Effects of Personal Predispositions on Selective Exposure Bibian E. I. V. van der Vis
University of Amsterdam
Faculty of Social and Behavioural Sciences
bibian.vandervis@student.uva.nl Student number: 10195599 Master’s Thesis
Graduate School of Communication
Master’s programme Communication Science: Political Communication Supervisor: Yphtach Lelkes
Abstract
The current information environment allows us to structure our own information flow, which facilitates the likelihood of selective exposure to likeminded content. This type of selectivity threatens democracy by encouraging polarization. Theories of cognitive dissonance and cognitive reflection have pointed at numerous personal
predispositions influencing selective exposure habits of individuals, but research has yet to include predispositions raised by both theories into one model. This study describes the individual and partial effect of a number of these personal
predispositions from both domains, namely attitude strength, defensive confidence, issue public membership and reasoning style, on ideological selectivity towards information from two political dimensions. Using a unique research design, created to overcome limitations of previous studies and reflecting a real-world approach most closely, an online selection experiment was conducted among Dutch citizens (N = 114). The results indicate a positive effect of attitude strength and a negative effect of issue public membership, moderated by the level of defensive confidence for one dimension, but not for another. Findings imply that our personal predispositions shape our selective exposure habits to political information from the socio-economic
dimension in order to reduce the occurrence of cognitive dissonance.
Keywords: selective exposure, personal predispositions, political dimensions, cognitive dissonance, cognitive reflection
Effects of Personal Predispositions on Selective Exposure
The modern digital era encourages exposure to one-sided information (Bennett & Iyengar, 2008; Cappella, Kim, & Albarracín, 2015; Iyengar & Hahn, 2009;
Iyengar, Hahn, Krosnick, & Walker, 2008; Taber & Lodge, 2006). This ideological selectivity has numerous consequences for democracies, such as polarization
(Arceneaux & Johnson, 2010; Arceneaux, Johnson, & Cryderman, 2013; Levendusky, 2013; Stroud, 2010; Taber & Lodge, 2006), political misconceptions (Bawden & Robinson, 2009; Caprara & Zimbardo, 2004; Prior, 2005) and limiting political discussion (Torcal & Maldonado, 2014). Ideological selectivity has been studied in a number of ways. Festinger (1957) explained that individuals desire likeminded content because counter-attitudinal information generates cognitive dissonance. Kahan (2013) rather proposed that attending counter-attitudinal information takes cognitive effort to conceptualize counterarguments, while individuals generally rely on heuristics to draw fast and associative conclusions. This inability to engage in systematic reasoning makes them aversive to counterarguments (Kahneman, 2003).
To date, it is unclear what role personal predispositions, such as existing attitudes and reasoning style, play in ideological selectivity. The understanding of personal predispositions is important because they are found to influence cognitive dissonance (Festinger, 1957) as well as cognitive reflection (Kahan, 2013). Research has yet to study a model that includes personal predispositions found to account for the occurrence of both cognitive dissonance and cognitive reflection. Such a model extends our understanding of the effects of personal predispositions on selective exposure. The central question in this study is: to what extent is ideological selectivity to political information influenced by personal predispositions?
First, it is tested how personal predispositions have an individual and a partial effect on selective exposure to information from two political dimensions. Second, the findings are compared to the results of previous studies, which have often used
methods generating limited generalizable results. This study attempts to overcome this by using a unique, real-world reflective research design. Lastly, the insights are applied to the understanding of the current political environment.
Theoretical foundations
Selective exposure encompasses the idea that individuals purposely select certain information to achieve certain psychological goals (Stroud, 2010; Zillmann & Bryant, 1985). Findings indicate that people have a tendency to turn to information that is politically likeminded, known as ideological selectivity (Brannon, Tagler, & Eagly, 2007; Graf & Aday, 2008; Hart et al., 2009; Torcal & Maldonado, 2014). The theory of cognitive dissonance provides one explanation for this phenomenon: people experience negative affect states when they are confronted with information that does not resonate with their beliefs (Festinger, 1957; Kunda, 1990). Therefore, people are motivated to avoid such information that counter argues their existing attitudes and opinions and they have a desire to select information from which they can reach preferred conclusions (Festinger, 1957; Kunda, 1990). In doing so, people create positive affective states (Westen, Blagov, Harenski, Kilts, & Hamann, 2006). Individuals rely on the strategy of motivated reasoning to mitigate cognitive dissonance and seek out information that confirms what they already believe
(Festinger, 1957; Kunda, 1990), resulting in the cognitive confirmation bias (Brannon et al., 2007; Hart et al., 2009; Iyengar et al., 2008; Knobloch-Westerwick & Meng, 2011; Taber & Lodge, 2006).
Other studies, however, ascribe cognitive reflection to the occurrence of ideological selectivity (Frederick, 2005; Kahan, 2013; Kahneman, 2003). Counter arguing opposing information requires a certain amount of cognitive effort (Meffert, Chung, Joiner, Waks, & Garst, 2006). In order for individuals to formulate such counterarguments, they must be able to engage in conscious reasoning (Kahan, 2013; Kahneman, 2003). Yet, a substantial segment of the public lacks this resource (Kahan, 2013; Meffert et al., 2006) and engages in a more heuristic-driven style of reasoning, aversive to counterarguments (Kahan, 2013; Kahneman, 2003). According to the theory of cognitive reflection, not cognitive dissonance but the inability to process information consciously accounts for individuals’ tendency to select likeminded content over counter-attitudinal content.
Research has yet to include personal predispositions from both explanations into one model to see the partial effect on selective exposure habits. Three
predispositions related to cognitive dissonance and one predisposition related to cognitive reflection are discussed and added to a model predicting ideological selectivity.
The notion that attitudes influence selective exposure forms the core argument of Festinger’s (1957) theory of cognitive dissonance, because people experience dissonance when they are confronted with information that does not resonate with their existing beliefs. Many other studies found additional evidence for the attitude strength influencing ideological selectivity through cognitive dissonance reduction (Hart et al., 2009; Iyengar et al., 2008; Knobloch-Westerwick & Meng, 2011; Kunda, 1990; Levendusky, 2013; Taber & Lodge, 2006). Brannon et al. (2007) extended this line of research and found that the preference for consistent information intensifies as the attitude is more strongly held. Knobloch-Westerwick and Meng (2011) added that
selection of consistent political information strengthens individuals’ political self-concept, which in itself again increases the preference for congenial information.
These theories stress that existing political attitudes influence selective
exposure habits of individuals (Festinger, 1957; Hart et al., 2009; Iyengar et al., 2008; Kunda, 1990; Taber & Lodge, 2006), where the preference for likeminded content increases as the attitude strength increases (Brannon et al., 2007). In avoiding cognitive dissonance, people engage in motivated reasoning and select information with which they agree and which confirms their existing attitudes (Festinger, 1957; Kunda, 1990; Taber & Lodge, 2006). The occurrence of cognitive dissonance when confronted with counter-attitudinal information is more pronounced as the existing attitude is more strongly held (Brannon et al., 2007; Festinger, 1957). Therefore, attitude strength is taken into account as the first personal predisposition to influence the selection of likeminded political information.
H1: The stronger the attitude towards a political issue, the more likely people are to select likeminded content about that political issue.
