Professional Documents
Culture Documents
Abstract
Politicians’ Twitter habits can vary considerably. Those who choose to use may do so as part of a broad discussion
community or as a one-to-many broadcaster. Because each user sees a different mix of tweets, a politician seen by
one user as interacting with the public may be seen by another as engaging in one-way communication, potentially
prompting different evaluations of the politician, the ongoing discussion, or even Twitter itself. This study uses an
experiment to test the effects of different engagement and framing styles in politician tweets on evaluations of the
politician, other discussants, and Twitter itself. Findings suggest that politicians who use Twitter to broadcast,
rather than engage with other users, not only receive worse evaluations themselves but that the negative evaluation
carries over to other users discussing the same topics, as well as to evaluations of the utility of Twitter as an
information source. This effect is attributed to intramedium interaction, in which reactions to one aspect of a
multimedia object carry over to other aspects of the same object.
1
College of Mass Communication & Media Arts, Southern Illinois University Carbondale, Carbondale, Illinois.
2
School of Journalism, Southern Illinois University Carbondale, Carbondale, Illinois.
8
POLITICIAN EVALUATIONS ON TWITTER 9
audiences develop with mediated personalities.14(p215) Like respond to strategic frames the way a typical person would
social presence, PSI is facilitated by direct address15 and fosters respond to value frames, for the same reasons. Discussion of
imagined intimacy and emotional closeness.16 Mediated per- strategy that explicitly favors one party may only have the effect
sonalities can achieve this intimacy using ‘‘the gestures, con- for one side.
versational style, and milieu of an informal face-to-face
gathering.’’14(p216) Lee and Shin7 argue that Twitter specifi-
Intramedium Interaction
cally may help foster PSI by displaying conversation threads
that have unfolded in real time, which positions readers as Twitter’s mixed-media setting might also mean individual
potential participants. Those reading a ‘‘conversational give- components of the screen influence perceptions of one an-
and-take between the candidate and other Twitter users might other. Holbert argued that as a variety of different media are
be able to vicariously participate in the virtual interaction.’’7(p3) consumed, they produce indirect, cumulative, and comple-
Studies have found that elected officials broadcast far mentary effects, a process called ‘‘intramedia mediation.’’35
more than they interact.17–19 While not pursuing interaction Other studies provide evidence of such contingent additive
may be rationale,20 direct address creates a more ‘‘real’’ effects of news consumption.36 Veenstra et al. focused this
experience of conversation for followers and can help voters concept on a single mixed-media environment encompassing
form a deeper more human picture of politicians. Perceived multiple elements, calling the process in this scenario ‘‘in-
presence has been shown to induce positive evaluations tramedium interaction.’’33
generally16,21,22 and, more specifically, improve overall In intramedium interaction, responses to one component
candidate evaluations and vote intention.6,7,23 prime reactions to components viewed later on the same
screen. For example, our research team previously found that
Social Identity Model of Deindividuation Effects responses to partisan attack ads on YouTube carried over to
evaluations of the comments section, despite comments’
However, party affiliation might interact with engage-
constancy.37 A similar priming process underlies the ‘‘nasty
ment.6 The social identity model of deindividuation effects
effect’’: incivility in comments sections can alter perceptions
(SIDE) suggests that personalization can weaken in-group
of constant news content.38 Likewise, participants’ responses
support by undercutting the salience of group identity.24
to varied levels of engagement, or to different frame use,
At the same time, it can reduce the impersonal out-group
may color evaluations of other users to whom the politician
prejudice of those in the opposing party. Conversely, an
does or does not respond.
unengaged communicator would produce a larger support
gap between in- and outgroups as those messages are more
likely to activate partisan heuristics.25 Hypotheses and Research Questions
Lee and Oh found that those with strong party affiliations
Based on the existing literature about politicians’ en-
responded more negatively to personalization in both in- and
gagement with the public and social media, as well as liter-
out-party Twitter messages.6 Lower ratings for a personal-
ature on response to value and strategic frames, we present
ized in-group candidate fit with the SIDE model, but those
the following hypotheses and research questions:
for the out-group figure were unexpected. Additionally, their
study found that among low identifiers, personalization in- H1: Evaluations of a Congressional representative will be
creased positive evaluation, but only of same-party candi- more positive when the representative (a) frames tweets in
dates. While interaction effects are likely, the literature lacks terms of values rather than strategy and (b) engages person-
a clear picture of which or why they are likely to occur. ally with other Twitter users.
H3: Evaluations of Twitter as a platform will be more faith political discussion.’’ These items were averaged to cre-
positive when the representative (a) frames the conversa- ate an index (r = 0.41, M = 3.21, SD = 0.73).
tion in terms of values rather than strategy and (b) engages
personally with other Twitter users.
