Professional Documents
Culture Documents
Johnson 2019
Johnson 2019
Benjamin K. Johnson, Rachel L. Neo, Marieke E.M. Heijnen, Lotte Smits, Caitrina van
Veen
PII: S0747-5632(19)30367-X
DOI: https://doi.org/10.1016/j.chb.2019.09.031
Reference: CHB 6155
Please cite this article as: Johnson B.K., Neo R.L., Heijnen M.E.M., Smits L. & van Veen C., Issues,
involvement, and influence: Effects of selective exposure and sharing on polarization and participation,
Computers in Human Behavior (2019), doi: https://doi.org/10.1016/j.chb.2019.09.031.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
Benjamin K. Johnsona*
Rachel L. Neob
Marieke E. M. Heijnenc
Lotte Smitsc
Caitrina van Veenc
a
University of Florida, Department of Advertising
benjaminkjohnson@ufl.edu
b
University of Hawaii at Manoa, School of Communications
c
Vrije Universiteit Amsterdam, Department of Communication Science
De Boelelaan 1081
*Corresponding author.
Abstract
Although research has amply demonstrated that people exhibit confirmatory biases associated
with exposure and information sharing on social media, there is a lack of research attempting to
parse out the respective effects of selective exposure and sharing on political outcomes,
especially in non-U.S. contexts. In this experiment, we tested the extent of confirmation bias in
Dutch Facebook users’ selection and sharing of opinionated news about three political issues.
messages were assessed for their influences on attitude polarization and political participation.
Value- and impression- involvement were considered as moderating factors. Findings indicate
that a confirmation bias is much more consistently observed in selective sharing than in selective
outcomes than pro-attitudinal selective exposure. Third, the effects of selective sharing and
exposure on political outcomes depend more on value involvement than impression relevant
involvement. Finally, between-topic differences were evident for the extent of confirmation bias
1. Introduction
The existence of a confirmation bias in the selection and sharing of news and information
tendency of people to choose and spend more time reading attitude-consistent information, can
yield effects on outcomes such as attitude polarization, feelings toward outgroups, voting
intentions, and political participation intentions (Feezell, 2016; Garrett et al., 2014; Wojcieszak,
Bimber, Feldman, & Stroud, 2016). In addition to attitude-consistent selective exposure, people
may also engage in attitude-consistent selective sharing. The widespread adoption of social
network sites such as Facebook and Twitter have created new platforms for partisan sharing of
news and political messages (Beam, Hutchens, & Hmielowski, 2016; Oeldorf-Hirsch & Sundar,
or selective sharing on social media influence political outcomes (e.g., Shin & Thorson, 2017;
Weeks et al., 2017), existing research on this topic still has several key limitations. First, most
studies have only examined either selective exposure or selective sharing as focal variables, but
not both. Some scholars have argued that selective sharing is a more overtly clarion reflection of
partisan leanings than selective exposure (Shin & Thorson, 2017) and that it involves greater
behavioral commitment (Lane et al., 2019). Our study thus tests the effects of (a) attitudinal
stance on selective exposure versus selective sharing and (b) how both selective exposure and
selective sharing on social media affect political outcomes among the same individuals. This
allows us to assess which of these two confirmation biases is a stronger predictor of political
outcomes.
Second, few studies on selective exposure or sharing have accounted for the moderating
EXPOSURE AND SHARING EFFECTS 3
role of involvement (Liao & Fu, 2013; Wheeless, 1974). Research has shown that involvement
levels affect partisan-based political behaviors (Perloff, 1989). However, not everyone exhibits
deep-seated levels of value involvement toward any given political issue or exhibits similar
Third, most studies on selective exposure (Bakshy, Messing, & Adamic, 2015) or sharing
on social media (e.g., Weeks et al., 2017) have been conducted in the United States. The United
States has a unique two-party political system with an electorate that is heavily polarized along
party lines (Garrett et al., 2014). It is imperative to conduct more experimental research outside
of the U.S. in order to provide stronger evidence for these two aforementioned confirmation
biases, especially in democracies such as the Netherlands with parliamentary systems involving a
fragmented diversity of political parties that must typically work together in coalition to form a
functioning government (Pellikaan, de Lange, & van der Meer, 2018). In the Netherlands,
specifically, there have been well-designed tests of partisan selective exposure using panel data
(e.g., Bos, Kruikemeier, & de Vreese, 2016), a quasi-experimental design (Trilling, van
Klingeren, & Tsfati, 2017), and experiments testing the effects of frames on selective exposure
(Brenes Peralta, Wojcieszak, Lelkes, & de Vreese, 2017; Hameleers, Bos, & de Vreese, 2018).
However, there is little to no evidence to date on selective sharing in Dutch contexts, and more
Fourth, most studies have examined selective exposure exclusively in the context of
highly polarized topics (e.g., Westerwick, Johnson, & Knobloch-Westerwick, 2017). In our
study, we use several topics which vary in their existing polarization and salience (cf. Y. M.
Kim, 2009).
EXPOSURE AND SHARING EFFECTS 4
To bridge these research gaps outlined above, this study reports the results of an
experiment in which compares the relative influence of selective exposure and sharing (to pro-
polarization. In particular, we investigate these processes in the contexts of three political topics
chosen for their varying relevance and salience for a Dutch sample: the relationship between
Ukraine and the EU; the entry of refugees into the Netherlands; and equal pay for men and
2. Literature review
Numerous studies have shown that individuals’ attitudinal stances will guide the
composition and degree of their information exposure (Garrett, 2009a; Hart et al., 2009;
Knobloch-Westerwick, 2015). Notably, confirmation bias describes the tendency where people
will gravitate toward information that aligns with their existing views (Lord, Ross, & Lepper,
1979). The psychological processes underlying confirmation bias can be explained by the theory
of cognitive dissonance (Festinger, 1957), which posits that people strive to minimize
psychological discomfort arising from performing actions that are inconsistent with their values
and beliefs. One such action is that of counter-attitudinal political news consumption (Jeong, Zo,
Lee, & Ceran, 2019). Although people do not avoid counter-attitudinal political information
entirely (Garrett, Carnahan, & Lynch, 2013), they nonetheless spend less time reading counter-
attitudinal than pro-attitudinal news stories (e.g., Frey, 1986; Garrett, 2009a).
Some scholars have argued that the control afforded by online platforms facilitates the
EXPOSURE AND SHARING EFFECTS 5
sheer volume of content made available by digital platforms may heighten the ability, desire, or
necessity for user selectivity in news use (S. Lee, Lindsey, & Kim, 2017). Indeed, research has
consistently demonstrated robust support for confirmation bias in online contexts such as social
The term selective exposure is often used to refer to an observed confirmation bias in
information exposure, where pro-attitudinal content is selected at a higher rate (Stroud, 2008) or
following Knobloch-Westerwick (2015), such that it refers to any systematic pattern in media
use. This accounts for a broader set of motivations for media use (including accuracy, impression
management, novelty, entertainment, etc.) as well as the possibility that some individuals, under
certain situations, may exercise a selective preference for counter-attitudinal messages (e.g., if
that information provides more utility; Knobloch-Westerwick & Kleinman, 2012). However, our
starting point for this investigation is a general expectation that a confirmation bias will be
H1: People will spend more time reading pro-attitudinal than counter-attitudinal news
2.2. The effect of selective exposure on political participation and opinion polarization
have effects on variables such as opinion polarization and political participation. Research has
demonstrated a robust positive link between pro-attitudinal news consumption and political
participation (Knobloch-Westerwick & Johnson, 2014; Stroud, 2010). And, the more people
attend to attitude-reinforcing political information, the more they will exhibit strong and even
EXPOSURE AND SHARING EFFECTS 6
extreme attitudes. Such confidence emboldens people to take political action, so that attitude-
affirming news on social media is likely to foster both online and offline political participation
(Beam, Hutchens, & Hmielowski, 2018; Feezell, 2016; Y. Kim & Chen, 2016).