A second predisposition said to influence cognitive dissonance was first noted by Albarracín and Mitchell (2004). They found that in some conditions, people with a strong belief in their own opinions displayed a decrease in preference for attitudinally consistent information. They concluded that individuals who believe they can
successfully defend their attitudes from external attacks have a greater reception of counter-attitudinal content (Albarracín & Mitchell, 2004). In a later study, Graf and Aday (2008) showed how people’s defensive confidence could be build up through reinforcement of their existing views. Participants free to self-select information from a platform using both attitudinally supporting and opposing information, exposed themselves to counter-attitudinal information after exposure to attitudinally consistent
information (Graf & Aday, 2008). They used the likeminded content to create a frame with which they could contextualize the counter-attitudinal content (Graf & Aday, 2008). Additional studies underline these results and state that the confirmation bias weakens among individuals whose attitudes were reinforced prior to exposure to counter-attitudinal content (Fischer & Greitemeyer, 2010; Hart et al., 2009).
These studies indicate that cognitive dissonance can be diminished if the confidence in ones own attitudes and opinions is strengthened. When people feel that they can successfully defend their ideas from external attack, the threshold for
cognitive dissonance is raised, therefore less likely to occur and selection of a counter-attitudinal article increases (Albarracín & Mitchell, 2004; Festinger, 1957; Fischer & Greitemeyer, 2010; Graf & Aday, 2008; Hart et al., 2009; Kunda, 1990). Because a high level of defensive confidence decreases the likelihood of the
occurrence of cognitive dissonance, it serves as the second predisposition influencing individuals’ selective exposure habits.
H2: People with a high level of defensive confidence are more likely to select counter-attitudinal political information than people with a low level of defensive confidence.
These expectations about the influence of attitude strength and defensive confidence on ideological selectivity do not differentiate between various political issues. However, does cognitive dissonance equally occur in the selection of information about any political issue? Converse (2006) stated that the common citizenry does not hold a strong attitude towards a wide variety of issues, because it does not develop a global point of view about politics. Rather, individuals develop such attitudes towards a few political issues (Converse, 2006; Kim, 2009). Citizens are incapable of devoting their attention to understanding various political issues
(Krosnick, 1990). Instead, citizens are knowledgeable about and have an intrinsic interest in some political issues (Atkin, 1971; Converse, 2006; Krosnick, 1990). Fragmented groups of citizens are formed in society, highly concerned about specific political issues, aiming to become specialists about their preferred political issue (Converse, 2006; Kim, 2009). Members of the issue public consider their preferred political issue as personally important (Converse, 2006; Kim, 2009; Krosnick, 1990) and personally relevant (Bolsen & Leeper, 2013; Chen, 2013; Popkin, 1991). When individuals have such personal ties with political issues, their selection criteria are driven more by the interest in the political issue than by partisan considerations (Atkin, 1971; Iyengar et al., 2008; Kim, 2009; Knobloch-Westerwick & Kleinman, 2012; Prior, 2013). Knobloch-Westerwick and Kleinman (2012) found that voters select information about unfavorable candidates when they are likely to win, because they want to be informed about upcoming political decisions and circumstances. This shows how information utility can override the confirmation bias (Garrett, 2013; Knobloch-Westerwick & Kleinman, 2012) and decreases the occurrence of cognitive dissonance. Members of an issue public strive to gain a well-rounded sense of the issue that concerns them and therefore consume information less biased than non-members (Chen, 2013; Kim, 2007). The stable attitudes of issue public non-members allow them to expose themselves to opposing views because they believe they will not be affected by counter-attitudinal messages (Chen, 2013; Converse, 2006; Knobloch-Westerwick & Meng, 2009; Krosnick, 1990; Krosnick & Telhami, 1995). The goal for issue publics is not to avoid cognitive dissonance, but to become specialists about their preferred issue (Iyengar et al., 2008; Kim, 2009; Knobloch-Westerwick & Kleinman, 2012).
These studies show how ideological considerations for information selection and the occurrence of cognitive dissonance can be surpassed. Information utility overrides the occurrence of cognitive dissonance and motivates an unbiased
information search (Garrett, 2013; Knobloch-Westerwick & Kleinman, 2012). Thus, issue public membership is considered a third personal predictor of selective exposure habits of individuals to political information.
H3: Individuals who are members of the issue public are more likely to select counter-attitudinal political information about their preferred issue than non-members of the issue public.
While these predispositions are related to avoiding cognitive dissonance, cognitive reflection has also been valued as an explanation for people’s affinity for likeminded content and it holds a personal predisposition responsible (Kahan, 2013; Kahneman, 2003). Individuals can process information according to a low-effort mode, based on heuristic cues where individuals draw fast and associative
conclusions or a high-effort mode, where individuals reach rule-based conclusions through systematic reasoning (Chaiken & Trope, 1999). Research mainly attributes a heuristic-driven reasoning style to individuals aversive to counterarguments (Kahan, 2013; Kahneman, 2003) and an effortful reasoning style to individuals open to counterarguments (Kahan, 2013). However, research indicates that people tend to select likeminded content because most of them lack the capability of effortful reasoning (Kahneman, 2003). So, rather than being unwilling to expose oneself to counter-attitudinal content, cognitive reflection theory states that most people are unable to process information consciously, because it requires cognitive effort (Kahan, 2013; Meffert et al., 2006), therefore becoming aversive to counter-attitudinal content.
One might expect that individuals able to engage in an effortful reasoning style are more likely to select counter-attitudinal information. Yet, a straightforward interpretation of these psychological modes is complex. Kahan (2013), for instance, concluded that an effortful reasoning style could also motivate people to select attitudinally consistent information in certain situations. In investigating information selection of US citizens about a political party, he found that people who can engage in conscious reasoning, but feel a certain loyalty towards an affinity group select likeminded content because it emphasizes their cherished position in the group. Yet, Dutch political parties hold a less strong ownership on certain political issues than do US political parties due to its multi-party system, where multiple parties overlap in issue-ownership (Hopmann, Elmelund-Praestekaer, Albæk, Vliegenthart, & De Vreese, 2012). Political issues are much less tied to one specific party, therefore less tied to a specific affinity group, possibly diminishing Kahan’s (2013) results for European countries with a multi-party system.
In general, the theory of cognitive reflection explains how a conscious and effortful reasoning style enables people to be more receptive of counterarguments. This increases the likelihood of selection of counter-attitudinal content, because an effortful reasoning style gives people the ability to formulate counterarguments (Kahan, 2013; Kahneman, 2003). If individuals are not able to engage in an effortful reasoning style, their heuristic-driven form of reasoning makes them aversive to counterarguments because they cannot defend themselves from external attack (Kahan, 2013; Kahneman, 2003). Following the theory of cognitive reflection, the reasoning style is added to the model predicting ideological selectivity.
H4: People who engage in a more conscious and effortful style of reasoning are more likely to select counter-attitudinal information about a political issue than people who engage in a more heuristic-driven style of reasoning.
These four personal predispositions are added to a model in order to investigate the partial effect on the selection of likeminded content. Such a model generates a more accurate prediction of selective exposure habits of individuals (Van Peet, Namesnik, & Hox, 2012).