Twitter platform evaluations. Evaluations of Twitter were
measured by agreement with 10 statements. Five items re-
RQ3: How, if at all, do (a) the representative’s tweet frame
lated to Twitter’s utility for discussion, such as ‘‘It provides a
and engagement style and (b) frame, engagement,
place for meaningful political discussion,’’ and ‘‘It’s a good
and party affiliation interact to influence evalua-
place to interact with public officials.’’ These items were av-
tions of Twitter?
eraged to create an index of positive assessments of Twitter as
a discussion platform (a = 0.81, M = 3.26, SD = 0.80). Another
five items related to Twitter’s utility as an information source,
Methods
such as ‘‘It’s a good source of news about politics.’’ These
Data for this study were collected using an online survey items were averaged to create an index of positive assessments
experiment with an embedded Twitter stimulus. Participants of Twitter as an information source (a = 0.86, M = 2.49,
were exposed to a series of manipulated tweets that were SD = 0.81). Full question wording and factor loading can be
based on real tweets sent by members of the Congress and found in Appendix 2.
the general Twitter public regarding the 2014 Farm Bill. This
topic was selected both because the public would likely have Party affiliation. Because Rep. Donald is a Democrat and
a range of familiarity with it and because it provided real makes a specific note of President Obama signing the Farm
tweets that fit well with the study’s manipulations. Nine Bill, Democratic Party affiliation is included as a factor in the
tweets were presented in a basic Twitter page template, analysis. One item asking for participants’ party affiliation was
which participants were told was a conversation captured collapsed to Democrats (40.2 percent) versus non-Democrats.
from Twitter, with some of the page elements removed or
disabled. The series began with a tweet from fictional Re- Results
presentative Mike Donald praising the passage of the Farm
Candidate evaluations
Bill, followed by responses from the public and further
tweets from Rep. Donald. A series of analysis of variance (ANOVA) tests were used to
Two factors were manipulated. First, the Representative’s address the hypotheses and research questions, each including
original tweet was framed in terms of either values (calling frame manipulation, engagement manipulation, and Demo-
the Farm Bill ‘‘a big win for farmers, consumers, commu- cratic affiliation as factors (for full ANOVA results, see Ta-
nities, and all of America’’) or electoral strategy (calling it ble 1). The first model tested influence on evaluations of Rep.
‘‘a big win for Democratic candidates across America in Donald. Main effects were found for engagement (F(1, 336)
2014’’). A manipulation check verified that participants had = 15.14, p < 0.001) and Democratic affiliation (F(1, 336) =
noticed the frame (t(344) = -4.10, p < 0.001). Another factor 4.92, p = 0.027) but not for frame (F(1, 336) = 1.37, p = 0.243).
manipulated engagement. In one condition, the politician Engagement prompted a more positive evaluation than did
used @-replies and second-person pronouns to respond di- nonengagement (3.70, SD = 0.58 vs. 3.42, SD = 0.69), sup-
rectly to tweets and included #FarmBill in his tweets. In the porting H1b but not H1a. Additionally, frame and Democratic
other, he made substantively equal tweets, but did not use affiliation interacted (F(1, 336) = 4.05, p = 0.045), with non-
others’ @names or hashtags, and linked to press releases on Democrats exposed to the strategic frame giving significantly
house.gov. Other users’ tweets were held constant. The full- lower evaluations (M = 3.37, SD = 0.69) than did all other
page template and manipulated text can be seen in Appendix 1. groups (non-Democrat–value frame, M = 3.60, SD = 0.63;
Manipulations were fully crossed for a 2 · 2 (value–strategy · Democrat–value, M = 3.61, SD = 0.68; and Democrat–strategy,
engaged–unengaged) experimental design.a M = 3.67, SD = 0.57). No other significant interactions were
Participants were recruited through mturk.com. Partici- found.
pants were paid $0.50, providing a total number of 344.
Limited to American residents, the sample was 29.8 percent Evaluations of other users
female and 78.1 percent white, with a mean age of 30.37
Next, we modeled evaluation of other users as the depen-
years (SD = 9.42), a median education of a Bachelor’s, and a
dent variable. Engagement again significantly influenced the
median household income of $40,000–60,000. The study
outcome (F(1, 336) = 13.74, p < 0.001), with a higher mean
was fielded on March 19, 2014.
than nonengagement (3.36, SD = 0.69 vs. 3.05, SD = 0.76),
supporting H2b. However, no other significant main or in-
Variable construction
teractive effects were found.
