ChapterPDF Available

Pussyfooting around November? A Longitudinal Analysis of Politicians' Twitter Use in 2014


Abstract and Figures

In the November 2014 Midterm Elections, social media was used by more members of Congress and with greater frequency than ever before. We employ an established metric for interpreting the short but influential posts made by members of Congress via Twitter to determine how they position themselves relative to other politicians, candidates, and issues. This is crucial as we invoke a traditional theory of a two-party system-the median voter model-to explain why positioning via Twitter fluctuates in the period surrounding Election Day. Based on more than 338,000 Twitter postings by 422 politicians, our analysis shows that the extent to which a member of Congress positions one's self is a function of how far ahead (or behind) Election Day is. Our findings are robust to local polynomial trend line-and fixed effects-based analyses. We also show that Republicans are much more volatile in their use of positioning, and that-for both parties-the minimum point at which members of Congress reduce their positioning is not Election Day itself but approximately two weeks before Election Day.
Content may be subject to copyright.
Title: Pussyfooting around November? A Longitudinal Analysis of Politicians’ Twitter Use in
Matthew A. Shapiro (corresponding author)
Illinois Institute of Technology
Libby Hemphill
Illinois Institute of Technology
Jahna Otterbacher
Open University Cyprus
Abstract: In the November 2014 Midterm Elections, social media was used by more members of
Congress and with greater frequency than ever before. We employ an established metric for
interpreting the short but influential posts made by members of Congress via Twitter to
determine how they position themselves relative to other politicians, candidates, and issues. This
is crucial as we invoke a traditional theory of a two-party system – the median voter model – to
explain why positioning via Twitter fluctuates in the period surrounding Election Day. Based on
more than 338,000 Twitter postings by 422 politicians, our analysis shows that the extent to
which a member of Congress positions one’s self is a function of how far ahead (or behind)
Election Day is. Our findings are robust to local polynomial trend line- and fixed effects-based
analyses. We also show that Republicans are much more volatile in their use of positioning, and
that – for both parties – the minimum point at which members of Congress reduce their
positioning is not Election Day itself but approximately two weeks before Election Day.
Keywords: electoral politics, information technology and politics; political communication;
median voter model; social media; Twitter
“… [E]each party strives to make its platform as much like the other’s as possible…. Each
candidate ‘pussyfoots,’ replies ambiguously to questions, refuses to take a definite stand in any
controversy for fear of losing votes” (Hotelling 1929: 54).
Elected officials and those running for political office typically achieve personal political
success (re)election to office – as a function of how effectively they communicate their
campaign pledges or receive partisan support. These messages can be distorted in the traditional
media by targeting partisan audiences (Bernhardt et al. 2008; Caillaud and Tirole 2007) or by
highlighting incumbents’ latest accomplishments (Healy and Lenz 2014). All of this is
exacerbated by problems identified in the public’s “issue-attention cycle” (Downs 1972). In short,
misinformation from the media polarizes, confuses, and is compounded by existing biases
(Kuklinski et al. 2000; Lewandowsky et al. 2012; Nyhan and Reifler 2010). We hold constant
but acknowledge these media-related effects as we focus on similar distortions arising through
social media. Specifically, we identify crucial fluctuations in the Twitter-based communications
by members of Congress during the 2014 Midterm Election.
We acknowledge that social media-based communications from elected officials are but
a part of a larger information-sharing campaign. In this way, we are effectively building on the
body of research which focuses on the language officials use in traditional media (Cook et al.
1983; Edwards III and Wood 1999; Entman 2007; Kedrowski 2000; Lee 2009) and on websites
that report statements and speeches of public officials (Gentzkow and Shapiro 2011). Social
media-based communication is important (Gulati and Williams 2007), as it is a proxy for
politicians’ ideology and policy preferences (Hemphill, Culotta, and Heston 2013; Shapiro and
Hemphill 2014) and predicts how a candidate will behave with (potential) constituents and
campaign funders (Williams and Gulati 2013). This paper, thus, is as much a study of how
members of the U.S. Congress altered their Twitter-based communications during the 2014
election“speech acts” in our vernacularas it is a campaign for broader acceptance of the
importance of social media-based communications by politicians.