Finally, the model possibly contains a moderation effect. Members of an issue public are driven by an intrinsic interest in the political issue (Atkin, 1971; Popkin, 1991) and strive to become specialists about the issue (Converse, 2006; Krosnick, 1990). Because information utility and issue importance have been found to override ideological selectivity (Knobloch-Westerwick & Kleinman, 2012), it is implied that for issue public members the consideration of being able to defend ones own attitudes might be inferior, since the goal is not to seek out information that resonates with their existing ideas, but to become fully informed about the issue (Converse, 2006; Kim, 2009; Krosnick, 1990). Therefore, cognitive dissonance is less likely to occur among issue publics confronted with counter-attitudinal information (Iyengar et al., 2008; Kim, 2009; Knobloch-Westerwick & Kleinman, 2012). Non-members of an issue public lack this intrinsic interest in the issue and cognitive dissonance is therefore more likely to occur when they are confronted with counter-attitudinal content. Non-members must therefore rely more on their level of defensive confidence in order to select counter-attitudinal information than issue public members. Thus, for issue public members, the level of defensive confidence is less important than for non-members of the issue public in selective exposure habits, because the goal of information selection for issue publics is not to avoid cognitive
dissonance but to become specialists about the preferred political issue (Atkin, 1971; Converse, 2006; Iyengar et al., 2008).
H5: The level of defensive confidence is less influential in the selective exposure habits of members of an issue public than in the selective exposure habits of non-members of an issue public.
Method Design
To test the hypotheses, an online survey experiment was developed with a 2 (political issue) x 2 (information bias) between-subjects design (table 1). Political issue was the first experimental factor with two levels, namely immigration and reducing income differences. These two political issues were chosen because they represent the socio-cultural and socio-economic dimensions on which European voters assess and evaluate politics (Kriesi et al., 2006; Van Der Brug & Van Spanje, 2009). Individuals’ attitudes on cultural issues are clearly differently structured than their attitudes on socio-economic issues (Van Der Brug & Van Spanje, 2009). This design allowed for further investigation of the differences in people’s assessment of the two dimensions and could complement earlier findings on the level of selective exposure.
Information bias was the second experimental factor with two levels, namely supportive and opposing arguments. Participants were confronted with one article about each political issue, either clearly supportive or opposing of their own attitudes towards that issue. This research design overcomes a limitation found in previous studies, where participants were offered various perspectives of the same topic to choose from (Brannon et al., 2007; Graf & Aday, 2008; Meffert et al., 2006).
sides of the same story in two separate articles. Rather, they cover the political issue in one article, possibly presenting different sides within one article. Therefore, this study approached a real-world setting as much as possible, found to reflect real-world habits most closely (Feldman, Stroud, Bimber, & Wojcieszak, 2013), and added a unique design to the existing body of research.
Table 1
2x2 Between-Subjects Research Design
Political issue reducing income differences
Political issue immigration
Pro reducing income differences
Anti reducing income differences
Pro immigration 1 3
Anti immigration 2 4
Note. Numbers 1 through 4 in the cells indicate the different conditions of the design. In addition, each condition included the same options to avoid political news, with an article about entertainment and sports. Research found that giving participants the option to tune out diminishes the polarizing effect of political information
(Arceneaux & Johnson, 2010; Arceneaux et al., 2013). While the option for non-political news like entertainment discourages selective exposure, especially to political news (Feldman et al., 2013), research indicates that the measurement of selective exposure is much more reflective of real-world habits when real-world options are included in research, such as entertainment and sports options (Arceneaux & Johnson, 2007; Feldman et al., 2013). Participants were free to select at least one of the four articles from the front-page of an online newspaper, containing information about immigration, reducing income differences, sports and entertainment.
Sample
Participants were reached by means of a convenience sample. The link to the online experiment was distributed via the social media website Facebook and
addressed only Dutch citizens. In addition, participants were asked to forward the link to their own Dutch social network, resulting in a snowball method. The link was online from April 22nd to May 9th, 2016. In total, 151 respondents participated in this study. However, 28 respondents did not complete the questionnaire and were
therefore excluded from the study. The final sample consisted of 114 respondents, among whom 55% were males and 45% were females, with an average age of 40.62 years (SD = 16.92). Regarding education, 13% had a high school degree, 4% an intermediate vocational education, 21% a higher vocational education and 61% had a university degree. While this sample was much better educated than the general Dutch population (CBS, 2016), research has shown that non-representative samples yield results that are similar to representative samples (Mullinix, Leeper, Druckman, & Freese, 2015).
Stimulus material
Based on current news and literature on immigration and economic inequality, four articles were constructed as stimulus material. For each political issue, two articles were developed representing either supportive or opposing information on the issue. The manipulation is mainly found in the headline and first sentence of the article seen on the front-page of the online newspaper. The content of the article is of lesser importance, because this study is only interested in the selection choice of the participants. This choice was made on the front-page, where participants chose the article they wished to read based on the headline and the first sentence of the article. To ensure that any differences detected were due to the bias of the article’s headline, the two articles about each political issue were comparable on the length of the headline and the first sentence to avoid a mere exposure effect. In addition, identical photographs were used for both stories per political issue. For the headline, only one
word differed, namely the one indicating the position of the article. The first sentence of the articles provided a key argument underlining the statement made in the
headline. The stimulus material is found in Appendix A. Table 2
Results Pilot Test for Manipulation Information Bias
Article M(SD) t
Pro-immigration 4.45(.52) 9.24***
Anti-immigration 1.82(.60) -6.50***
Pro-reducing income differences 3.45(.52) 2.89
Anti-reducing income differences 2.27(.65) -3.73**
Note. N = 11. Analyses indicate deviation from value 3 (neutral). **p<.01, ***p<.001
In April 2016, a pilot test for the headlines of the articles was carried out on another sample. This pilot test was conducted to determine that participants received the stimulus material as intended (table 2). The sample consisted of 11 participants, among whom 5 were males and 6 were females, with an average age of 30.21 years (SD = 11.81). Participants were asked to rate the headlines, including the first
sentence of the article and the photograph, on the bias towards the political issue. A 5-point Likert scale was used ranging from (1) very negative to (5) very positive. A t-test analyzed whether their scores were significantly deviant from (3) neutral. For the article supportive of reducing income differences, changes had to be made due to its indistinctness. Another t-test indicated that the new article was now significantly deviant from neutral (M = 4.36, SD = .51), t(10) = 5.16, p < .000. The other three headlines were assessed as expected.
Procedure
The online survey experiment was created with Qualtrics, a research program developed for online data collection and analysis. Participants were first informed that this experiment was interested in news selection and that its goal was to obtain insight
in people’s selection habits in different news environments. They were also informed about the expected time to participate in the study and information was provided on the guaranteed anonymity and right to end participation at any given time. If
participants agreed to the terms and conditions, they were redirected to the study. First, participants were asked a number of questions designed to gauge their political attitudes, defensive confidence, issue importance and issue relevance, and reasoning style. After these questions, the participants were shown an introduction page about the selection procedure for the front-page of the newspaper. They were asked to select one headline from the front-page which would redirect them to the full article. If they were interested in another article, they could return to the front-page to select a second, third or fourth headline. After this introduction page, participants were randomly assigned to either one of four conditions. They were then free to self-select as many articles as they wanted. After the participants were confronted with the front-page and read the article(s), a manipulation check was carried out to determine whether they paid enough attention to the stimulus material.