Evaluations of candidate and other Twitter users. Several
statements were presented on a five-point agree (5)–disagree Evaluations of Twitter
(1) scale to assess feelings toward Rep. Donald and the other
H3 was tested in two ANOVAs, with evaluations of Twitter
Twitter discussants. Five items focused on Rep. Donald, such
as a discussion platform and as an information source as the
as ‘‘He is interested in his constituents’ concerns.’’ These items
dependent variables. There were no significant main or inter-
were averaged to create an index of positive assessments of
active effects on evaluations of Twitter as a discussion plat-
Rep. Donald (Cronbach’s a = 0.78, M = 3.56, SD = 0.64). Two
form. However, evaluations of Twitter as an information
items measured views of the other users: ‘‘They are open to
source were subject to a main effect of Democratic affiliation
other viewpoints,’’ and ‘‘They use Twitter to engage in good
(F(1, 334) = 5.44, p = 0.020), as well as a three-way interaction
POLITICIAN EVALUATIONS ON TWITTER 11
tendency moderates its effects. Communication Research 29. Ball-Rokeach SJ, Loges WE. (1996) Making choices: me-
2013; 40:27–51. dia roles in the construction of value-choices. In: Seligman
17. Graham T, Broersma M, Hazelhoff K, et al. Between C, Olson JM, Zanna MP, eds. The psychology of values: the
broadcasting political messages and interacting with voters: Ontario Symposium. Mahwah, NJ: Erlbaum.
the use of Twitter during the 2010 UK General Election 30. Jamieson KH. (1992) Dirty politics: deception, distraction,
campaign. Information, Communication & Society 2013; and democracy. New York, NY: Oxford University Press.
16:692–716. 31. Lawrence RG. Game-framing the issues: tracking the
18. Graham T, Jackson D, Broersma M. New platform, old strategy frame in public policy news. Political Commu-
habits? Candidates’ use of twitter during the 2010 British nication 2000; 17:93–114.
and Dutch General Election campaigns. New Media & 32. Patterson TE. (1994) Out of order. New York, NY: Vintage
Society 2014; DOI: 10.1177/1461444814546728. Books.
19. Waters RD, Williams JM. Squawking, tweeting, cooing, 33. Veenstra AS, Sayre B, Shah DV, et al. Frames and
and hooting: analyzing the communication patterns of knowledge in mixed media: how activation changes in-
government agencies on Twitter. Journal of Public Affairs formation intake. CyberPsychology & Behavior 2008;
2011; 11:353–363. 11:443–450.
20. Stromer-Galley J. On-line interaction and why candidates 34. Scholl RM, Pingree RJ, Gotlieb MR, et al. Here’s what
avoid it. Journal of Communication 2000; 50:111–132. you’ll learn from this news story: prior framing and
21. Bente G, Rüggenberg S, Krämer NC, et al. Avatar-mediated learning reasons from news. Electronic News (in press).
networking: increasing social presence and interpersonal trust 35. Holbert RL. Intramedia mediation: the cumulative and
in net-based collaborations. Human Communication Research complementary effects of news media use. Political Com-
2008; 34:287–318. munication 2005; 22:447–461.
22. Skalski P, Tamborini R. The role of social presence in in- 36. Shen F, Eveland WP. Testing the intramedia interaction
teractive agent-based persuasion. Media Psychology 2007; hypothesis: the contingent effects of news. Journal of
10:385–413. Communication 2010; 60:364–387.
23. Thorson KS, Rodgers S. Relationships between blogs as 37. Veenstra AS, Park CS, Lyons BA, Kang CS, Iyer N. In-
eWOM and interactivity, perceived interactivity, and para- tramedium interaction and the third-person effect: how
social interaction. Journal of Interactive Advertising 2006; partisans respond to youtube ads and comments. Cyberp-
6:34–44. sychology, Behavior, and Social Networking 2015; 18:
24. Spears R, Postmes T, Lea M, et al. When are net effects 406–410.
gross products? The power of influence and the influence of 38. Anderson AA, Brossard D, Scheufele DA, et al. The ‘‘nasty
power in computer-mediated communication. Journal of effect’’: online incivility and risk perceptions of emerging
Social Issues 2002; 58:91–107. technologies. Journal of Computer-Mediated Communica-
25. Postmes T, Spears R. Behavior online: does anonymous tion 2014; 19:373–387.
computer communication reduce gender inequality? Person-
ality and Social Psychology Bulletin 2002; 28:1073–1083.
26. Hopmann DN, Van Aelst P, Legnante G. Political balance Address correspondence to:
in the news: a review of concepts, operationalizations and Benjamin A. Lyons
key findings. Journalism 2012; 13:240–257. College of Mass Communication & Media Arts
27. Schemer C, Wirth W, Matthes J. Value resonance and value Southern Illinois University Carbondale
framing effects on voting intentions in direct-democratic 1100 Lincoln Drive
campaigns. American Behavioral Scientist 2012; 56:334–352. Mailcode 6601
28. Shah DV, Domke D, Wackman, DB. ‘‘To thine own self be Carbondale, IL 62901
true’’: values, framing and voter decision-making strate-
gies. Communication Research 1996; 23:509–560. E-mail: lyonsb@siu.edu
(Appendix follows /)
14 LYONS AND VEENSTRA
Appendix 1. Stimulus
POLITICIAN EVALUATIONS ON TWITTER 15