In the following pages, we explain how we can understand elected officialsuse of social
media in traditional theories of electoral behavior. We then examine longitudinally, from
November 1, 2013 to February 26, 2015,1
how members of Congress altered their
communication patterns in line with Hotelling’s classic argument, captured in the prefacing
quotation. Our findings show that, regardless of party and chamber, members of Congress alter
their message particularly for general elections.
Tapping the Median Voter Model
Hotelling’s (1929) claims laid the foundation for a more developed understanding of how
a two-party system like that in the U.S. operates. Namely, election seekers present themselves as
moderates in order to maximize their vote-share (Bernstein et al. 1988; Canes-Wrone et al. 2002;
1 The latter date was selected to allow us to clean and analyze the data in a timely manner.
Downs 1957). It is a compelling and elegant argument despite the fact that the public is
ideologically divided and especially so among those who are most engaged in politics
(Abramowitz and Saunders 2008; Green et al. 2002; Greenberg 2004; Hetherington and Weiler
Democratic voters have shifted ideologically less to the left than Republican voters have
shifted to the right (Abramowitz and Saunders 2008; Abramowitz 2013; Abramowitz 2014;
Mann and Ornstein 2012; Brady and Han 2006), possibly a function of cross-party
communication differences. That is, while repetitious communications allow information to enter
the recipient’s working memory (Chong and Druckman 2007; Chong and Druckman 2011),
dramatic changes in the content of this information around Election Day may make it difficult
for voters to track shifts in the ideological position of their candidates (Lauderdale 2013). This
would be particularly problematic given alignments between social identity and party affiliation
(Bartels 2002; Campbell et al. 1960; Hetherington 2001; Mason 2013) as well as the role of party
branding (Groeling 2010; Brady et al. 2000; Jones 2010; Kawato 1987).
We make two propositions. First, the incentives for maximizing vote share attract both
parties to the median voter but especially around the election. We conceptualize this attraction in
the form of decreased rhetoric about one’s position about an issue; i.e. to invoke Hotelling
(1929), more “pussyfooting.” Second, given rightward shifts in the median voter, as implied
above, the decrease in partisan rhetoric around Election Day is expected to differ between parties.
In light of the connections between social identity and party affiliation, it is difficult to predict
how such differences might take shape. On the one hand, Republicans may be more intense in
their rhetoric overall, and thus the expected decrease around Election Day could result in a level
of rhetoric which still exceeds that of Democrats. On the other hand, Republicans may be more
dramatic in how they approach the median voter, decreasing their use of rhetoric closer to the
election and increasing it rapidly afterward. It is worth noting that these conditions are not
mutually exclusive.
There are several additional directions in which we can examine the phenomenon of
changes in communication from members of Congress over time. An obvious choice, for
example, is cross-chamber differences, particularly given claims that Senators are more even
tempered than Representatives (Grofman et al. 1991; Theriault 2006; Kernell 1973; Hare and
Poole 2014). Given differences in how the use of rhetoric is used from politicians to politician,
we focus here primarily on how variance in the frequency of communications affects inter-party
Data Collection & Coding
Twitter remains a less preferred communication vehicle for the majority of American
voters. Yet, there were approximately 7 million followers of members of Congress on Twitter as
2 See for
of the fall of 2012,3
The collection of Twitter posts was a function of the Twitter Database Server (Green
2011), existing Twitter-collectors (Hemphill 2011), and new Twitter-collectors such as
PurpleTag which use the Twitter REST API to collect a maximum of 3,200 Twitter posts from
each legislator.
and recent research has identified a positive association between Twitter
posts by members of Congress and New York Times content (Shapiro and Hemphill 2014). We
assume, thus, that Twitter is used as a key part of politicians’ communication tools. For example,
the simple count of Twitter posts by members of Congress from November 1, 2013 to February
26, 2015 shows clear lulls around the winter break, spikes around the President’s State of the
Union address, and spikes surrounding other policy-related issues.