Finally, the participants were asked demographic questions about their gender, age and highest completed education. Participants were then debriefed about the study and informed that the articles were created solely for the purpose of this study and included partly fictitious information. Completing the questionnaire took on average 14.40 minutes (SD = 7.15). The full questionnaire can be found in Appendix B. Measures
Political attitudes. The attitudes towards several political issues were measured using a 7-point Likert scale, ranging from -3 (completely disagree) to 3 (completely agree). Attitudes towards immigration and reducing income differences were measured, in addition to attitudes towards three other political issues, in order to
avoid a priming effect when participants were confronted with the stimulus material. The political attitudes for each issue were measured using a three-item scale. The items were taken from previous research.
Immigration. The attitudes towards immigration were measured with the following
items: (1) immigrants make the country a worse place to live (Van Der Brug & Van Spanje, 2009), (2) immigrants should not be permitted to come into our country if they compete with our own workers (Sampson & Smith, 1957) and (3) our country should permit the immigration of foreign people even if it lowers our standard of living (Sampson & Smith, 1957). Statements 1 and 2 were recoded, so the higher the score participants indicated per item, the more supportive they were of immigration. An explanatory factor analysis indicated that the scale was one-dimensional,
explaining 70.75% of the variance. The 3-item scale also proved to be reliable with a Cronbach’s alpha of .79 (M = .38, SD = 1.35).
Reducing income differences. The attitudes towards reducing income differences
were measured with the following statements: (1) the government should take
measures to reduce differences in income levels (Van Der Brug & Van Spanje, 2009), (2) we must contribute to reduce income differences (Kraaykamp, Van Snippenburg, & Ultee, 2015), and (3) greater income equality should be promoted within society (Kraaykamp et al., 2015). The higher the score participants indicated, the more supportive they were of reducing income differences. An explanatory factor analysis indicated that the scale was one-dimensional, explaining 79.31% of the variance. The 3-item scale was also reliable with a Cronbach’s alpha of .87 (M = -.40, SD = 1.61).
Additional political issues. The statements on the environment, education and
the analyses of this study. They were only included to avoid a priming effect. They were therefore taken from the programs of different Dutch political parties.
In order to construct the strength of participants’ attitudes supportive or opposing of immigration and reducing income differences, the scales were recoded. Participants who indicated a score of 0 or lower on the general attitude scale were considered to be opposing of the political issue. Participants who indicated a score of 0 or higher were considered to be supportive of the political issue. Participants with a neutral attitude towards the political issue were considered to have a weak attitude and were coded as both weak opposing and weak supportive. Participants who scored on either ends of the general attitude scale (-3 or +3) were considered to have a strong attitude. Two additional variables were constructed for the attitude strength towards each political issue, ranging from (1) weak to (4) strong attitude, namely strength pro-attitude and strength anti-pro-attitude.
Defensive confidence. To measure participants’ defensive confidence levels, the Defensive Confidence Scale provided by Albarracín and Mitchell (2004) was used. The scale was composed of 12 items, to which respondents could answer on a 5-point scale, ranging from (1) not at all characteristic of me to (5) extremely
characteristic of me. Examples of statements are ‘I can defend my points of view when I want to’ or ‘I am unable to defend my own opinions successfully’. Four items needed to be recoded for this scale, namely statements 3, 4, 10 and 12. An
explanatory factor analysis indicated that the scale measured two dimensions, with component 1 (eigenvalue = 5.37) explaining 44.73% of the variance and component 2 (eigenvalue = 1.38) explaining 11.53% of the variance. Because the second
prove to be reliable (Cronbach’s α = .86), the scale as used by Albarracín and Mitchell (2004) was constructed (M = 3.67, SD = .55).
In addition to these statements measuring participants’ general level of defensive confidence, participants were also asked about issue-specific defensive confidence on the five political issues. However, this study is only interested in the defensive confidence levels about immigration and reducing income differences. This issue-specific defensive confidence was measured using three items from the scale provided by Albarracín and Mitchell (2004) for each political issue.
Immigration defensive confidence. An explanatory factor analysis indicated that the
scale was one-dimensional, explaining 66.47% of the variance. The 3-item scale also proved to be reliable with a Cronbach’s alpha of .75 (M = 3.74, SD = .66).
Reducing income differences defensive confidence. An explanatory factor analysis
indicated that this scale was also one-dimensional, explaining 72.06% of the variance. The 3-item scale was reliable with a Cronbach’s alpha of .80 (M = 3.61, SD = .72).
Issue publics. Research on issue publics usually focuses on issue importance (Kim, 2009; Krosnick, 1990), because multiple attributes related to issue public membership have been shown to significantly correlate (Wojcieszak, 2012).
However, issue publics should be considered as a construct made up of more than one item (Chen, 2013). To measure issue public membership in a more reliable way, issue relevance needs to be included as an additional attribute (Bolsen & Leeper, 2013; Chen, 2013). Therefore, two items were used to measure issue public membership for each political issue. Participants were asked to rate their opinions on two statements on a 7-point Likert scale, ranging from (1) very unimportant/irrelevant to me to (7) very important/relevant to me: (1) to what extent is immigration/reducing income differences personally important to you, and (2) to what extent is
immigration/reducing income differences personally relevant to you. Membership of an issue public was constructed for both immigration and reducing income
differences. None of the items needed to be recoded. An explanatory factor analysis indicated that the scale for the immigration issue public was one-dimensional, explaining 71.77% of the variance. The 2-item scale also proved to be reliable with a Cronbach’s alpha of .61 (M = 4.28, SD = 1.20). For the reducing income differences issue public, an explanatory factor analysis indicated that the scale was
one-dimensional, explaining 82.95% of the variance. The 2-item scale was also reliable with a Cronbach’s alpha of .79 (M = 3.64, SD = 1.54).
Reasoning style. Frederick (2005) developed The Cognitive Reflection Test (CRT), which measures individuals’ disposition to engage in a conscious and effortful form of reasoning as opposed to a heuristic-driven form of reasoning (Appendix C). It is a performance-based measure of individuals’ disposition to approach intellectual problems in an open-minded manner (Campitelli & Labollita, 2010; Frederick, 2005). The CRT provided three brainteasers with open-ended questions, where it is not unusual that a large number of the participants does not answer any question correctly (Weller et al., 2013). Participants were given one point for each question answered correctly. A final score for the CRT ranged from (0) heuristic-driven style of reasoning to (3) conscious and effortful style of reasoning. The mean score for the subjects in this study was 1.62 (SD = 1.08).
Selective political news exposure. Qualtrics unobtrusively kept track of the different articles participants selected from the front-page and the order in which participants selected the articles. Two measures of exposure were used. Participants were coded on the selected political article for both immigration and reducing income differences, where they either (1) selected or (0) not selected the articles. Also, for
each political article the participants selected, they were coded on the content of the political article, where they were exposed to either (1) likeminded or (0) counter-attitudinal content. Whether the article was likeminded or counter-counter-attitudinal was derived from their attitudes towards the issues.
An overview of the descriptive statistics of the different variables and their correlations can be found in Appendix D.