Social media provide ease of access to content, coupled with the presence of social peers,
making it an important context for interacting with news and political information. As such,
social media appear to be especially consequential for political participation (Bode, Vraga,
Borah, & Shah, 2014; Boulianne, 2015). Furthermore, consuming attitude-affirming news
increases the salience of a person’s political identity, thereby causing a person to think and
behave in ways that signify their party affiliation (Stroud, 2010). For instance, selective exposure
causes people to develop more polarized opinions on controversial issues that align with their
party’s values (Y. Kim, 2015; Stroud, 2010; Westerwick, Johnson, & Knobloch-Westerwick,
2017). From this, we can expect that accessing pro-attitudinal news in a social media context will
H2: The amount of time spent reading pro-attitudinal news stories will be positively
associated with (a) online political participation, (b) offline participation, and (c) opinion
polarization.
the views espoused by pro-attitudinal information (Garrett et al., 2014). People might gain a
better understanding of their political opponents’ issue positions through exposure to counter-
consumption might make people hold ambivalent issue attitudes and fewer intentions to engage
might increase participation intentions (Y. Kim & Chen, 2016), especially if incivility is present
(Hwang, Kim, & Huh, 2014). And, some research on the relationship between counter-attitudinal
information exposure and opinion polarization shows null or nuanced results, such that exposure
to ideologically diverse content online might not influence polarization outcomes (J. Lee, Choi,
Kim, & Kim, 2014) or might yield a mix of polarization and depolarization effects, depending on
diminishes political behavior (Matthes, 2012) and weakens (i.e., depolarizes) attitudes
(Westerwick et al., 2017). In the social media context, recent surveys with Americans found that
counter-attitudinal discussion on social media was linked to less political participation (Lu,
Heatherly, & Lee, 2016) and counter-attitudinal news exposure on Facebook produced
depolarization over time (Beam et al., 2018). To that end, the evidence suggests that counter-
attitudinal political information exposure will reduce the degree to which issue opinions are
H3: The amount of time spent reading counter-attitudinal news stories will be negatively
associated with (a) online political participation, (b) offline participation, and (c) opinion
polarization.
The presence of an expressive goal has been shown to enhance confirmation bias in
message exposure (Smith, Fabrigar, Powell, & Estrada, 2007). This need to present one’s
identity and opinions are even more pertinent when it comes to selective sharing, as the sharing
is often directed at a large audience (if not public) and plays a role in individual self-presentation
(Coppini et al., 2017). In the context of social media, news sharing encompasses outwardly
EXPOSURE AND SHARING EFFECTS 8
expressive activities such as recommending, posting, or forwarding various forms of news and
political information to the members of one’s social network (Kümpel, Karnowski, & Keyling,
2015), plus other interactions such as comments or likes which signal engagement with news.
Network analyses of social media data have consistently shown that people selectively
share attitude-affirming political information such as hyperlinked blogposts (Adamic & Glance,
2005) or tweets (Barbera, Jost, Nagler, Tucker, & Bonneau, 2015; Colleoni, Rozza, &
Arvidsson, 2014). However, such big data analyses have certain limitations. For instance, big
data analyses are often restricted to publicly available data, potentially yielding unrepresentative
samples (Garrett, 2013). Thus, it is equally important to corroborate findings from analyses of
big data with research carried out using other types of methodologies such as surveys or
experiments (Garrett, 2013). Furthermore, survey and experimental designs allow for greater
inferences about mental processes and causal steps, and have unique strengths with regard to
selective exposure research (Clay, Barber, & Shook, 2013). As such, this study uses an online
experiment embedded within a web-based behavior-tracking study of social media news use to
examine whether people are more likely to share pro-attitudinal social media news items over
much more consistently observed phenomenon on social media platforms than selective
exposure (Shin & Thorson, 2017). People do not always fully attune themselves to pro-
attitudinal information (Garrett et al., 2013). There are times when they will deliberately attend
to counter-attitudinal political information, e.g., because they want to develop arguments against
attitudinal information could create conflict or misunderstanding with one’s network. Pro-
EXPOSURE AND SHARING EFFECTS 9
attitudinal information sharing on social media represents an overt reflection of one’s partisan
values (Shin & Thorson, 2017). Notably, Shin and Thorson (2017) found that people are more
likely to re-tweet fact-checking messages that favor the in-party over those that favor the out-
party as an overt statement of loyalty and support for their in-party. All in all, people may be less
H4: People will be more likely to share pro-attitudinal than counter-attitudinal news
RQ1: Is the sharing of pro- versus counter-attitudinal news stories on social media a
2.4 The effect of selective sharing on political participation and political polarization
Current research has mostly focused on establishing the positive relationship between
generalized political information sharing and political participation (Gil de Zúñiga, Molyneux, &
Zheng, 2014; Vaccari et al., 2015). Online information sharing increases political participation
by increasing issue knowledge and enabling people to co-organize political events (Kwak,
Williams, Wang, & Lee, 2005; F. Lee, Chen, & Chan, 2017; Valenzuela, 2013). Given that
leanings (Shin & Thorson, 2017), it is likely that such selective sharing will make one’s partisan
identity even more cognitively salient than selective exposure alone. Although there is some
cross-sectional evidence that news sharing on social media may increase network heterogeneity
(i.e., contact with others with divergent opinions; Choi & Lee, 2015), it can foster relationships
with like-minded others. News sharing can also produce feelings of certainty and efficacy
(Oeldorf-Hirsch & Sundar, 2015). It is a logical inference that the sharing of pro-attitudinal
information will have stronger effects than pro-attitudinal news exposure on empowering people
EXPOSURE AND SHARING EFFECTS 10
to collaborate with other in-party members on events or causes benefiting their political party.
of issue stance, the sharing of pro-attitudinal information is arguably a much more effective
catalyst of partisan beliefs and affiliation than selective exposure (Shin & Thorson, 2017). When
people engage in such partisan-based social categorization, opinion polarization occurs as they
develop extreme views about group norms toward issue positions (Hogg, 2014). Panel survey
data suggest that social media use contributes to political engagement and subsequent
polarization of attitudes (C. Lee, Shin, & Hong, 2018). This study proposes that:
H5: The sharing of pro-attitudinal news stories on social media will be positively
associated with (a) online political participation, (b) offline political participation, and (c)
opinion polarization.
RQ2: Is the sharing of pro-attitudinal news stories on social media a more consistently
positive predictor of (a) online political participation, (b) offline political participation,
participation and opinion polarization are less understood. As outlined above, people are more
likely to share pro-attitudinal over counter-attitudinal information (Shin & Thorson, 2017).
However, scholars have pointed out that the act of sharing counter-attitudinal political
information indicates an outward openness toward embracing, or at the very least reflecting
upon, alternative political views (Lane et al., 2019). This suggests that people who openly share
counter-attitudinal political information are willing to openly admit that they are politically
undecided, and will consequently be less likely to take concrete political action.
EXPOSURE AND SHARING EFFECTS 11
Furthermore, when people are unabashedly willing to consider the validity of alternative
viewpoints (Mutz, 2002), it is likely that they will not hold extreme issue positions that conform
information (Lane et al., 2019). It is thus plausible that sharing counter-attitudinal political
information will be even more effective than consuming counter-attitudinal political information
H6: The sharing of counter-attitudinal news stories on social media will be negatively
associated with (a) online political participation, (b) offline political participation, and (c)
opinion polarization.
consistently negative predictor of (a) online political participation, (b) offline political
stories?