4 For the period from November 1, 2013 to February 26, 2015, we accumulated
nearly 338,000 Twitter postings by 422 politicians.5 Classifying positioning Twitter posts is the
result of an iterative process of establishing inter-coder reliability across a spectrum of action-
based categories and then using an automated algorithm to classify thousands more statements.
This approach addresses many if not all of the concerns raised in Grimmer and Stewart (2013)
about text analysis at such a massive scale. Six codes were ultimately identified: narrating,
positioning, directing to information, requesting action, giving thanks, and “other”, the latter
dropped in our subsequent analysis. Codes were not mutually exclusive, and Cohen’s kappa
scores (Cohen 1968) for each code indicate very strong agreement between coders. Given the
labor intensiveness of hand-coding each Twitter post, we automated the coding process by
training binary classifiers for each of the five action codes. We used MALLET (Machine
Learning for LanguagE Toolkit; McCallum 2002), which employs supervised learning
algorithms to exploit the words in Twitter posts in order to determine whether or not they exhibit
each of the five actions.6
The algorithm produces probabilistic results – each Twitter post has a
value between 0 and 1 for each action code, and the value indicates how likely that code applies
to the Twitter post. For instance, a value of 0.80 for Positioning means there’s an 80 percent
likelihood the Twitter post is actually a positioning statement. Typically, the differences between
high, medium, and low positioning Twitter posts are a function whether or not a bill, person,
party, or the Twitter post author’s opinion is mentioned.
Our analysis should begin with an examination of all five speech acts over the 2014
election cycle; however, we target only positioning for the purposes of this paper. We use local
polynomial trend lines (Fan and Gijbels 1996), fitting the trend line using locally weighted least
3 The number of followers is determined by aggregating the followers for each individual
member of Congress. There are, thus, likely to multiple counts of the same individual,
particularly members of the media, political pundits, fans of Congress, etc.
4 See for details.
5 Twitter’s API makes nearly all Twitter posts available, but occasionally Twitter posts are not
indexed and are therefore not made available programmatically.
6 Further details about the development of our coding scheme and the automated classifier are
available in Hemphill, Otterbacher, and Shapiro (2013) and Otterbacher et al. (2012).
squares. This is consistent with our expectation that more recent events matter more than those
across the entire time period. To definitively show that positioning is occurring in line with our
expectations, we engage in a second analysis which restricts the time period from August 2014 to
the end of February 2015 and employ fixed effects modeling techniques. The fixed effects
approach is especially important given the potential for unidentified attributes of individual
members of Congress to drive shifts in positioning.
Twitter is used by many but dominated by a few, shown in Table 1. Given the possibility
that a handful of users can determine the impact of our graphical results, we control for variance
across high- and low-frequency Twitter users in Congress by grouping members of Congress
into quartiles based on their Twitter usage from November 1, 2013 to February 25, 2015.7 Party-
based differences are presented in Figure 1, where each local polynomial trend line represents a
quartile (lowest frequency Twitter posters are in the first quartile; highest frequency Twitter
posters are in the third quartile).8
In this figure, we observe that both Democrats and Republicans
begin to decrease their positioning from late-July 2014 in anticipation of Election Day (the
vertical blue line), and that the greatest differences among quartiles centers on Election Day. We
also observe that both parties increase their positioning after Election Day. More similarities
between parties include the divergence of quartile-based trend lines from November 2013 up to
Election Day as well as the increased positioning which occurs from April 2014, peaking in mid-
summer 2014. We attribute this peak to politicians’ appeals during primary elections to the
median voter within their respective parties. In sum, our expectation that November 2014 will
coincide with the lowest point for positioning in order to attract the median voter is conditionally
satisfied. A close inspection of Figure 1 reveals, though, that positioning is least likely to occur
not on Election Day itself but rather approximately two weeks before.