Analyses
To test the individual and partial effect of attitude strength (H1), defensive confidence (H2), issue public membership (H3) and reasoning style (H4) on the selection of likeminded content, logistic regression analyses were conducted. For each regression, the unstandardized coefficients with standard errors were reported in a table, as well as the χ2-value, degrees of freedom, odds ratio, p-value and R. Additionally, fixed effects models were assessed which took into account repeated measures and made optimal use of the data. For these analyses, the unstandardized coefficients with standard errors, as well as the t-value, p-value and Akaike corrected were described. A priori power analyses indicated that for the logistic regression analyses and the fixed effects models a minimum of 138 respondents was needed (f2 = .15, α = .05, 1-β = .95). Finally, H5 was subject to a moderation analysis with the centered values for issue public membership and defensive confidence. The unstandardized coefficients, standard errors, t-value, p-value and 95%-confidence intervals were reported. An a priori power analysis indicated that a minimum of 89 respondents was needed (f2 = .15, α = .05, 1-β = .95).
Results
To check if random assignment to the conditions was successful, three tests were performed. First, an ANOVA was conducted for the distribution of age. There
was no main effect for the content of the immigration article, F(1,3) = 1.79, p = .184, η2 = .02, nor for the content of the article about reducing income differences, F(1,3) = .02, p = .903, η2 = .00, nor an interaction between both, F(1,3) = .00, p = .977, η2 = .00. In addition, there were no differences in the distribution of gender for the content of the immigration article, χ2(1, N = 114) = .32, p = .572, nor for the content of the article about reducing income differences, χ2(1, N = 114) = 5.20, p = .173, nor were there differences in the distribution of education for the content of the immigration article, χ2(5, N = 114) = 8.78, p = .118, nor for the content of the article about reducing income differences, χ2(5, N = 114) = 6.34, p = .275. This indicated that the randomisation was successful for every variable, and they did not need to be
controlled for.
To determine whether participants paid enough attention to the stimulus material, a manipulation check was conducted after exposure to the front-page of the online newspaper. Participants were asked about the tone of the headlines on
immigration and reducing income differences. They could answer on a 5-point Likert scale ranging from (1) very negative to (5) very positive. Participants also had an ‘I don’t know’ option. The participants confronted with a supportive article of
immigration indicated a significantly more positive tone of the article (M = 4.20, SD = .87) than the participants confronted with an opposing article of immigration (M = 1.88, SD = .55), t(83) = 15.07, p < .001. In addition, participants confronted with a supportive article of reducing income differences indicated a significantly more positive tone of the article (M = 3.98, SD = .87) than participants confronted with an opposing article of reducing income differences (M = 2.31, SD = .82), t(92) = 9.35, p < .001. These results are in accordance with the pilot test and indicate that the
In general, participants selected 1.64 articles (SD = .73) from the front-page of the online newspaper. Participants were more likely to select a political article (M = .91, SD = .29) than a sports article (M = .20, SD = .40), t(117) = 14.74, p < .001, or an entertainment article (M = .17, SD = .38), t(117) = 14.95, p < .001. When selecting an article about immigration, the average difference (M = .07, SD = 1.00) between participants selecting a likeminded (M = .53, SD = .50) or a counter-attitudinal article (M = .47, SD = .50) was not significant, t(76) = .57, p = .572. In addition, when selecting an article about reducing income differences, the average difference (M = -.10, SD = 1.00) between participants selecting a likeminded (M = .45, SD = .50) or a counter-attitudinal article (M = .55, SD = .50) was also not significant, t(72)= -.82, p = .416.
Logistic regressions were conducted to predict the effects of the personal predispositions on the selection of likeminded content. The binary logistic regressions for the individual effect of each predisposition on selective exposure can be found in Appendix E (tables E1 and E2). By adding each predictor to the model (Appendix E, tables E3 and E4), the multilevel regression models are constructed and discussed in order to investigate the hypotheses. For an overview of the effects for each of the conditions, see Appendix F.
Table 3 shows that the personal predispositions accounted for 8.4% of the variance in the selection of likeminded content about immigration. The multilevel model, however, did not indicate a good fit, χ2(5) = 5.01, p = .414. The level of general defensive confidence did exert a marginally significant effect, b = 1.11, SE = .64, p = .084. However, the level of general defensive confidence indicated a positive effect on the selection of likeminded content, an effect opposite of what was
confidence were 3.04 times more likely to select likeminded content than participants with a lower level of defensive confidence.
Table 3
Multilevel Logistic Regression Model Predicting the Partial Effect of the Personal Predispositions on the Selection of Likeminded Information about Immigration
b(SE) OR 95% CI for OR Intercept -1.54(1.62) Attitude strength -.19(.36) .83 [.41, 1.68] General defensive confidence 1.11(.64)+ 3.04 [.87, 10.59] Specific defensive confidence -.49(.52) .62 [.22, 1.71] Issue publics -.11(.22) .89 [.58, 1.38] Reasoning style .17(.23) 1.19 [.76, 1.85] χ2(5) 5.01 R .084
Note. n = 77. Only respondents are included who selected the article about immigration. OR = odds ratio, CI = 95 percent confidence interval
+p<.10
The same opposite effect, although insignificant, was found for attitude strength, b = -.19, SE = .36, p = .602, and reasoning style, b = .17, SE = .23, p = .441. These first results for the immigration article implied a rejection of the first
hypotheses: neither a strong attitude (H1), nor a low level of defensive confidence (H2), nor non-membership of the issue public (H3), nor a heuristic-driven reasoning style (H4) significantly increased the selection of likeminded content for the
immigration article.
Table 4 indicated that the personal predispositions accounted for 9.1% of the variance in the selection of likeminded content about reducing income differences. Again, the multilevel model did not provide a good fit, χ2(5) = 5.17, p = .395. The influence of issue public membership indicated a marginally significant effect on the selection of likeminded content, b = .28, SE = .17, p = .097. However, the result was in the opposite direction of what was expected in H3. The result even implied that
members of the issue public were 1.32 times more likely to select likeminded content. These first results for the article about reducing income differences implied a
rejection of the first hypotheses: neither a strong attitude (H1), nor a low level of defensive confidence (H2), nor non-membership of an issue public (H3), nor a heuristic-driven reasoning style (H4) significantly increased the selection of likeminded content for the article about reducing income differences. Table 4
Multilevel Logistic Regression Model Predicting the Partial Effect of the Personal Predispositions on the Selection of Likeminded Information about Reducing Income Differences b(SE) OR 95% CI for OR Intercept .43(1.76) Attitude strength .05(.35) 1.05 [.53, 2.06] General defensive confidence -.32(.57) .72 [.24, 2.22] Specific defensive confidence -.08(.58) .92 [.30, 2.87] Issue publics .28(.17)+ 1.32 [.95, 1.84] Reasoning style -.17(.24) .84 [.53, 1.34] χ2(5) 5.17 R .091
Note. n = 73. Only respondents are included who selected the article about reducing income differences. OR = odds ratio, CI = 95 percent confidence interval.