The construct of involvement has played a crucial role in explaining how message
processing affects judgments across many domains of media and communication research (e.g.,
Brown & Basil, 1995; Kwak, 1999; Slater & Rouner, 1992, 1996). Scholars have identified three
main types of involvement: value relevant, outcome relevant, and impression relevant (Cho &
Boster, 2005). Although all three types of involvement trigger attitudes related to one’s self-
concept, each type of involvement results in different persuasive outcomes (Cho & Boster,
2005). To date, few studies have examined how these specific types of involvement affect the
degree to which selective exposure and sharing behaviors predict political outcomes. This study
EXPOSURE AND SHARING EFFECTS 12
the activation of attitudes that are linked to important values” (Johnson & Eagly, 1989, p. 290).
A person’s political identity and values are integral to their self-concept (Sherif, Sherif, &
Nebergall, 1965), and will inform their judgments and political information consumption
acceptable beliefs narrows, and they tend to reject a wide set of attitude positions (Cho & Boster,
2005; Fazio, Zanna, & Cooper, 1977). Under such conditions, these types of people will be prone
to engaging in behaviors or exhibiting attitudes that align with their values or beliefs, and
correspondingly just as likely to avoid adopting value-inconsistent behaviors and attitudes (Fazio
et al., 1977).
Given the above, we can predict that individuals with high levels of value involvement
affirm their political values, such as selectively consuming or sharing pro-attitudinal social
media news. In addition, people with high levels of value involvement are particularly likely to
be politically active, and tend to have very extreme opinions (Perloff, 1989). It is plausible that
such pro-attitudinal political information consumption and sharing behaviors will translate into
political action and opinion polarization at high levels of value involvement. By contrast,
consuming and sharing counter-attitudinal political information are behaviors that ostensibly
conflict with one’s political values (e.g., Lane et al., 2019). People with high levels of value
to shape their political opinions or decisions to partake in political activities. As such, we predict
that the negative effects of counter-attitudinal news exposure and sharing on political
participation and opinion polarization will be weakest when value involvement is high.
H7: People with higher levels of value involvement will be more likely to (a) spend time
reading and (b) share pro-attitudinal versus counter-attitudinal news stories on social
media.
H8: People with higher levels of value involvement will show stronger positive effects of
pro-attitudinal news exposure on (a) online political participation, (b) offline political
H9: People with higher levels of value involvement will show weaker negative effects of
counter-attitudinal news exposure on (a) online political participation, (b) offline political
H10: People with higher levels of value involvement will show stronger positive effects
of pro-attitudinal news sharing on (a) online political participation, (b) offline political
H11: People with higher levels of value involvement will show weaker negative effects
involvement that gauges the degree to which individuals inherently care about the social
concern about how others perceive them (Cho & Boster, 2005). Thus, people who exhibit high
levels of impression involvement tend to deliberately engage in public displays of behavior that
in-party solidarity in front of an imagined audience (Liu, Rui, & Cui, 2017; Shin & Thorson,
the degree to which a person embodies the values of their political affiliation, or seeks greater
affiliation (Huddy, Mason, & Aarøe, 2015). With this in mind, it is plausible that those with high
levels of impression involvement will be most likely to share pro-attitudinal over counter-
attitudinal social media news, and in turn become more politically active after selectively sharing
pro-attitudinal news as explicit signs that they conform to in-party norms. By contrast, sharing
counter-attitudinal political news is antithetical to partisan group norms (Lane et al., 2019).
Consequently, individuals who exhibit high levels of impression involvement might be reluctant
to openly share counter-attitudinal political news and allow such news items to undermine their
desire to participate in political activities. When they do share such content and make a semi-
opinions and dampen their participation intentions. As such, the negative effects of counter-
attitudinal social media news exposure on online and offline political participation is likely to be
one’s political values than political participation (Huddy et al., 2015). Regardless of opinion
extremity, people can choose to keep their opinions to themselves without letting others know
(Cho & Boster, 2005). However, shared opinions are likely to lead to commitment and
EXPOSURE AND SHARING EFFECTS 15
reinforcement (Lane et al., 2019). Thus, the moderating role of impression involvement on the
relationship between attitudinal stance and opinion polarization via selective sharing is different
from that of value involvement. Those who desire to make a positive impression on their social
network (Smith et al., 2007) should shift their beliefs and behaviors in the direction of the news
they have consumed, and especially the content they have shared. The following hypotheses are
proposed:
H12: People with higher levels of impression involvement will be more likely to (a)
spend time reading and (b) share pro-attitudinal versus counter-attitudinal news stories on
social media.
H13: People with higher levels of impression involvement will show stronger positive
effects of pro-attitudinal news exposure on (a) online political participation, (b) offline
H14: People with higher levels of impression involvement will show stronger negative
H15: People with higher levels of impression involvement will show stronger positive
effects of pro-attitudinal news sharing on (a) online political participation, (b) offline
H16: People with higher levels of impression involvement will show stronger negative
effects of counter-attitudinal news sharing on (a) online political participation, (b) offline
The study’s hypotheses and research questions are visualized in Figure 1, which illustrates the
within-subjects differences in pro- and counter-attitudinal exposure and sharing with brackets on
EXPOSURE AND SHARING EFFECTS 16
the left-hand side of the model, the between-subjects effects of exposure and sharing on the
dependent variables with paths on the right-hand side of the model, and the moderating effects of
value and impression involvement on main effects indicated with the moderator variables at the
Figure 1. Conceptual model with hypotheses and research questions. Brackets represent within-
subjects effects (tested with paired-sample t-tests, by topic) and paths to dependent variables
represent between-subjects effects (tested with a regression model for each DV, by topic; results
appear in Tables 1 and 2). Value and impression involvement were tested with subsequent
ANCOVAs and regression models. H7a tests moderation of H1, H7b tests moderation of H4,
H8a tests moderation of H2a, H8b tests moderation of H2b, and so on.
EXPOSURE AND SHARING EFFECTS 18
3. Method
An online experiment was conducted, using a 3 (issue: Ukraine vs. refugees vs. equal
pay) x 2 (stance: pro- vs. counter-attitudinal) within-subjects manipulation of topic and stance. A
select articles to read and subsequently express intentions to share. The study data, syntax for
3.1. Participants
A convenience sample of adult Dutch Facebook users was recruited via invitations
distributed through university students’ social media networks and flyers in campus and public
settings. As an incentive, participants were entered into a drawing for €5 giftcards. A total of 399
individuals completed the questionnaire. Three cases were removed because of invalid self-
reported age, another eight were removed because elapsed time to submit the survey was +/-3 SD
(i.e., > 5 hrs), and a further four were removed because total stimuli browsing time was +/-3 SD
(i.e., > 1 hr). This left a final sample for hypothesis testing of N = 384. This sample was 62.5%
women, and ranged in age from 18 to 68 (M = 29.72, SD = 10.59). The majority (96.6%) had
Dutch nationality and were born in the Netherlands (93.5%). The majority (63.4%) had earned a
3.2. Procedure
Data were collected from 21 March to 5 April 2016, prior to a referendum on the EU’s
relationship with Ukraine. After completing baseline attitudes and distractors, participants were
presented with an overview page of news articles, from which they could select one or more
intentions, involvement, and other variables. In general, participants reported moderate attitudes
EXPOSURE AND SHARING EFFECTS 19
toward the referendum, polarized attitudes toward refugees, and positive attitudes toward pay
equality (see details in Section 3.4). Neutral individuals on the three topics were omitted from
analysis.