In terms of party-based differences, Republicans exhibit much more dynamic shifts in
their use of positioning while using Twitter, possibly challenging voters’ understanding and/or
ability to track Republican ideology. The crucial cross-party distinction lies in our comparison of
quartiles of frequency. For Democrats, members of Congress that are most likely to position are
the least frequent users of Twitter. Yet, the opposite is true for Republicans. More importantly,
while we continue to see for Democrats a clear shift from decreased-to-increased positioning in
the two weeks preceding Election Day, a close examination of the first and second Republican
quartiles indicates that there are small shifts back and forth within the last two weeks of the
election. Specifically, Republicans in the first quartile increase their positioning before
remaining static in the last few days before the election; Republicans in the second quartile shift
between slight increases and decreases in the last couple of weeks, but then increase their
7 These quartiles are not based on distributions within party and/or chamber.
8 All parties other than Republicans are grouped with Democrats.
positioning in the last few days before the election. Despite these differences, the overall pattern
is one in which there is a negative correlation between frequency of Twitter posting and strength
of positioning for Democrats. The correlation is positive for Republicans.
Complementing the previous analysis are fixed effects results which predict over time the
positioning by party and chamber subgroup. Party-level effects can be superseded by those of
individual politicians (Jacobson 2009; Jacobson 1978; Stokes and Miller 1962), so a fixed effects
modeling approach focusing on each individual member of Congress allows us to control for and
guard against the possibility of omitted time invariant measures that also might be correlated
with positioning. The marginal effects, i.e. the predicted changes relative to November 2014, are
presented in Table 2. Statistically significant differences between each month and November
2014 are represented by bold font. We realize that November 2014 represents the entire month
and not solely the election on November 4 and, thus, that the coefficients in marginal effects are
at best representations of general time trends. Nonetheless, the results remained unchanged when
using two week bands of time for the period preceding and following the 2014 election.
The dynamics revealed in Table 2 are not entirely consistent with the expectations set
forth in Hypothesis 1. Rather than moderating language incrementally more in the months
leading up to the election, members of Congress tend to fluctuate in their positioning. Relative to
November 2014, House Democrats, for example, position less in September and October only.
Senate Democrats, House Republicans, and Senate Republicans all position less in August and
October only. Finally, for all four subgroups, positioning increases in the months following the
election, implying that the period surrounding Election Day is indeed a minimum point.
Researchers have not reached consensus regarding how the Internet impacts the political
campaign process (Druckman et al. 2014), but the evidence provided here confirms that
communications via social media are consistent with and contribute to the corpus of electoral
behavior and media-based communication. Our research confirms that the November 2014
election – possibly every general election from this point forth – can be used to predict how
members of Congress will use rhetoric via Twitter. Higher levels of positioning by members of
Congress between election cycles and lower levels of extreme language use in the period leading
up to the election have the potential to draw in the median voter and dictate what is reported in
the traditional media. These findings are robust to both local polynomial trend line- and fixed
effects-based analyses. Yet, there are a number of alternative explanations for our findings, and
we make a broad call for future and sustained research on political behavior and communication
through social media.
In terms of inducing theory about political communication, we observed that members of
Congress represent themselves and/or their colleagues in very different ways in the final days
before the election. We do observe conflicting evidence with regard to our expectation that
members of Congress will be more restrained in their use of positioning. For both parties, the
final two weeks of an election is a period during which politicians seem to assert themselves by
positioning more. This may occur because a member of Congress believes that more extreme
positioning in the final days before the election will produce an appearance of confidence about
the election’s outcome. Yet, roughly one quarter of all voters do not finalize their vote choice
earlier than two weeks before the election (Nir and Druckman 2008), a share that has continued
to grow in recent years (Box-Steffensmeier et al. n.d.). Perhaps politicians expect that voters
have essentially made up their minds about their vote choice, and thus the final two weeks
provide an opportunity to further secure the votes of the one’s existing base of support. Whatever
the case – and to again repeat the call for future research on this topic – the late-campaign shift
represents but another piece of Twitter-derived evidence not previously identified for members
of Congress en masse or longitudinally.
