Running head SEXUALITY, MEDIA, AND ATTRACTION 1 SEXUALITY,.docx
Spreading the Message
1. Spreading the Message:
Predicting Party Hashtag Use by Members of Congress on Twitter
John Murchison ‘16, Joli Holmes ‘17, & Grace Wong ’18, and Logan Dancey, Assistant Professor.
John Murchison
Wesleyan University 2016
jmurchison@wesleyan.edu
Contact Casas, Andreu, and John Wilkerson. 2015. ”A Delicate Balance: Republican Party Branding during the 2013
Government Shutdown.” Working paper. http://andreucasas.com/casas_wilkerson_party_brand.pdf
Egan, Patrick. J. 2013. Partisan Priorities: How Issue Ownership Drives and Distorts American Politics. New York:
Cambridge University Press.
Grossmann, Matt, and David A. Hopkins. 2015. “Ideological Republicans and Group Interest Democrats: The
Asymmetry of American Party Politics.” Perspectives on Politics 13 (No. 1): 119-139
Petrocik, John R. 1996. ”Issue Ownership in Presidential Elections, with a 1980 Case Study.” American Journal of
Political Science 40 (no. 3): 825-850.
References
• In recent years, members of congress have increasingly relied on
Twitter as means of communicating messages to the public.
There is evidence that parties intentionally coordinate messaging
on Twitter (Casas and Wilkerson, 2015). The advent of hashtags
allows us to track message adoption and volume.
• Drawing on theories of party behavior and strategy, we test two
expecations about party messaging. First, parties will message on
issues they "own" (Petrocik 1996; Egan 2014). Second, individual
messaging incentives will vary across parties with Republicans
driven more by ideological considerations and Democrats driven
more by group-based considerations (Grossman and Hopkins
2015).
Introduction
• All tweets posted by members of the 114th Congress from January
to December 2015 were collected from Twitter using the
Streaming API.
• The tweet dataset contains over 270,000 tweets from 240
House Republicans and 187 House Democrats
• 52.4 % of the tweets came from Republicans and 47.6% of
the tweets came from Democrats.
• Other variables, both district-level and member-level, were
collected from GovTrack, Pew Research, and the U.S. Census.
• The analysis in this paper focuses on a subset of tweets: those
using one of the top 100 hashtags in our dataset in 2015.
• To test the issue ownership expectation, we coded hashtags into
categories used by Egan (2014). Of the top 100 hashtags, over 50
could be categorized (e.g., #Obamacare coded as health,
#irandeal as foreign policy, and #climatechange as environment).
• Figure 1 plots partisan usage differential against public
opinion advantage to assess the issue ownership expectation.
• To explore in differences across parties, we also identified a group
of hashtags that members of both parties used in their
messaginging: #immigration, #medicare, #jobs, ACA (#aca,
#obamacare, #getcovered, #kingvburwell), and Planned
Parenthood (#standwithpp, #pp, #defundpp).
• We fit a negative binomial model for various hashtags/categories
predicting usage count by ideological variables (caucus
membership), constituency variables, and total number of
tweets by the member. We compare constituency and
ideological effects by party.
• Constituency variables by hashtag/category: #immigration-
%Hispanic/Latino; #medicare-%Over 60; #jobs-%Unemployed;
#ACA-%Uninsured, Planned Parenthood-%Highly Religious;
Trade-%Manufacturing jobs.
Data & Methods
• Figure 2 presents the ideological variable results for the negative
binomial models predicting hashtag/category count. The y-axis
shows the marginal increase in predicted use (by count) of the
hashtag for members of the most extreme caucus compared to
the most moderate caucus (see Figure 1).
• Figure 3 shows the same marginal difference by hashtag, but
using constituency variables (see data and methods section)
instead of caucus membership. The y-axis shows the marginal
increase in predicted use with an increase of one standard
deviation in the constituency variable.
• We find evidence that both the ideological identification of the
member and more group-based constituency characteristics
predict hashtag usage across both parties. When messaging on
the same issue, Democrats and Republicans seem motivated by
similar considerations. On this admittedly small set of issues,
then, we see that the relative importance of ideological vs. more
group-based considerations varies across issues but not
systematically across parties.
• Figure 4 shows a positive but weak correlation between issue
ownership and hashtag usage.
• Hashtag usage is characterized by partisanship, with Republicans
and Democrats often using different hashtags. However, there is a
significant amount of overlap in what issues parties message on.
For example, both parties message on planned parenthood and
Obamacare, but they tend to use different hashtags (e.g.,
#defundpp vs. #standwithpp or #obamacare vs. #getcovered).
Parties are thus willing to message on both issues they own and
issues the other party owns.
Discussion
Data & Methods Results
Acknowledgements
Thanks to Pavel Oleinikov, Manolis Kaparakis, Wesleyan’s Quantitative Analysis Center, &
the Department of Government for supporting this project.
Joli Holmes
Wesleyan University 2017
jholmes@wesleyan.edu
Grace Wong
Wesleyan University 2018
ghwong@wesleyan.edu
Figure 2: Marginal Effect of Membership in Most Extreme Ideological Caucus, by Party and
Topic/Hashtag (reference group: most moderate caucus)
Results
Figure 1: Density of Ideological Positioning (NOMINATE) of Caucus Membership Figure 4: Public perception of owned issues compared to actual use.
Correlation = .194, p-value = .15
Figure 3: Marginal Effect of 1 Std. Deviation Increase in Relevant Constituency Variable, by
Party and Topic/Hashtag
Logan Dancey
Wesleyan University
ldancey@wesleyan.edu