3.3. Stimuli
Eight news articles were adapted from Dutch newspaper stories. Six reflected opposing
stances on the three target issues of (a) affirming the EU-Ukraine association in a referendum,
(b) accepting more migrant refugees to the Netherlands, and (c) ensuring pay equality for men
and women. The pro- or anti- stances of the articles was confirmed with a pre-test in which 20
Dutch university students rated headlines for their stance on the issue. The full articles were
rewritten and edited for length and to match the position and tone of their headlines. Two
distractor stories were also included, on soft news topics (single parenthood and monogamy).
Each article preview consisted of a headline and lead (word count: M = 32.75, SD =
3.65), a byline with name and timestamp (e.g., “N. Meyer, yesterday 11:03”). These eight
previews were presented (in a randomized order) in two columns and four rows on an overview
page, from which participants could select articles for further reading (they were instructed to
choose at least one). If a particular article was selected for reading, they were directed to a full
page that presented the entire article (word count: M = 651.5, SD = 43.83). When multiple
articles were selected, the presentation order of articles was varied. Time spent reading
individual articles was unobtrusively recorded by the questionnaire platform (Hastall &
Knobloch-Westerwick, 2013).
To enhance ecological validity, each article (and article preview) was assigned a number
of Facebook likes and reactions, and each article was assigned a number of Facebook comments
that appeared after the articles. The valence of reactions and comments was intended as a
EXPOSURE AND SHARING EFFECTS 20
manipulation of opinion climate, but did not yield effects and is not analyzed further.
3.4. Measures
3.4.1. T1 attitudes
against to 100 = completely in favor. Toward the Ukraine referendum, “the association
agreement between Ukraine and the EU,” M = 48.39, SD = 25.41; 5.7% did not report an
attitude, 30.2% reported a negative attitude, 31.8% reported a neutral attitude at the midpoint of
50, and 32.3% reported a positive attitude. Toward refugees, “allowing more migrants in the
Netherlands,” M = 64.14, SD = 26.45; 4.4% did not report an attitude, 22.4% reported a negative
attitude, 10.4% reported a neutral attitude at the midpoint, and 62.8% reported a positive attitude.
Toward equal pay, “pay equality between men and women,” M = 83.07, SD = 31.39; 4.2% did
not report an attitude, 14.1% reported a negative attitude, 3.9% reported a neutral attitude at the
midpoint, and 77.9% reported a positive attitude. In addition to these scalar attitudes, scores were
trichotomized for the purpose of classifying exposure and sharing as pro-attitudinal or counter-
Participants selected an average of 2.85 (SD = 1.53) of the eight available articles: six
political articles about the three target issues, plus two lifestyle news distractors. They spent M =
160.16 seconds, SD = 310.47, reading political articles and M = 45.58 seconds, SD = 77.45,
reading distractor articles. The time spent on political articles, by issue, and by stance, was
recoded to reflect whether the stance of the article was pro-attitudinal or counter-attitudinal for
the individual participant (based on their trichotomized attitudes). This procedure yielded
measures of time spent reading pro-attitudinal and counter-attitudinal articles for Ukraine,
EXPOSURE AND SHARING EFFECTS 21
After the browsing opportunity, participants were presented with the headline for each
article, one at a time, and asked three questions: “How likely is it that you would ‘like’ this
article on Facebook?” “How is it likely that you would ‘share’ this article on your timeline on
Facebook?” and “How is it likely that you would post a comment under this article on
Facebook?” Participants indicated their likelihood, from 1 = not likely to 7 = very likely. These
mean scores for these three items—liking, sharing, and commenting—were taken as an index of
sharing intentions (cf. Alhabash, Almutairi, Lou, & Kim, 2019), with α’s ranging from .689 to
.874 for all six political articles. Sharing intentions were variable but relatively low, with a grand
mean across articles of 1.61 (SE = .06). The least shared article type was counter-attitudinal
Ukraine (M = 1.31, SD = 0.75) and the most shared article type was pro-attitudinal refugees (M =
2.04, SD = 1.30).
Attitudes were re-administered for each issue with 101-point thermometers. Polarization
was measured by subtracting t1 attitudes from t2 attitudes. This difference was multiplied by -1
for those with initially negative attitudes, so that a positive score on polarization indicated
greater extremity in the direction of the original attitude, and a negative score indicated
from t1 to t2 for all three issues: Ukraine, M = -6.31, SD = 15.95, refugees, M = -1.39, SD =
Participants indicated, regarding politics in general, “how willing you are to…” engage in
EXPOSURE AND SHARING EFFECTS 22
different behaviors, 1= not at all willing to 7 = very willing. Three items (adapted from
Kruikemeier, van Noort, Vliegenthart, & de Vreese, 2014) measured online participation (“Sign
a political petition online,” “Send an e-mail to someone to influence him or her politics,” and
“Have a discussion about politics online”), α = .671, M = 2.83, SD = 1.31, and five items
(adapted from Shah et al., 2007) measured offline participation (“Sign a political petition,”
“Have a face-to-face discussion about politics,” and “Place a poster or sticker or wear a button
with a political message”), α = .807, M = 3.20, SD = 1.33. Online and offline participation were
Participants were asked “If the Ukraine referendum were held today, would you go
vote?” A majority of 53.9% said yes, 45.8% said no, and only one participant declined to answer.
3.4.7. Involvement
Items were adopted from the Cho and Boster (2005) measure of attitudinal involvement,
and worded with regard to political attitudes. Six items measured value involvement (e.g., “My
life would change if my political views were different”), α = .778, M = 3.88, SD = 1.18, and five
items measured impression involvement (“People could judge me on the basis of my political
Other variables included in the dataset but not included in these analyses are Facebook
use, news use, public opinion perceptions, perceived knowledge, affect, political interest,
avoidance.
Given the use of multiple topics, and the mix of within-subjects and between-subjects
effects predicted, as well as the focus on three dependent variables, the data were analyzed in a
series of tests. Each test was conducted separately for each topic, owing to different patterns of
individuals holding neutral attitudes (e.g., an individual with a pro- or anti-refugee attitude might
First, the within-subjects effects of pro- versus counter-attitudinal exposure (H1) and
sharing (H4) were tested with paired-sample t-tests. Then, the effects of exposure and sharing
were modeled as predictors in regression models for the outcomes of online participation, offline
participation, and opinion polarization. These regression models test the between-subjects main
effects outlined in H2, H3, H5, and H6. Next, value involvement and impression involvement
were tested as moderators of these main effects. Within-subjects ANCOVA tested whether value
(H7) and impression (H12) involvement moderated pro-attitudinal versus counter-attitudinal (a)
exposure and (b) sharing. Interaction effects were probed with the MEMORE macro (Montoya &
Hayes, 2017). Then, to test the two involvement variables as moderators of the effects of
exposure and sharing on the dependent variables (H8-11 and H13-16), regression models
PROCESS macro (Hayes, 2018) was used to probe interactions with the Johnson-Neyman
technique. Finally, Research Questions 1, 2, and 3 are addressed by comparing the relative
4. Results
To test the basic confirmation bias hypotheses of H1 and H4, paired-sample t-tests
compared pro-attitudinal and counter-attitudinal exposure and sharing for each topic. Topics
were analyzed separately owing to the exclusion of attitude-neutrals. For selective exposure,
there was a significant effect for the refugee topic alone, t(326) = 3.52, p < .001, d = .195, where
more seconds were spent reading pro-attitudinal (M = 33.89, SD = 65.24) than counter-
attitudinal (M = 16.75, SD = 63.99) articles. No effect was evident for exposure to the
referendum (p = .730, d = .022) or pay equality (p = .584, d = .041) topics. H1 was partially
supported. For selective sharing, all three topics showed pro-attitudinal effects: the referendum,
t(231) = 4.26, p < .001, d = .280, refugees, t(310) = 8.92, p < .001, d = .506, and pay equality,
t(337) = 3.48, p < .001, d = .189. H4 was fully supported, and RQ1 was answered: selective
sharing was stronger and more consistently observed than selective exposure.