Figures and Tables
Table 1. Most frequent Twitter posters by congressional chamber and political party, November
1, 2013 to February 26, 2015
House Democrats
Senate Democrats
House Republicans
Senate Republicans
Chaka Fattah 3,228
Sanders 3,240
Boehner 3,244
McConnell 2,796
Keith Ellison
Patty Murray
John McCain
Kyrsten Sinema 3,190
Blumental 2,961 Bradley Byrne 2,814 Rand Paul 1,889
Wilson 3,161
Murphy 2,867 Billy Long 2,702 Ted Cruz 1,603
Paul Tonko 3,118
Baldwin 2,568
Pittenger 2,311 Daniel Coats 1,498
Tony Cardenas 2,532
Heitkamp 2,557 Keith Rothfus 2,022 Roy Blunt 1,457
Langevin 2,299
Schumer 2,134 Renee Ellmers 1,975 Orrin Hatch 1,419
Dina Titus 2,214
Cardin 1,934 Steve Scalise 1,841 Mark Kirkfe 1,407
Eric Swalwell
Richard Durbin
James Inhofe
Garamendi 2,150 Thomas Carper 1,292
Fitzpatrick 1,631 Michael Crapo 1,297
Figure 1. Local polynomial probabilities of positioning, 11/1/2013 to 2/26/2015, by party
Note: Vertical line indicates Election Day, 2014. Quartiles represent groups of members of
Congress according to frequency of Twitter posts.
.3 .35 .4 .45
Democrat Republican
first quartile
second quartile
third quartile
Positioning probability
Table 2. Probabilities (as percentages) of positioning by party and chamber in the months
surrounding November 2014
August 2014
September 2014
October 2014
November 2014
December 2014
January 2015
February 2015
Note: Marginal measurements based on fixed effects modeling; bold text indicates significant
differences at the p<0.1 level relative to probability of positioning in November 2014.
Abramowitz, Alan I. “Long-term trends and short-term forecasts: the transformation of US
presidential elections in an age of polarization.” PS: Political Science & Politics 47.2
(2014): 289–292.
———. “The emerging Democratic presidential majority: lessons of Obama’s victory.” In
American Political Science Association 2013 Annual Meeting, 2013.
Abramowitz, Alan I., and Kyle L. Saunders. “Is Polarization a Myth?” The Journal of Politics
70.2 (2008): 542–555.
Bartels, Larry M. “Beyond the running tally: partisan bias in political perceptions.” Political
Behavior 24.2 (2002): 117–150.
Bernhardt, Dan, Stefan Krasa, and Mattias Polborn. “Political Polarization and the Electoral
Effects of Media Bias.” Journal of Public Economics 92.5-6 (2008): 1092–1104.
Bernstein, Robert, Gerald C. Wright, and Michael Berkman. “Do U.S. senators moderate
strategically?” American Political Science Review 82.1 (1988): 237–245.
Box-Steffensmeier, Janet et al. “The long and short of it: The unpredictability of late deciding
voters.” Electoral Studies (n.d.).
Brady, David W., Robert D’Onofrio, and Morris P. Fiorina. “The nationalization of electoral
forces revisited.” In Continuity and Change in House Elections, edited by David W. Brady,
John Cogan, and Morris P. Fiorina. Palo Alto: Stanford University Press, 2000.
Brady, David W., and Hahrie C. Han. “Polarization then and now: A historical perspective.” In
Red And Blue Nation?: Characteristics And Causes of America’s Polarized Politics, 119–
174. Washginton, D.C.: Brookings and Hoover Press, 2006.
Caillaud, Bernard, and Jean Tirole. “Consensus Building: How to Persuade a Group.” American
Economic Review 97.5 (2007): 1877–1900.
Campbell, Angus et al. The American voter. vols. New York: Wiley, 1960.
Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. “Out of step, out of office:
electoral accountability and House members’ voting.” American Political Science Review
96.1 (2002): 127–140.
Chong, Dennis, and James N. Druckman. “Framing public opinion in competitive democracies.”
American Political Science Review 101.4 (2007): 637–655.