+p<.10
However, these logistic regressions did not take full advantage of the nature of the data. Respondents in this experiment were given the option to return to the front-page and select multiple articles. Therefore, the variables could be treated as repeated measures. For an optimal use of data, an additional set of analyses was conducted to get a more detailed picture of the influence of the personal characteristics on the selection of likeminded content for both political issues. A logistic mixed regression model, which allows the intercept to vary by respondent, was conducted to analyze the fixed effects of the five factors on the selection of likeminded content (tables 5
and 6). Appendix G shows the fixed effects of the individual factors on the selection of likeminded content.
Table 5
Fixed Effects Models Predicting the Selection of Likeminded Information about Immigration
Model 1 Model 2 Model 3 Model 4 Model 5
b(SE) t b(SE) t b(SE) t b(SE) t b(SE) t
Intercept -.14 (.53) 1.48 (1.36) 1.55 (1.39) 1.32 (1.49) 1.27 (1.51) Attitude strength .15 (.60) .24 .05 (.61) .07 .16 (.63) .25 .20 (.63) .31 .22 (.64) .34 General defensive confidence -.41 (.33) -1.27 -.19 (.47) -.41 -.20 (.47) -.43 -.21 (.48) -.43 Specific defensive confidence -.25 (.41) -.62 -.28 (.41) -.67 -.29 (.42) -.69 Issue publics .07 (.17) .42 .08 (.17) .43 Reasoning style .04 (.19) .22 Akaike corrected 519.12 521.23 521.43 523.28 524.86
Note. n = 119. Only respondents who selected the article about immigration in one of the four trials are included.
For the article about immigration, the fixed effects did not generate a different picture of the influence of the factors on the selection of likeminded content than the multiple logistic regression model (see table 5, model 5). Additional factors did not improve the model-fit and none of the factors provided a significant effect. The coefficients for issue public membership, b = .08, SE = .17, p = .666, and reasoning style, b = .04, SE = .19, p = .825, even indicated an opposite effect of what was expected in the third and fourth hypotheses. These findings provided definitive evidence for the rejection of the first four hypotheses for the selection of likeminded content about immigration: people were not more likely to select likeminded
strongly held (H1), they had a low level of defensive confidence (H2), they did not belong to the issue public (H3), and when they engaged in a more heuristic-driven form of reasoning (H4).
Table 6
Fixed Effects Models Predicting the Selection of Likeminded Information about Reducing Income Differences
Model 1 Model 2 Model 3 Model 4 Model 5
b(SE) t b(SE) t b(SE) t b(SE) t b(SE) t
Intercept -.09 (.39) -.23 (1.32) -1.52 (1.55) -1.49 (1.61) -1.64 (1.63) Attitude strength .41 (.50) .81 .43 (.52) .81 .69 (.56) 1.24 1.33 (.63)* 2.12 1.32 (.63)* 2.10 General defensive confidence .04 (.33) .12 -.52 (.46) -1.15 -.84 (.50)+ -1.69 -.83 (.50)+ -1.68 Specific defensive confidence .80 (.45)+ 1.77 1.40 (.49)** 2.85 1.35 (.49)** 2.75 Issue publics -.44 (.15)** -2.96 -.42 (.15)** -2.78 Reasoning style .14 (.21) .68 Akaike corrected 504.97 505.40 509.10 525.12 526.68
Note. n = 115. Only respondents who selected the article about reducing income differences in one of the four trials are included.
+p<.10, *p<.05, **p<.01
An interesting picture was derived however from the fixed effects of the different factors on the selection of likeminded content about reducing income differences (see table 6, model 5). These results indicated that participants with a stronger attitude selected more likeminded stories about reducing income differences than participants with a weaker attitude, b = 1.32, SE = .63, p = .038. These results were in accordance with H1. In addition, a marginally significant effect was found for the level of defensive confidence, where participants with a higher level of general defensive confidence were more likely to select counter-attitudinal information than
participants with a lower level of general defensive confidence, b = -.83, SE = .50, p = .097. This was also an expected effect, but it did not provide conclusive evidence for the support of H2, because the level of specific defensive confidence had a significant positive effect on the selection of likeminded content, b = 1.35, SE = .49, p = .007. This result indicated that participants who felt they could successfully defend their ideas and opinions about reducing income differences specifically were more likely to select likeminded content than participants who felt less confident about these specific ideas and opinions. Therefore, only partial support was found for H2 for the issue of reducing income differences. Furthermore, members of the issue public were more likely to select counterattitudinal content than nonmembers of the issue public, b = -.42, SE = .15, p = .006. This result was expected in H3. Finally, the results provided no significant effect for the reasoning style of participants and the selection of likeminded content, b = .14, SE = .21, p = .500. Therefore, H4 was rejected.
Taken together, partial support was found for the effects of the personal predispositions on the selection of likeminded content about reducing income
differences. Figure 1 shows the mean selection scores for the significant factors. The selection of likeminded content increased among participants with a stronger attitude towards reducing income differences (see panel 1, figure 1) and non-members of the issue public (see panel 2, figure 1). In addition, selection of counter-attitudinal content increased among participants with a higher level of general defensive confidence, yet a higher level of specific defensive confidence increased the selection of likeminded content about reducing income differences (see panels 3 and 4, figure 1).
Figure 1. Mean scores for the significant factors in the fixed effects model predicting the selection of likeminded information about reducing income differences. 0 = Selected counter-attitudinal content, 1 = Selected likeminded content.
Finally, to test H5, a moderation analysis was conducted in order to examine whether issue publics were less reliant on their level of defensive confidence in selecting counter-attitudinal content than non-members of the issue public (table 7). The findings showed support for a moderation effect between issue public
membership and defensive confidence and the selection of counter-attitudinal content. A main effect of issue publics indicated that members of the issue public selected more counter-attitudinal information than non-members, an effect also found in the fixed effects model (table 6), but the general level of defensive confidence moderated this effect, b = -.62, SE = .27, p = .023. In addition, the specific level of defensive confidence also moderated the effect of issue public membership on the selection of counter-attitudinal content, b = -.34, SE = .17, p = .046. This implied that for
members of the reducing income differences issue public, the level of general defensive confidence and the level of specific defensive confidence were less
important than for non-members of the issue public. Non-members of the issue public
0 0,2 0,4 0,6 0,8 1 1. Attitude strength Selection of lik emi nd ed c on te nt Weak Medium Strong 0 0,2 0,4 0,6 0,8 1 2. Issue publics Selection of lik emi nd ed c on te nt Non-member Member 0 0,2 0,4 0,6 0,8 1 3. General defensive confidence Selectino of lik emi nd ed c on te nt Low Medium High 0 0,2 0,4 0,6 0,8 1 4. Specific defensive confidence Selection of lik emi nd ed c on te nt Low Medium High
relied more on their level of defensive confidence when selecting counter-attitudinal content than members of the issue public. Therefore, H5 was supported: the level of defensive confidence was less influential in the selective exposure habits of members of an issue public than in the selective exposure habits of non-members of an issue public.
Table 7
Moderation Effects with Centered Values for Issue Public Membership and General and Specific Defensive Confidence
Moderation effect b(SE) t 95% CI
Issue Publics x General Defensive Confidence Issue publics centered
General defensive confidence centered Moderation -.38(.14)** -.10(.36) -.62(.27)* -2.67 -.27 -2.31 [-.66, -.10] [-.80, .61] [-1.15, -.09] Issue Publics x Specific Defensive Confidence
Issue publics centered
Specific defensive confidence centered Moderation -.18(.12) .10(.26) -.34(.17)* -1.53 .38 -2.02 [-.42, .05] [-.42, .62] [-.66, -.01] Note. n = 115. Only respondents who selected the article about reducing income differences are included.