EXPOSURE AND SHARING EFFECTS 25
Table 1
Relationships of Selective Exposure and Sharing with Online and Offline Participation
Involvement Interactions
Value*Pro Exposure .649* .121 .344 .487 .238 .413*
Value*Counter Expo. .417 -.337 .450 .163 -.143 -.048
Value*Pro Sharing -.332 -.203 .040 -.262 .013 -.167
Value*Counter Sharing -.112 .332 .313 -.390 .247 .288
Model 3: Impression ∆R2=.027 ∆R2=.002 ∆R2=.005 ∆R2=.018 ∆R2=.005 ∆R2=.004
Involvement Interactions
Impress*Pro Exposure .095 -.130 .059 .135 -.196 .037
Impress*Counter Expo. .429 -.039 .270 .656* .176 -.269
Impress*Pro Sharing -.492 .179 .260 -.269 -.001 .312
Impress*Counter Shar. .964* -.048 -.019 .222 -.118 -.286
N= 231 310 336 231 310 336
Note. Standardized coefficients. UA Ref = Ukraine-EU association referendum. *p < .05, **p <
.01, ***p < .001. Model 2 and Model 3 each build on the main effects in Model 1; Model 2
tested all value-involvement interactions simultaneously, and Model 3 tested all impression-
involvement interactions simultaneously.
EXPOSURE AND SHARING EFFECTS 26
Next, regression models for each topic (pro-attitudinal and counter-attitudinal exposure
and sharing as predictors) tested effects on online and offline participation as well as opinion
polarization, controlling for value and impression involvement (main effect models at the top of
Tables 1 and 2). Intention to share the pro-attitudinal article was associated with more online
participation intentions and offline participation intentions for the Ukraine referendum and
refugee topics (Table 1, columns 1, 2, 4 and 5). Sharing pro-attitudinal articles about equal pay
was positively associated with offline participation (Table 1, column 6). For the refugee and pay
equality topics, pro-attitudinal sharing was positively linked to polarization, and attitude-
Table 2
From these main effect findings outlined above, H5a-c were largely supported: the
sharing of pro-attitudinal social media news stories was positively associated with offline and
online political participation, as well as opinion polarization. By contrast, H2a-c, which predicted
polarization, were not supported at all. These discrepancies in findings answer RQ2, which
questioned whether sharing of pro-attitudinal social media news stories will be a more
consistently positive predictor of political participation and polarization than exposure to pro-
attitudinal social media news stories, which was indeed the case. H3a-c, which specified
relationships between the amount of time spent reading counter-attitudinal news stories and both
political participation and polarization, proved untenable. However, H6c, which hypothesized
that sharing counter-attitudinal social media news stories will be negatively associated with
opinion polarization was supported for the equal pay and refugee issues. Based on this finding,
RQ3c was answered, such that the sharing of counter-attitudinal news stories was a stronger
predictor of issue depolarization than exposure to counter-attitudinal news stories (except in the
In addition, sharing counter-attitudinal articles about pay equality was positively linked to
online participation (Table 1, column 3). Contrary to H6a, this finding provides some evidence
that counter-attitudinal information sharing was positively rather than negatively associated with
online political participation. This result also addresses RQ3a: Counter-attitudinal sharing was
not a strong negative predictor (in the case of pay equality was even a positive predictor) relative
attitudinal sharing and offline political participation across all three issues, lending no support to
H6b. This also addresses RQ3b regarding the relative contribution of exposure and sharing.
EXPOSURE AND SHARING EFFECTS 29
An additional regression model (not in table) tested whether exposure to and sharing of
Ukraine articles were associated with the intention to vote in the referendum (an outcome
specific to this topic). Only pro-attitudinal sharing was a (positive) predictor, β = .174, p = .035.
sharing for stories on the referendum, F(1, 229) = 7.74, p = .006, ηp2 = .033, refugees, F(1, 308)
= 12.33, p < .001, ηp2 = .038, and equal pay, F(1, 334) = 3.91, p = .049, ηp2 = .012. Probing these
interactions with Johnson-Neyman technique via the MEMORE macro (Montoya & Hayes,
2017) found that moderate and high levels of value involvement (≥ 3.16 for Ukraine, ≥ 2.25 for
refugees, and ≥ 3.16 for pay equality, where M = 3.88) were linked to more pro-attitudinal over
counter-attitudinal message sharing, whereas those low on value involvement show no difference
.080), and equal pay (p = .169) articles. H7a was not supported.
Impression involvement did not moderate selective sharing, (ps > .25), lending no
support to H12b. With regard to H12a, neither did impression involvement moderate the bias in
Then, regression models were extended to test value involvement and impression
attitudinal sharing, and counter-attitudinal sharing) on the DVs of interest (Model 2 in Tables 1
EXPOSURE AND SHARING EFFECTS 30
and 2). Alternatively, impression involvement was tested as a simultaneous moderator of focal
variables in Model 3 (bottom of Tables 1 and 2). A number of significant interactions emerged,
as indicated in the Tables. The PROCESS macro (Hayes, 2018) was used to probe and interpret
There was partial support for H8a-b which hypothesized that pro-attitudinal information
exposure will have positive effects on offline and online participation at high levels of value
involvement. People high on value involvement (the 13.79% at or above a score of 5.306) had a
positive effect of Ukraine pro-attitudinal exposure on online participation. Also, people high on
value involvement (the 14.54% at or above a score of 5.012) had a positive effect of equal pay
value involvement (the 1.72% at or below a score of 1.824) there was a negative effect of
involvement (the 7.42% at or below a score of 2.003) had a negative effect of equal pay pro-
People high on value involvement (the 39.57% at or above a score of 4.058) had a
with moderate to high on value involvement (the 89.52% at or above a score of 2.486) had a
negative effect of counter-attitudinal sharing on their polarization for the issue of equal pay,
where higher involvement led to stronger depolarization effects. In all, these findings run counter
to H11c, which hypothesized that the negative effect of counter-attitudinal information sharing
on opinion polarization is weakest among individuals with high levels of value involvement.
above a score of 4.374) had a positive effect of Ukraine counter-attitudinal exposure on offline
participation. Next, people low on impression involvement (the 6.90% at or below a score of
2.062) had a negative effect of Ukraine counter-attitudinal sharing on online participation, while
people high on impression involvement (the 21.12% at or above a score of 4.328) had a positive
effect of Ukraine counter-attitudinal sharing on online participation. This finding ran counter to
H16a, which hypothesized that the negative effects of counter-attitudinal information sharing on
online political participation will be stronger among people high on impression involvement.
In addition, value and impression involvement did not moderate any effects of exposure
None of the other moderation effect hypotheses or research questions were supported
with statistically significant results. With regard to H8c, H9c, and H10c, which hypothesized that
the respective effects of information exposure and pro-attitudinal information sharing on opinion
polarization would depend on value involvement, there was no support. Furthermore, regarding
H13c, H14c, H15c, and H16c, impression involvement did not moderate the effects of either
Also, the effects of pro-attitudinal information sharing on the two types of political
participation did not depend on value involvement. Additionally, moderation analyses showed
that the effects of counter-attitudinal exposure and sharing on both types of political participation
did not depend on value involvement. H9a-b, H10a-b, and H11a-b received no support. With
respect to H13a-b, the relationship between pro-attitudinal information exposure and both types
of political participation also did not depend on impression involvement. The relationship
between counter-attitudinal information exposure and online political participation also did not
EXPOSURE AND SHARING EFFECTS 32
both types of political participation did not depend on impression involvement, lending no
political participation did not depend on impression involvement, lending no support to H16b.