———. “Public-elite interactions: Puzzles in search of researchers.” In The Oxford Handbook of
American Public Opinion and the Media, edited by Robert Y. Shapiro, Lawrence R. Jacobs,
and George C. Edwards. Oxford: Oxford University Press, 2011.
Cohen, Jacob. “Weighted kappa: Nominal scale agreement provision for scaled disagreement or
partial credit.” Psychological Bulletin 70.4 (1968): 213–220. Available:
Cook, Fay Lomax et al. “Media and agenda setting: Effects of the public, interest group leaders,
policy makers, and policy.” Public Opinion Quarterly 47 (1983): 16–35.
Downs, Anthony. An Economic Theory of Democracy. vols. New York: Harper Collins, 1957.
———. “Up and down with ecology - the ‘issue-attention cycle.’” The Public Interest
28.Summer (1972): 38–50.
Druckman, James N., Martin J. Kifer, and Michael Parkin. “U.S. congressional campaign
communications in an Internet age.” Journal of Elections, Public Opinion & Parties 24.1
(2014): 20–44.
Edwards III, George C, and B Dan Wood. “Who influences whom? The president, congress, and
the media.” American Political Science Review 93.2 (1999): 327–344.
Entman, Robert M. “Framing bias: Media in the distribution of power.” Journal of
Communication 57.1 (2007): 163–173.
Fan, J., and I. Gijbels. Local Polynomial Modelling and Its Applications. vols. London: Chapman
& Hall, 1996.
Gentzkow, Matthew, and J.M. Shapiro. “Ideological segregation online and offline.” Quarterly
Journal of Economics 126.4 (2011): 1799–1839.
Green, A. “Twitter database server (Beta 0.10),” 2011. Available:
Green, D., B. Palmquist, and E. Schickler. Parisan Hearts and Mind: Political Parties and the
Social Identities of Voters. vols. New Haven: Yale University Press, 2002.
Greenberg, Stanley B. The Two Americas. vols. New York: St. Martin’s Press, 2004.
Grimmer, Justin, and Brandon M. Stewart. “Text as data: the promise and pitfalls of automatic
content analysis methods for political texts.” Political Analysis 21.3 (2013): 267–297.
Groeling, Tim. When Politicians Attack: Party cohesion in the media. vols. New York:
Cambridge University Press, 2010.
Grofman, Bernard, Robert Griffin, and Amihai Glazer. “Is the Senate more liberal than the
House? Another look.” Legislative Studies Quarterly 16.2 (1991): 281–295.
Gulati, Girish J., and Christine B. Williams. “Closing the gap, raising the bar: candidate web site
communication in the 2006 campaigns for Congress.” Social Science Computer Review
25.4 (2007): 443–465.
Hare, Christopher, and Keith T. Poole. “The polarization of contemporary American politics.”
Poilty 46 (2014): 411–429.
Healy, Andrew J., and Gabriel S. Lenz. “Substituting the end for the whole: why voters respond
primarily to the election-year economy.” American Journal of Political Science 58.1 (2014):
Hemphill, L. “Twitter-collectors (144e2f0f6),” 2011. Available:
Hemphill, Libby, Aron Culotta, and Matthew Heston. Framing in social media: how the U.S.
Congress uses Twitter hashtags to frame political issues. vols., 2013.
Hemphill, Libby, Jahna Otterbacher, and Matthew A. Shapiro. “What’s Congress doing on
Twitter?” In Proceedings of the 2013 Conference on Computer Supported Cooperative
Work, 877–886, 2013.
Hetherington, Marc J. “Resurgent mass partisanship: the role of elite polarization.” American
Political Science Review 85.3 (2001): 619–631.
Hetherington, Marc J., and Jonathan D. Weiler. Authoritarianism and polarization in American
politics. vols. Cambridge: Cambridge University Press, 2009.
Hotelling, Harold. “Stability in competition.” The Economic Journal 39.153 (1929): 41–57.
Jacobson, Gary C. “The effects of campaign spending in congressional elections.” American
Political Science Review 72.2 (1978): 469–491.
