parties-documents/Kimball.pdf
Körösényi, A. (2013). Political polarization and its consequences on democratic accountability.
Corvinus Journal of Sociology and Social Policy, 4(2), 3-30. Retrieved from
http://unipub.lib.uni-corvinus.hu/2245/1/84-334-1-PB.pdf
Lau, R. R., & Redlawsk, D. P. (2001). Advantages and disadvantages of cognitive heuristics in political decision making. American Journal of Political Science, 45(4), 951-971. Retrieved from http://www.jstor.org/stable/2669334
Levendusky, M. S. (2013). Why do partisan media polarize viewers? American Journal of
Political Science, 57(3), 611-632. http://dx.doi.org/10.1111/ajps.12008
Liu, F. C. S. (2010). Polarized news media and the polarization of the electorate. International
Journal of Artificial Life Research, 1(1), 35-50.
http://dx.doi.org/10.4018/jalr.2010102103
Luskin, R. C. (1990). Explaining political sophistication. Political Behavior. 12(4), 331-361. http://dx.doi.org/10.1007/BF00992793
36
McCright, A. M., Dunlap, R. E., & Marquart-Pyatt, S. T. (2016). Political ideology and views about climate change in the European Union. Environmental Politics, 25(2), 338-358. http://dx.doi.org/10.1080/09644016.2015.1090371
McManus, J. (1995). A market-based model of news production. Communication Theory, 5(4), 301-338. http://dx.doi.org/10.1111/j.1468-2885.1995.tb00113.x
Macdonald, S. E., & Rabinowitz, G. (1998). Solving the paradox of nonconvergence: Valence, position, and direction in democratic politics. Electoral Studies, 17(3), 281-300.
http://dx.doi.org/10.1016/S0261-3794(98)00032-8
Mouw, T., & Sobel, M. E. (2001). Culture wars and opinion polarization: The case of abortion.
American Journal of Sociology, 4, 913-943. http://dx.doi.org/10.1086/320294
Munson, S. A., & Resnick, P. (2010). Presenting diverse political opinions: How and how much?
Proceedings of the 2010 CHI conference on Expressing and Understanding Opinions in
Social Media. April 10-15. Atlanta, GA. http://dx.doi.org/10.1145/1760000/1753543
Nai, A. (n.d.). "If Trump says so... then no." Resistance to persuasion and source cues under
conditions of source unlikeability. Manuscript under review.
Nederhof, A. J. (1985). Methods of coping with social desirability bias: A review. European
Journal of Social Psychology, 15(3), 263-280. http://dx.doi.org/10.1002/ejsp.2420150303
Overby, L. M., & Barth, J. (2006). Radio advertising in American political campaigns. The persistance, importance, and effects of narrowcasting. American Politics Research, 34(4), 451-478. http://dx.doi.org/10.1177/1532673X06288201
Petty, R. E., & Cacioppo, J. T. (1979). Issue involvement can increase or decrease persuasion by enhancing message-relevant cognitive responses. Journal of Personality and Social
37
Pfeifer, M. (2009). Public opinion on state responsibility for minimum income protection: A comparison of 14 European countries. Acta Sociologica, 52(2), 117-134.
http://dx.doi.org/10.1177/0001699309104000
Pildes, R. H. (2014). Romanticizing democracy, political fragmentation, and the decline of American government, Yale Law Journal, 124(3), 804-852. Retrieved from
http://heinonline.org/hol-cgi-bin/get_pdf.cgi?handle=hein.journals/ylr124§ion=23 Prior, M. (2007). Post-broadcast democracy. New York: Cambridge University Press.
Rae, N. C. (2007). Be careful what you wish for: The rise of responsible parties in American national politics. Annual Review Political Science, 10, 169-191.
http://dx.doi.org/10.1146/annurev.polisci.10.071105.100750
Resnick, P., Garrett, R. K., Kriplean, T., Munson, S. A., & Stroud, N. J. (2013). Bursting your (filter) bubble: Strategies for promoting diverse exposure. Proceedings of the 2013
conference on Computer Companion, 987-1-4503-1332-2l,
http://dx.doi.org/10.1145/2441955.2441981
Shuper, P. A., & Sorrentino, R. M. (2004). Minority versus majority influence and uncertainty orientation: Processing persuasive messages on the basis of situational expectancies. The
Journal of Psychology, 144(2), 127-147. http://dx.doi.org/10.3200/SOCP.144.2.127-148
Stokes, D. E. (1963). Spatial models of party competition. American Political Science Review,
57(2), 368-377. http://dx.doi.org/10.2307/1952828
Sullivan, A. (2017, September 19). America wasn't built for humans. New York Magazine. Retrieved from http://nymag.com/daily/intelligencer/2017/09/can-democracy-survive-tribalism.html
38
Tormala, Z. L., & Petty, R. E. (2002). What doesn't kill me makes me stronger: The effects of resisting persuasion on attitude change. Journal of Personality and Social Psychology,
83(6), 1298-1313. http://dx.doi.org/10.1037//0022-3514.83.6.1298
van der Veer, P. (2006). Pim Fortuyn, Theo van Gogh, and the politics of tolerance in the Netherlands. Public Culture, 18(1), 111-124. http://dx.doi.org/10.1215/08992363-18-1-111
Vergeer, M. (2014). Peers and sources as social capital in the production of news: Online social networks as communities of journalists. Social Science Computer
Review, 33(3), 277- 297. http://dx.doi.org/10.1177/0894439314539128
Wheeler, S. C., Briñol, P., & Hermann, A. D. (2007). Resistance to persuasion as self-regulation: Ego-depletion and its effects on attitude change processes. Journal of Experimental
Social Psychology, 43(1), 150-156. http://dx.doi.org/10.1016/j.jesp.2006.01.001
Zick, A., Küpper, B., & Hövermann, A. (2011). Intolerance, prejudice and discrimination: A European report. Friedrich Ebert Stiftung. Retrieved from
https://library.fes.de/pdf-files/do/07908-20110311.pdf
Zuwerink, J. R., & Devine, P. G. (1996). Attitude importance and resistance to persuasion: It's not just the thought that counts. Journal of Personality and Social Psychology, 70(5), 931-944. http://dx.doi.org/10.1037/0022-3514.70.5.931
39
Appendices
Appendix A: articles for the conditions and buffer article
40
41
42
43