In short, H1 was supported (partial, by topic), H4 was supported, H5 was supported, H6c
was supported (partial, by topic), H7b was supported, and H8a-b was supported (partial, by
topic). Participants had confirmation bias in their exposure and sharing. Sharing stories affected
5. Discussion
In sum, this study’s findings are multi-faceted and complex. In terms of main effects,
results provide partial support for the hypothesis that people will spend more time reading pro-
attitudinal than counter-attitudinal social media news articles. The presence of a confirmation
bias was contingent on the issue: the refugee issue was the only topic to produce more pro-
attitudinal than counter-attitudinal exposure, and it also produced greater selective sharing than
the other two topics. The topic of refugees was the most polarized (at baseline) of the issues
examined in this study, which likely contributes to this difference with the other two issues.
Furthermore, pro- and counter-attitudinal information exposure did not have any main effects on
In contrast, findings showed support for the hypothesis that people will be more likely to
pro-attitudinal information sharing was positively associated with online and offline political
participation for issues pertaining to the Ukrainian referendum and refugees. However, there was
EXPOSURE AND SHARING EFFECTS 33
the issue of equal pay. Consistent with predictions, the sharing of pro-attitudinal news articles
about refugees and equal pay had positive effects on opinion polarization. Also, the sharing of
counter-attitudinal news articles about refugees and equal pay were negatively associated with
opinion polarization.
information sharing and political participation across issue types provide fairly strong support for
the argument that pro-attitudinal information sharing is a much stronger sign of one’s
commitment to their political views than pro-attitudinal partisan-based information exposure (cf.
Shin & Thorson, 2017). When people overtly express their identification with specific political
values or opinions in the form of pro-attitudinal information sharing, such information sharing is
likely to translate into political action. However, it is important to note that the link between
they control for baseline attitudes. The effects of sharing on polarization are absent for the
Ukraine referendum, which was the least salient and least contentious issue of the three. Given
the confusion or apathy among the Dutch electorate regarding the vote (cf. Van der Brug, van der
Meer, & van der Pas, 2018), sharing information about it on social media would involve less of
With regard to interaction effects, there was some modest evidence that the positive
online and offline political participation tended to be strongest among those with high levels of
EXPOSURE AND SHARING EFFECTS 34
value involvement for the issues of Ukraine referendum and equal pay. This set of findings
provides support for the assumption that people tend to be more politically defensive at high
levels of value involvement. When people are politically closed-minded, they are likely to
action. However, these findings only provide one part of the overall picture: Although not
hypothesized, there were also negative associations between pro-attitudinal news exposure and
political participation when value involvement levels were low for the issues of the Ukraine
referendum and equal pay. Perhaps, at low levels of value involvement, people are more
amenable to allowing opposing political viewpoints to influence their political decisions. And
when people are willing to consider alternative political views, they could experience attitude
ambivalence and develop weakened political attitudes that result in lower levels of political
believe that an active majority agrees with them and that their own political participation is not
depended on value involvement, for the issues of Ukrainian referendum and equal pay. However,
these findings ran counter to the study’s hypotheses. Specifically, the negative influence of
counter-attitudinal news sharing on opinion polarization was strongest when value involvement
in the issue of the Ukrainian referendum was high. Likewise, the negative effect of counter-
attitudinal news sharing on opinion polarization regarding equal pay was strongest when value
involvement was high. This finding is surprising because people tend to be closed to alternate
viewpoints at high levels of value involvement. It would appear, however, that those high on
value involvement were especially prone to persuasion (i.e., depolarization) if they engaged
EXPOSURE AND SHARING EFFECTS 35
relationships that information exposure and sharing have with political participation and
association between counter-attitudinal news sharing on online participation among those with
low levels of impression involvement on the issue of the Ukrainian referendum. This finding
provides qualified evidence for the idea that people with low levels of impression involvement
are unlikely to be strongly committed toward publicly performing actions that align with their
partisan values. When impression involvement levels are low, people seem alright with explicitly
letting their social media followers know that they are considering alternative political stances.
Also, there was a positive association between counter-attitudinal news sharing and online
participation among those with high levels of impression involvement on the referendum issue.
Perhaps, people with high levels of impression involvement openly share counter-attitudinal
news with the goal of openly deriding such information. Such denigration of counter-attitudinal
news in front of an imagined social media audience allows them to explicitly prove that they are
committed to their partisan values, strengthening their resolve to take political action.
5.1. Implications
In terms of theoretical contributions, this study builds upon existing research on political
information consumption and self-expression patterns on social media in the following ways.
First, this study parses out the respective effects of selective exposure and selective sharing on
political outcomes. We demonstrate that people are much more likely to engage in selective
sharing than selective exposure on social media and that selective sharing on social media is a
much more robust predictor of political outcomes than selective exposure. This underscores the
EXPOSURE AND SHARING EFFECTS 36
importance of delineating between exposure and sharing when examining how selectivity in
selective exposure such as the SESAM (Knobloch-Westerwick, 2015) would therefore benefit
from distinguishing between exposure, sharing, and other modes of content consumption and
consumption and self-expression on social media can have serious political consequences (Beam
et al., 2018; Lane et al., 2019). This study is thus one of few that examines the effects of both
selective exposure and sharing on different types of political outcomes, specifically participation
and opinion polarization. Our consistent result that sharing is more consequential than exposure
is consistent with theories of public commitment such as identity shift (Lane et al., 2019). People
might harbor polarized issue attitudes internally, but such attitudes are not synonymous with
explicit political action (e.g., Miller & Conover, 2015). For instance, our study shows that pro-
attitudinal information sharing was positively associated with political participation whereas
counter-attitudinal sharing was negatively associated with polarized issue attitudes. This pattern
of findings clearly illustrates that political participation and opinion polarization are conceptually
distinct outcomes. More theoretical development is needed to articulate different processes for
Third, it seems intuitively obvious that highly involved individuals will have very
selective political information consumption and expressive behaviors (Cho & Boster, 2005). Yet
few studies on political social media use have actually tested involvement as a moderator of
selectivity and its effects. Our study addressed this research gap by examining whether effects of
selective sharing and exposure on political outcomes depended on involvement levels. Further
theoretical refinement could articulate conditions under which involvement is consequential, and
EXPOSURE AND SHARING EFFECTS 37
future research should include accuracy involvement (Hart et al., 2009; Johnson & Eagly, 1989).
Fourth, many experiments examining online-based selective exposure (for reviews see
Clay et al., 2013 and Knobloch-Westerwick, 2015) have been conducted in the context of news
websites. However, little experimental work has employed social media contexts (but see Kaiser,
Keller, & Kleinen-von Königslöw, in press; Mothes & Ohme, 2019), even though large
majorities of individuals have become reliant on Facebook and other social media for news
rather than online news sites (Bakshy et al., 2015; Settle, 2018). As such, this study improves
upon previous study designs by using a free choice paradigm to examine partisan selective
This study’s findings have practical implications for harnessing social media platforms
for political engagement. Given that this study’s findings largely show that pro-attitudinal news
politicians can encourage their voter base to share political messages that favor their in-party
through their online social networks. Such selective sharing not only helps to keep other people
politically informed, but also makes people themselves more likely to take tangible political
action. Furthermore, messages tailored to match peoples’ partisan leanings and levels of value
involvement can be disseminated through online social networks to foster political engagement.