———. The Politics of Congressional Elections. vols. 7th ed. New York: Pearson Longman,
Jones, David. “Partisan polarization and congressional accountability in House elections.”
American Political Science Review 54.2 (2010): 323–337.
Kawato, Sadafumi. “Nationalization and partisan realignment in congressional elections.”
American Political Science Review 81.4 (1987): 1235–1250.
Kedrowski, Karen M. “How members of congress use the media to influence public policy.” In
Media Power in Politics, edited by Doris A Graber. Washington, D.C.: Congressional
Quarterly, Inc., 2000.
Kernell, Sam. “Is the Senate more liberal than the House?” The Journal of Politics 35.2 (1973):
Kuklinski, James H. et al. “Misinformation and the currency of democratic citizenship.” The
Journal of Politics 62.3 (2000): 790–816.
Lauderdale, Benjamin E. “Does inattention to political debate explain the polarization gap
between the U.S. Congress and public?” Public Opinion Quarterly 77.S1 (2013): 2–23.
Lee, Han Soo. “News media conditions presidential responsiveness to the public.” In APSA
Annual Conference. Toronto, 2009.
Lewandowsky, Stephan et al. “Misinformation and its correction: continued influence and
successful debiasing.” Psychological Science in the Public Interest 13.3 (2012): 106–131.
Mann, Thomas E., and Norman J. Ornstein. It’s Even Worse Than It Looks: How the American
constitutional system collided with the new politics of extremism. vols. New York: Basic
Books, 2012.
Mason, Lilliana. “The rise of uncivil agreement: issue versus behavioral polarization in the
American electorate.” American Behavioral Scientist 57.1 (2013): 140–159.
McCallum, Andrew Kachites. “MALLET: A Machine Learning for Language Toolkit,” 2002.
Nir, Lilach, and James N. Druckman. “Campaign mixed-message flows and timing of vote
decision.” International Journal of Public Opinion Research 20.3 (2008): 326–346.
Nyhan, Brendan, and Jason Reifler. “When corrections fail: the persistence of political
misperceptions.” Political Behavior 32.2 (2010): 303–330.
Otterbacher, Jahna, Libby Hemphill, and Matthew A. Shapiro. “Tweeting vertically? Elected
officials’ interactions with citizens on Twitter.” In CeDEM (Conference for E-Democracy
and Open Government) Asia 2012, 2012.
Shapiro, Matthew A., and Libby Hemphill. “Policy-related communications and agenda setting:
Twitter, New Yorks Times, and the widening soapbox.” Available at SSRN:, 2014.
Stokes, Donald E., and Warren E. Miller. “Party government and the saliency of Congress.”
Public Opinion Quarterly 26.4 (1962): 531–546.
Theriault, Sean M. “Party polarization in the U.S. Congres: member replacement and member
adaptation.” Party Politics 12.4 (2006): 483–503.
Williams, Christine B., and Girish J. Gulati. “Social networks in political campaigns: Facebook
and the congressional elections of 2006 and 2008.” New Media & Society 15.1 (2013): 52–
This paper examines partisan communications of incumbent members of Congress during the nine weeks leading up to the 2016 U.S. election. The central premise is rooted in the median voter theorem, which is coupled with theories of political activation and reinforcement, to show how politicians communicate in order to attract support from large swaths of the public. We analyze the partisanship of tweets posted by incumbents in Congress using mixed-effects models to examine the relationships between party, time, and race competitiveness on the degree of partisanship expressed by politicians. Our results reveal that Democrats and Republicans exhibited different partisanship signaling patterns in the weeks before the election. Specifically, Democrats decreased their partisanship, perhaps to appeal to the median voter, while Republicans stayed consistent in their partisanship, potentially using Twitter to activate and reinforce voters rather than to win them over.