*p<.05, **p<.01
Conclusion and discussion
The motivation for this study was to assess four personal predispositions believed to be relevant for the selection of likeminded information about political issues because they influence the occurrence of cognitive dissonance and the ability to engage in cognitive reflection: the strength of the attitude towards the political issue, the level of defensive confidence, membership of the issue public and the style of reasoning. The individual influence of these four factors has been empirically supported in previous studies (Albarracín & Mitchell, 2004; Brannon et al., 2007; Chen, 2013; Hart et al., 2009; Kahan, 2013; Kim, 2009; Knobloch-Westerwick & Kleinman, 2012; Taber & Lodge, 2006), and the goal of this study was to test how
these predispositions relate to one another and what their partial effect is on the selection of likeminded content of political articles. Two political issues were assessed, representing the cultural and the economic dimension of politics (Van Der Brug & Van Spanje, 2009). The results for the two issues were clearly distinct.
When predicting the selection of likeminded content for the immigration article, none of the tested personal predispositions exerted a significant individual or partial effect. For the article about reducing income differences, however, the
selection of likeminded content increases as the strength of the attitude increases and among non-members of the issue public. These findings are in accordance with previous results (Brannon et al., 2007; Chen, 2013; Hart et al., 2009; Kim, 2009). In addition, a moderation analysis revealed that the level of defensive confidence, both in general and specific to this issue, acts as a moderator on the relationship between issue public membership and the selection of counter-attitudinal information. The perception of the personal importance and relevance of the issue seems to override defensive confidence. However, people who do not perceive the issue to be personally important or relevant rely on their defensive confidence to a greater extent in their selective exposure habits. Non-members with a low level of defensive confidence are less likely to select counter-attitudinal information than members of the issue public with a low level of defensive confidence. Finally, no support is found for the
influence of reasoning style on selective exposure. Whether people are conscious and effortful in their reasoning or more heuristic-driven has no effect on the selection of likeminded content. These findings are in contrast with previous work (Kahan, 2013; Kahneman 2003).
For the first time, research has now shown a difference in selective exposure to political news from different political dimensions. The results imply that personal
predispositions are not accountable for selective exposure to cultural issues, but they do influence selective exposure to economic issues. An uncontrollable environmental factor might account for the results for immigration. This study was conducted in the Netherlands during a great migration crisis. Dutch citizens are confronted with political news about immigration on a daily basis, resulting in a primed issue that has become top-of-mind. Knobloch-Westerwick and Kleinman (2012) found that
individuals want to anticipate upcoming political circumstances, whereby information utility overrides the confirmation bias. Therefore, personal predispositions are no longer relied upon to avoid cognitive dissonance. However, caution is advised when ascribing the findings to only this environmental factor. Mind that the Netherlands is also in a period of recession, a consequence of the recent economic crisis. Perhaps cultural issues are in general considered a municipal good more often and more quickly than economic issues, which in this study seem to be more personally bound. Cultural issues then affect us as a community, increasing information utility
(Knobloch-Westerwick & Kleinman, 2012). Yet, issues that affect us personally might increase our desire to draw preferred conclusions (Kunda, 1990), increasing the likelihood of selective exposure through cognitive dissonance reduction (Festinger, 1957). Future research should take these considerations into account when replicating the findings.
This study uses a unique research design, one that approaches a real-world setting most closely, resulting in a more precise reflection of real-world habits (Arceneaux & Johnson, 2007; Feldman et al., 2013). Most past studies have worked in an experimental setting with a research design that limits the generalizability of the results because it does not reflect a real-world setting (Brannon et al., 2007; Graf & Aday, 2008; Meffert et al., 2006). This study overcomes this limitation by offering
participants an actual reflection of a newspaper, raising the reliability of the results and improving its generalizability. The use of a new research design might explain the unsupported hypotheses, which are mostly based on findings from situations where people could choose between likeminded and counter-attitudinal content. Here that option was not given and shows how findings from experimental settings are not replicated in a real-world setting. It calls for further investigation of previous, limited generalizable results and questions our knowledge of the influence of personal predispositions on the selection of likeminded content.
The results of this study have implications for political communication research because it extends our knowledge of selective exposure, especially among issue public members. Membership to an issue public has proven to lead to highly interested groups of individuals, who do not need to be highly knowledgeable to form a strong and stable attitude about the issue (Chen, 2013; Kim, 2009; Krosnick, 1990). Now, it appears that issue public members also do not need to be confident of their ability to defend their own ideas and opinions regarding their preferred issue to expose themselves to counter-attitudinal information. Their intrinsic interest in the issue overrides this obstacle of defending ones own ideas successfully. Future research can further investigate the interaction between issue importance, defensive confidence and attitude strength. Where Albarracín and Mitchell (2004) and Brannon et al. (2007) found that people with strong attitudes only expose themselves to
counter-attitudinal information when they are high in defensive confidence, results from this study imply that this relationship in influenced by issue importance.
Lastly, this study has some limitations. First, it should be mentioned that the sample size is somewhat modest and more highly educated than the general Dutch public. To overcome the modest sample size, the repeated measures approach has
slightly improved the reliability, since it multiplies the sample size. While recent research states that unrepresentative samples provide results similar to representative samples in experiments (Mullinix et al., 2015), a larger sample is advised for future research because it might produce other results and increase the power of the study. Second, the participants were forced to choose at least one article from the front-page and were not offered an option to avoid all articles. Such forced responses can skew the results (Arceneaux & Johnson, 2007) and damage the external validity.
Additionally, people are more prone to select the article at the top of the list, because they perceive it to be more important or significant, resulting in a primacy effect (Murdock, 1962). The immigration article was the top story of the front-page of the online newspaper in every condition. The forced responses might have strengthened this primacy effect. Finally, the scale used for the level of general defensive
confidence was reliable but not valid. The decision to use the scale as introduced by Albarracín and Mitchell (2004) was made because research had proven it to be valid in various situations. Yet, it should be mentioned that in this study the scale was two-dimensional and is therefore a threat to the internal validity.
Does this study document something that should concern citizens in a democratic system? Yes and no. Personal predispositions limit individuals’
information gathering, fuel cognitive biases and motivate us to only seek that what strengthens our prejudices. A democracy can only function when citizens contribute to the political discussion. However, the results in this study imply that our
predispositions do not limit our involvement in every political issue. The trick is to be aware of our personal prejudices facilitating a one-sided information flow that
restricts political involvement in this digital era of communication, where we are able to structure our information flow to our own wishes.
References
Albarracín, D., & Mitchell, A. L. (2004). The Role of Defensive Confidence in Preference for Proattitudinal Information: How Believing That One Is Strong Can Sometimes Be a Defensive Weakness. Personality and Social Psychology Bulletin, 30(12), 1565-1584.
Arceneaux, K., & Johnson, M. (2007, April). Channel Surfing: Does Choice Reduce Videomalaise? Paper presented at the annual meeting of the Midwest Political Science Association, Chicago.