Based on this study’s findings, campaign planners can identify voters from their own camp with
high levels of value involvement on social media and increase political participation levels
among these individuals by encouraging them to share pro-attitudinal news stories. Furthermore,
research has shown that Facebook news use can lead to depolarization (Beam et al., 2018).
Similarly, our study’s findings show that sharing counter-attitudinal political information in a
Facebook context leads to less polarized issue attitudes, regardless of value involvement levels.
EXPOSURE AND SHARING EFFECTS 38
With this in mind, social media platforms can devise strategies to promote counter-attitudinal
information sharing, e.g., by designing advertisements to encourage people from all types of
value involvement levels to like, share, or comment on counter-attitudinal (or perhaps balanced,
cf. Brenes Peralta et al., 2017) news. Ultimately, such counter-attitudinal news sharing on
Facebook is likely to be beneficial to democracy by causing even people with higher levels of
value involvement to hold less extreme attitudes and embrace alternative issue viewpoints.
In terms of study limitations, this study did not measure political participation levels prior
to stimuli exposure. This prevents us from making causal claims about how selective exposure
and selective sharing affect political participation. Nevertheless, this study still provides
consistent evidence for the association between selective sharing and political participation.
Furthermore, this study operationalized selective sharing by using items gauging the likelihood
of commenting on the news article or “liking” the news article. On the one hand, these items can
be regarded as generalized indicators of expressive social media political use (e.g., Macafee &
De Simone, 2012). Future research can explore differences in the affordances of social media
sharing (e.g., liking vs. sharing). Moreover, measuring actual sharing (i.e., participants’ sharing
behavior in their own networks rather than sharing intentions in a questionnaire), perhaps in an
unobtrusive manner, can address limitations associated with self-reported intentions. Also, this
study did not examine mechanisms related to confirmation bias such as cognitive dissonance or
motivated reasoning. Future research examining partisan selectivity on social media should
incorporate such mechanisms into their study design, to better explain the links between
selectivity and its effects on political outcomes. A further limitation is the focus on three specific
topics (although varied in salience and baseline polarization) and variation that might exist in
EXPOSURE AND SHARING EFFECTS 39
stimuli (e.g., in argument strength). Future work should continue to include more variation in
topics and message design. Nevertheless, this study provides strong evidence of selective
party system, and makes a particularly unique contribution by juxtaposing exposure and sharing.
With regard to directions for research, future studies can examine how social media cues
such as the number of “likes” or “reactions” moderate the effects of selective exposure and
selective sharing on political outcomes. Furthermore, traditional public opinion theories such as
the spiral of silence posit that people will assess opinion congruence by paying attention to cues
in their social environments. If people perceive their opinions to be in the minority, they will
important to examine whether the tenets of such theories receive support in the context of online
social environments (Liu et al., 2017). Future studies can thus examine the respective effects of
selective exposure and selective sharing on other outcome variables such as political efficacy or
specific opinion expressions in social media. In addition, cross-national comparisons can help to
validate study findings and examine differences across political systems and cultures. Future
research should examine how the present predictions and findings will fare in other countries.
5.3. Conclusion
In sum, this study builds upon existing research on selective exposure and sharing in the
following ways: First, this study uses a free-choice experimental design to provide much stronger
evidence for the effects of information stance on selective sharing than on selective exposure.
Second, selective sharing is a much more consistent predictor of political outcomes than
selective exposure. Third, the effects of selective sharing and exposure on political outcomes
depend more on value involvement than impression involvement. Fourth, issues matter. By using
EXPOSURE AND SHARING EFFECTS 40
a variety of topics that varied in their preexisting salience and polarization, we introduced
variation that may explain the differences seen in confirmation bias and its effects. Future
research should even more systematically vary issue characteristics to test how they shape
selectivity effects. As social media platforms continue to play an important role in the political
process, more research needs to be conducted to understand how individual level predispositions
and socio-environmental factors can exacerbate or attenuate the effects of partisan selective
References
Adamic, L. A., & Glance, N. (2005). The political blogosphere and the 2004 US election:
Divided they blog. In J. Adibi, M. Grobelnik, D. Mladenic, & P. Pantel (Eds.), LinkKKD
’05: Proceedings of the 3rd International Workshop on Link Discovery (pp. 36-43). New
Alhabash, S., Almutairi, N., Lou, C., & Kim, W. (2019). Pathways to virality:
Barbera, P., Jost, J. T., Nagler, J., Tucker, J. A., & Bonneau, R. (2015). Tweeting from left to
Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and
Beam, M. A., Hutchens, M. J., & Hmielowski, J. D. (2016). Clicking vs. sharing: The
Beam, M. A., Hutchens, M. J., & Hmielowski, J. D. (2018). Facebook news and (de)polarization:
Reinforcing spirals in the 2016 US election. Information, Communication, & Society, 21,
940-958. doi:10.1080/1369118X.2018.1444783
Bode, L., Vraga, E. K., Borah, P., & Shah, D. V. (2014). A new space for political behavior:
Bos, L., Kruikemeier, S., & de Vreese, C. (2016). Nation binding: How public service
EXPOSURE AND SHARING EFFECTS 42
doi:10.1371/journal.pone.0155112
Boulianne, S. (2015). Social media use and participation: A meta-analysis of current research.
doi:10.1080/1369118X.2015.1008542
Brenes Peralta, C., Wojcieszak, M., Lelkes, Y., & de Vreese, C. (2017). Selective exposure to
balanced content and evidence type: The case of issue and non-issue publics about
climate change and health care. Journalism & Mass Communication Quarterly, 94, 833-
861. doi:10.1177/1077699016654681
Brown, W. J., & Basil, M. D. (1995). Media celebrities and public health: Responses to “Magic”
Johnson’s HIV disclosure and its impact on AIDS risk and high risk behaviors. Health
Cho, H., & Boster, F. J. (2005). Development and validation of value-, outcome-, and
doi:10.1177/0093650204273764
Choi, J., & Lee, J. K. (2015). Investigating the effects of news sharing and political interest on
doi:10.1016/j.chb.2014.11.029
Clay, R., Barber, J. M., & Shook, N. J. (2013). Techniques for measuring selective exposure: A
doi:10.1080/19312458.2013.813925
Colleoni, E., Rozza, A., & Arvidsson, A. (2014). Echo chamber or public sphere? Predicting
political orientation and measuring political homophily in Twitter using Big Data.
EXPOSURE AND SHARING EFFECTS 43
Coppini, D., Duncan, M. A., McLeod, D. M., Wise, D. A., Bialik, K. E., & Wu, Y. (2017). When
Fazio, R. H., Zanna, M. P., & Cooper, J. (1977). Dissonance and self-perception: An integrative
Feezell, J. T. (2016). Predicting online political participation: The importance of selection bias
and selective exposure in the online setting. Political Research Quarterly, 69, 495-509.
doi:10.1177/1065912916652503
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA, USA: Stanford University
Press.
Garrett, R. K. (2009a). Echo chambers online?: Politically motivated selective exposure among
doi:10.1111/j.1083-6101.2009.01440.x
2466.2009.01452.x
Garrett, R. K. (2013). Selective exposure: New methods and new directions. Communication
Garrett, R. K., Carnahan, D., & Lynch, E. K. (2013). A turn toward avoidance? Selective
EXPOSURE AND SHARING EFFECTS 44
doi:10.1007/s11109-011-9185-6
Garrett, R. K., Dvir Gvirsman, S., Johnson, B. K., Tsfati, Y., Neo, R., & Dal, A. (2014).