Full-text available
Influencing the policy agenda has long been viewed as one of the most important sources of political I power. For decades, scholars have maintained that the president has the most significant role in setting the policymaking agenda in Washington, but little systematic empirical work has been done to measure the president's influence. We explore the president's success in focusing the issue attention of Congress and the mass media by evaluating time-series measures of presidential, mass media, and congressional attention to five issues: crime, education, health care, U.S.-Soviet relations, and the Arab-Israeli conflict. We find that most of the time the president reacts, responding primarily, to fluctuations in media attention and world events. in domestic policy, we find a more interactive relationship, one that appears to offer the president the opportunity to act in an entrepreneurial fashion to focus the attention of others ill the system on major presidential initiatives.
Fostering a positive brand name is the chief benefit parties provide for their members. They do this both by coordinating their activities in the legislative process and by communicating with voters. Whereas political scientists have generally focused on the former, dismissing partisan communication as cheap talk, this book argues that a party's ability to coordinate its communication has important implications for the study of politics. The macro-level institutional setting of a party's communication heavily influences that party's prospects for cohesive communication. Paradoxically, unified government presents the greatest challenge to unified communication within the president's party. As this book argues, the challenge stems primarily from two sources: the constitutional separation of powers and the intervening role of the news media.
Agenda setting efforts by our elected officials have been studied thus far under the assumptions that symbols and the like propagate an understanding of the problem, and that this happens through the traditional media (e.g., newspapers, television news). The rise of social media, and Twitter in particular, enhances opportunities for agenda setters to share information with others. Indeed, the volume of communications from our elected officials in Congress is enormous – hundreds of thousands of Twitter postings each year – and there has been scant research on what all this social media activity means for the policy making process. We examine Twitter data for all of 2013, focusing on areas relating to the environment, health care, the economy, among others. We match up these communications to those presented in the traditional media in order to gauge how these Congress-based Twitter posts and New York Times articles co-occur. There is overlap, and it is striking. The patterns for discussions about the budget and the Affordable Care Act are nearly identical for both information sources, but patterns are less correlated for issues relating to immigration, the environment, and energy. While eschewing causal claims, we do believe that this is mounting evidence that members of Congress are attempting to use social media to impact the public agenda. Competing with media outlets, politicians seek first mover status in identifying a problem.
In the 2006 midterm elections, even more campaigns and interest groups had an online presence than in 200?, and their activities had matured relative to previous years (Rainie and Horrigan 2007). Moreover, citizens seeking information about U.S. Senate races increased fivefold over the 2002 midterm election level and doubled for U.S. House races. Although mainstream media continued to dominate the content that citizens viewed online, 20 percent reported going directly to a candidate's Web site to learn about the campaign (Rainie and Horrigan 2007). Television remained the medium of choice, but the Internet's financial role continued to enlarge. Estimates put the total for online fund-raising at $100 million and online campaign advertising at $?0 million (Cornfield and Rainie 2006).
Politics and political conflict often occur in the written and spoken word. Scholars have long recognized this, but the massive costs of analyzing even moderately sized collections of texts have hindered their use in political science research. Here lies the promise of automated text analysis: it substantially reduces the costs of analyzing large collections of text. We provide a guide to this exciting new area of research and show how, in many instances, the methods have already obtained part of their promise. But there are pitfalls to using automated methods—they are no substitute for careful thought and close reading and require extensive and problem-specific validation. We survey a wide range of new methods, provide guidance on how to validate the output of the models, and clarify misconceptions and errors in the literature. To conclude, we argue that for automated text methods to become a standard tool for political scientists, methodologists must contribute new methods and new methods of validation.
Recent studies of the U.S. Congress have demonstrated a substantial difference in partisan polarization between legislators' votes and citizens' survey responses about those votes. But perhaps public polarization would increase if citizens were more atten- tive to political debates in Congress? Using matching techniques on natural variation in citizens' political information levels, I show that citizens who are informed about the partisan alignment of issues have a similar preference distribution to Congress, even after the former are re-weighted to resemble the entire public along salient political, social, and demographic dimensions. In contrast, using a survey experiment, I show that cue and argument treatments only partially reduce the discrepancy between the views expressed by the public and the voting behavior of Congress on the same issues. Both experimental and observational studies have significant limitations for measuring counterfactuals involving public opinion, and so our understanding of the polarization gap remains unfortunately limited.