Arceneaux, K., & Johnson, M. (2010, September). Does Media Fragmentation Produce Mass Polarization? Selective Exposure and a New Era of Minimal Effects. Paper presented at the annual meeting of the American Political Science Association, Washington.
Arceneaux, K., Johnson, M., & Cryderman, J. (2013). Communication, Persuasion, and the Conditioning Value of Selective Exposure: Like Minds May Unite and Divide but They Mostly Tune Out. Political Communication, 30(2), 213-231. Atkin, C. K. (1971). Instrumental utilities and information seeking. In P. Clarke (Ed.),
New models for communication research (pp. 205-239). Beverly Hills: Sage Publication.
Bawden, D., & Robinson, L. (2009). The dark side of information: overload, anxiety and other paradoxes and pathologies. Journal of Information Science, 35(2), 180-191.
Bennett, W. L., & Iyengar, S. (2008). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58(4), 707-731.
Publics. Political Communication, 30(3), 329-348.
Brannon, L. A., Tagler, M. J., & Eagly, A. H. (2007). The moderating role of attitude strength in selective exposure to information. Journal of Experimental Social Psychology, 43(4), 611-617.
Campitelli, G., & Labollita, M. (2010). Correlations of cognitive reflection with judgments and choices. Judgment and Decision Making, 5(3), 182-191. Cappella, J. N., Kim, H. S., & Albarracín, D. (2015). Selection and transmission
processes for information in the emerging media environment: Psychological motives and message characteristics. Media psychology, 18(3), 396-424. Caprara, G. V., & Zimbardo, P. G. (2004). Personalizing politics: a congruency model
of political preference. American Psychologist, 59(7), 581-594. CBS (2016, May 17). Hoogstbehaald onderwijsniveau en
onderwijsrichting. [Quarterly review of education level]. CBS StatLine
(Second quarter, population). Centraal Bureau voor de Statistiek, The Hague. Chaiken, S., & Trope, Y. (1999). Dual-process theories in social psychology. New
York: Guilford Press.
Chen, H. (2013). Capturing the Nature of Issue Publics: Selectivity, Deliberation, and Activeness in the New Media Environment (Doctoral dissertation). Retrieved from https://repositories.lib.utexas.edu.
Converse, P. E. (2006). The Nature of Belief Systems in Mass Publics (1964). Critical Review, 18(1-3), 1-74.
Feldman, L., Stroud, N. J., Bimber, B., & Wojcieszak, M. (2013). Assessing Selective Exposure in Experiments: The Implications of Different Methodological Choices. Communication Methods and Measures, 7(3-4), 172-194.
Fischer, P., & Greitemeyer, T. (2010). A New Look at Selective-Exposure Effects: An Integrative Model. Current Directions in Psychological Science, 19(6), 384-389.
Frederick, S. (2005). Cognitive Reflection and Decision Making. The Journal of Economic Perspectives, 19(4), 25-42.
Garrett, R. K. (2013). Selective Exposure: New Methods and New Directions. Communication Methods and Measures, 7(3-4), 247-256.
Graf, J., & Aday, S. (2008). Selective attention to online political information. Journal of Broadcasting & Electronic Media, 52(1), 86-100.
Hart, W., Albarracín, D., Eagly, A. H., Brechan, I., Lindberg, M. J., & Merrill, L. (2009). Feeling validated versus being correct: a meta-analysis of selective exposure to information. Psychological bulletin, 135(4), 555-588.
Hopmann, D. N., Elmelund-Praestekaer, C., Albæk, E., Vliegenthart, R., & De Vreese, C. H. (2012). Party media agenda-setting. How parties influence election news coverage. Party Politics, 18(2), 173-191.
Iyengar, S., & Hahn, K. S. (2009). Red media, blue media: Evidence of ideological selectivity in media use. Journal of Communication, 59(1), 19-39.
Iyengar, S., Hahn, K. S., Krosnick, J. A., & Walker, J. (2008). Selective Exposure to Campaign Communication: The Role of Anticipated Agreement and Issue Public Membership. The Journal of Politics, 70(1), 186-200.
Kahan, D. M. (2013). Ideology, motivated reasoning, and cognitive reflection. Judgment and Decision Making, 8(4), 407-424.
Kahneman, D. (2003). Maps of Bounded Rationality: Psychology for Behavioral Economics. American Economic Review, 93(5), 1449-1475.
Investigating the Moderating Effects of Situational Information Processing Goals in Issue Publicsʼ Web Behavior. Communication Research, 34(2), 185-211.
Kim, Y. M. (2009). Issue Publics in the New Information Environment: Selectivity, Domain Specificity, and Extremity. Communication Research, 36(2), 254-284.
Knobloch-Westerwick, S., & Kleinman, S. B. (2012). Preelection Selective Exposure: Confirmation Bias Versus Informational Utility. Communication Research, 39(2), 170-193.
Knobloch-Westerwick, S., & Meng, J. (2009). Looking the other way: Selective exposure to attitude-consistent and counterattitudinal political information. Communication Research, 36(3), 426-448.
Knobloch-Westerwick, S., & Meng, J. (2011). Reinforcement of the Political Self Through Selective Exposure to Political Messages. Journal of
Communication, 61(2), 349-368.
Kraaykamp, G., Van Snippenburg, L. B., & Ultee, W. C. (2015). Occupational class and occupational status as an explanation of income, attitudes towards
inequality, political confidence and political apathy. Mens en Maatschappij, 64(1), 42-65.
Kriesi, H., Grande, E., Lachat, R., Dolezal, M., Bornschier, S., & Frey, T. (2006). Globalization and the transformation of the national political space: Six European countries compared. European Journal of Political Research, 45(6), 921-956.
Krosnick, J. A. (1990). Government Policy and Citizen Passion: A Study of Issue Publics in Contemporary America. Political Behavior, 12(1), 59-92.
Krosnick, J. A., & Telhami, S. (1995). Public attitudes toward Israel: A study of the attentive and issue publics. International Studies Quarterly, 39(4), 535-554. Kunda, Z. (1990). The Case for Motivated Reasoning. Psychological Bulletin, 108(3),
480-498.
Levendusky, M. S. (2013). Why do partisan media polarize viewers? American Journal of Political Science, 57(3), 611-623.
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.
Mullinix, K. J., Leeper, T. J., Druckman, J. N., & Freese, J. (2015). The
generalizability of survey experiments. Journal of Experimental Political Science, 2(2), 109-138.
Murdock, B. B. (1962). The serial position effect of free recall. Journal of experimental psychology, 64(2), 482-488.
Popkin, S. L. (1991). The reasoning voter: Communication and persuasion in presidential campaigns. Chicago: University of Chicago Press.
Prior, M. (2005). News vs. entertainment: How increasing media choice widens gaps in political knowledge and turnout. American Journal of Political Science, 49(3), 577-592.
Prior, M. (2013). Media and Political Polarization. Annual Review of Political Science, 16(1), 101-127.
Sampson, D. L., & Smith, H. P. (1957). A Scale to Measure World-Minded Attitudes. The Journal of Social Psychology, 45(1), 99-106.
Stroud, N. J. (2010). Polarization and partisan selective exposure. Journal of Communication, 60(3), 556-576.