Gil de Zúñiga, H., Molyneux, L., & Zheng, P. (2014). Social media, political expression, and
Hameleers, M., Bos, L., & de Vreese, C. H. (2018). Selective exposure to populist
Hart, W., Albarracín, D., Eagly, A. H., Brechan, I., Lindberg, M. J., & Merrill, L. (2009). Feeling
regression-based approach (2nd ed.). New York, NY, USA: Guilford Press.
doi:10.1177/0963721414540168
Hwang, H., Kim, Y., & Huh, C. U. (2014). Seeing is believing: Effects of uncivil online debate
Jeong, M., Zo, H., Lee, C. H., & Ceran, Y. (2019). Feeling displeasure from online social media
EXPOSURE AND SHARING EFFECTS 45
postings: A study using cognitive dissonance theory. Computers in Human Behavior, 97,
231-240. doi:10.1016/j.chb.2019.02.021
Kaiser, J., Keller, T. R., & Kleinen-von Königslöw, K. (in press). Incidental news exposure on
Facebook as a social experience: The influence of recommender and media cues on news
Kim, Y. (2015). Does disagreement mitigate polarization? How selective exposure and
Kim, Y., & Chen, H.-T. (2016). Social media and online political participation: The mediating
Kim, Y. M. (2009). Issue publics in the new information environment: Selectivity, domain
doi:10.1177/0093650208330253
doi:10.1177/0093650211400597
Kruikemeier, S., van Noort, G., Vliegenthart, R., & de Vreese, C. H. (2014). Unraveling the
effects of active and passive forms of political Internet use: Does it affect citizens’
doi:10.1177/1461444813495163
EXPOSURE AND SHARING EFFECTS 46
Kümpel, A. S., Karnowski, V., & Keyling, T. (2015). News sharing in social media: A review of
current research on news sharing users, content, and networks. Social Media + Society,
Kwak, N. (1999). Revisiting the knowledge gap hypothesis: Education, motivation, and media
Kwak, N., Williams, A. E., Wang, X., & Lee, H. (2005). Talking politics and engaging politics:
doi:10.1177/0093650204271400
Hastall, M. R., & Knobloch-Westerwick, S. (2013). Caught in the act: Measuring selective
105. doi:10.1080/19312458.2012.761190
Huddy, L., Mason, L., & Aarøe, L. (2015). Expressive partisanship: Campaign involvement,
political emotion, and partisan identity. American Political Science Review, 109, 1-17.
doi:10.1017/S0003055414000604
Knobloch-Westerwick, S., & Johnson, B. K. (2014). Selective exposure for better or worse: Its
mediating role for online news’ impact on political participation. Journal of Computer-
Lane, D. S., Lee, S. S., Liang, F., Kim, D. H., Shen, L., Weeks, B. E., & Kwak, N. (2019). Social
media expression and the political self. Journal of Communication, 69, 49-72.
doi:10.1093/joc/jqy064
EXPOSURE AND SHARING EFFECTS 47
Lee, C., Shin, & Hong, A. (2018). Does social media use really make people political polarized?
Direct and indirect effects of social media use on political polarization in South Korea.
Lee, F. L. F., Chen, H.-T., & Chan, M. (2017). Social media use and university students’
Lee, J. K., Choi, J., Kim, C., & Kim, Y. (2014). Social media, network heterogeneity, and
Lee, S. K., Lindsey, N. J., & Kim, K. S. (2017). The effects of news consumption via social
media and news information overload on perceptions of journalistic norms and practices.
Leippe, M. R., & Elkin, R. A. (1987). When motives clash: Issue involvement and response
Liao, Q. V., & Fu, W.-T. (2013). Beyond the filter bubble: Interactive effects of perceived threat
Brewster, & S. Bødker (Eds.), CHI ’13: Proceedings of the SIGCHI conference on
human factors in computing systems (pp. 2359-2368). New York, NY, USA: ACM.
doi:10.1145/2470654.2481326
Liu, Y., Rui, J. R., & Cui, X. (2017). Are people willing to share their political opinions on
Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The
EXPOSURE AND SHARING EFFECTS 48
Lu, Y., Heatherly, K. A., & Lee, J. K. (2016). Cross-cutting exposure on social networking sites:
Macafee, T., & De Simone, J. J. (2012). Killing the bill online? Pathways to young people's
Matthes, J. (2012). Exposure to counterattitudinal news coverage and the timing of voting
Miller, P. R., & Conover, P. J. (2015). Red and blue states of mind: Partisan hostility and voting
doi:10.1177/1065912915577208
Montoya, A. K., & Hayes, A. F. (2017). Two condition within-participant statistical mediation
doi:10.1037/met0000086
Mothes, C., & Ohme, J. (2019). Partisan selective exposure in times of political and
42-53. doi:10.17645/mac.v7i3.2183
Oeldorf-Hirsch, A., & Sundar, S. S. (2015). Posting, commenting, and tagging: Effects of
doi:10.1016/j.chb.2014.11.024
Pellikaan, H., de Lange, S. L., & van der Meer, T. W. G. (2018). The centre does not hold:
Coalition politics and party system change in the Netherlands, 2002-2012. Government
Perloff, R. M. (1989). Ego-involvement and the third person effect of televised news
Settle, J. E. (2018). Frenemies: How social media polarizes America. Cambridge, UK:
Shah, D. V., Cho, J., Nah, S., Gotlieb, M. R., Hwang, H., Lee, N. J., Scholl, R. M., & McLeod,
doi:10.1111/j.1460-2466.2007.00363.x
Sherif, C. W., Sherif, M., & Nebergall, R. E. (1965). Attitude and attitude change: The social
Shin, J., & Thorson, K. (2017). Partisan selective sharing: The biased diffusion of fact-checking
doi:10.1111/jcom.12284
Slater, M. D., & Rouner, D. L. (1992). Confidence in beliefs about social groups as an outcome
of message exposure and its role in belief change persistence. Communication Research,
Smith, S. M., Fabrigar, L. R., Powell, D. M., & Estrada, M.-J. (2007). The role of information-
Stroud, N. J. (2008). Media use and political predispositions: Revisiting the concept of selective
Trilling, D., van Klingeren, M., & Tsfati, Y. (2016). Selective exposure, political polarization,
and possible mediators: Evidence from the Netherlands. International Journal of Public
Vaccari, C., Valeriani, A., Barberá, P., Bonneau, R., Jost, J. T., Nagler, J., . . . Tucker, J. A.
(2015). Political expression and action on social media: Exploring the relationship
between lower- and higher-threshold political activities among twitter users in Italy.
Valenzuela, S. (2013). Unpacking the use of social media for protest behavior: The roles of
information, opinion expression, and activism. American Behavioral Scientist, 57, 920-
942. doi:10.1177/0002764213479375
Van der Brug, W., van der Meer, T., & van der Pas, D. (2018). Voting in the Dutch “Ukraine-
0107-z
EXPOSURE AND SHARING EFFECTS 51
Weeks, B. E., Lane, D. S., Kim, D. H., Lee, S. S., & Kwak, N. (2017). Incidental exposure,
selective exposure, and political information sharing: Integrating online exposure patterns
379. doi:10.1111/jcc4.12199
selective exposure to political online information: Source bias versus content bias.
Wheeless, L. R. (1974). The effects of attitude, credibility, and homophily on selective exposure
Wojcieszak, M., Bimber, B., Feldman, L., & Stroud, N. J. (2016). Partisan news and political
doi:10.1080/10584609.2015.1051608
Research Highlights
• An experiment showed Dutch adults were more selective in sharing than exposure.