Live-Tweeting a Presidential Primary Debate

Abstract

Twitter offers a function called live-tweeting that allows users to communicate about events with each other in real time. This study examines the use of live-tweeting during a 2012 Republican Primary Debate by examining the 181,780 tweets posted during the nationally televised debate. Live-tweeting offers users an opportunity to engage in public conversation about political events and thus potentially influence the framing of what occurred. Our study examines whether citizens utilize the opportunity to contribute to the political narrative or if elite users dominate political conversations through an analysis of tweets used by both groups. Findings show that there were very few differences between the elite and nonelite social media conversations and that elite users views were spread farther than nonelite views.

Keywords

social media presidential debates 2012 election public sphere framing

It is generally agreed that televised political debates have a positive function in a representative democracy. Generally speaking, political debates work to inform the electorate (Benoit, Hansen, & Verser, 2003) and can mobilize citizens to participate in politics (McKinney & Chattopadhyay, 2007). In the United States, all presidential general election and most presidential primary election debates are televised nationally and accompanied by extensive pundit analysis. Interestingly, Chaffee and Dennis (1979) suggest that this debate commentary may have a greater impact on the viewing audience than the debate itself.

Media coverage of political events is often accomplished by the use of specific frames. A frame is a lens applied to an issue or event that highlights certain aspects of the issue or event and downplays others (Entman, 2004). The way an issue or event is framed has been found to affect individuals’ knowledge, attitudes, and behaviors related to that issue or event (see Borah, 2011 for a review). Entman (2003, 2004) outlines a process, called cascading activation, through which a frame is created and passed from the media to the audience. Within the cascading activation process, a frame moves to the public as individuals see the frame in the media and adopt it into their understanding of the event (Entman, 2003).

However with the advent of social media platforms like Twitter, the potential exists for the public to create their own frames regarding political events and not simply rely on the frames created by media organizations. Twitter allows for users to annotate an event in real time and share those messages with others. Users can post a frame of an event on their chosen social media that may in turn influence the overall framing of the event. As Adam Sharp, the head of Twitter’s government news and social innovation division, explains, “Twitter is where people go to connect in real time to events and talk about them” (“Twindex,” 2012 at 2:53). Are citizens actually using social media to create their own political frames? If the answer is yes, the potential exists for a more robust political sphere that includes more diverse voices.

This project explores this question. Given the emerging capacity of social media for users to participate in large scale, public discussions of major political events, the study explores how the frames from tweets of media and political elites differ from the tweets of nonelite (or public) sources. Differences will be examined by analyzing the content and structure of the conversation that occurred on Twitter during the Republican Florida primary debate that occurred on January 26, 2012, and was broadcast nationally by Cable News Network (CNN). The goal is to see whether citizens used Twitter media to create unique frames in commentary posted during a presidential primary debate or if the frames contained in the body of tweets created by citizens generally corresponded with those of the political and media elite.

Social Media, Frames, and Political Debates

In a study of the political revolts that occurred in Iran after the 2009 elections, Rahimi (2011) conceptualized Twitter as a locality in which dissent could be discussed and expressed by the citizens toward the government. Ampofo, O’Loughlin, and Anstead (2011) analyzed the communication relationships in social media between citizens and political elites in the process of opinion formation and found that people tended to adopt different roles to help teach people who were confused about the issues. Tumasjan, Sprenger, Sandner, and Welpe (2011) argued that Twitter was a locality in which political deliberation occurred during the 2010 German elections, finding that the frequency of mentions of political parties during the election campaign predicted the political party that eventually won the election. It may be the case that social media has the potential to expand opportunities for citizen discussion of political issues.

However, others argue that the promise of social media and citizen discussion has not been realized. For example, Larsson and Moe (2011) explored Twitter posts in the context of the Swedish 2010 federal election and found that Twitter use was largely constrained to a small group of individuals. Gaffney (2010) argued that reports of Twitter use and influence during the Iranian Revolts in 2009 were largely overblown. Morozov (2009, 2012) agreed, further arguing that it would be easy for governments and other powers to use the Internet and social media to control populations.

The current project extends this line of research by examining the talk that occurs during the live-tweeting of a presidential nomination debate. Some research suggests that media and candidate goals may be different from citizens’ goals when it comes to issues addressed in a presidential debate. Jackson-Beeck and Meadow (1979) conducted a content analysis of the 1960 and 1976 debates, comparing the issue agenda of the candidates and moderators with public polling data regarding the issues important to citizens. Ultimately, they discovered three distinct issue agendas among the three groups. This finding was replicated by McKinney (2005), who found that the issues mentioned in the 2004 town hall debate were different from those that were important to the public. The point is that if citizens and elites have differing issue agendas during a debate this should be reflected in conversation on Twitter.

To understand the online conversation during the debate, we focus specifically on how it was framed. A frame is the emphasis given to some aspects of an event and the lack of emphasis given to other aspects (Entman, 2004). This emphasis can affect the network of ideas and feelings associated with particular events, candidates, or issues that exists in the minds of the audience (Berkowitz, 1984; Borah, 2011). This is important because debate exposure has been shown to influence the evaluations that viewers have of candidate’s image and issue stances and therefore can affect an individual’s cognitive network (McKinney & Carlin, 2004). However, while some debate effects result from viewing the actual debate event, effects may occur as the result of exposure to news coverage of the event (Fridkin, Kenney, Gershon, Shafer, & Woodall, 2007; Hellweg, Pfau, & Brydon, 1992; Steeper, 1978; Tsfati, 2003).

Can we explicate frames from social media conversation during a political event? Researchers have begun to examine the frames developed through social media (e.g., Hamdy & Gomaa, 2012). In addition, news outlets are more frequently reporting on what people are saying over social media, emphasizing individual posts that have been picked from a group or reporting on the posts as a whole (see Rusack, 2012; The Times-Union, 2012, for examples of both methods).

For our analysis, we analyze the Twitter frames of entire groups. Using this method may limit our understanding of nuances contained within a particular frame, but it allows us to discover differences between the frames constructed by media/political elites and the public. Entman (2003, 2004) outlines two concepts that are important to the popularity of a frame: magnitude, which corresponds to the frequency of uses of framing words, and cultural resonance, which centers around the importance of individual terms and ideas and how they may dictate the nature of the frame. We utilize both of these concepts in our analysis.

Guided by our overall interest in the differences of Twitter dialogue between elite and nonelite users, we pose the first research question:

Research Question 1a: What amount of debate tweets come from media and political elite sources and nonelite sources?

In addition to regular tweets, Twitter allows users to “re-tweet” messages. This involves pushing a button next to the tweet, subsequent to which the tweet is posted to all the followers of the retweeter and can in turn be viewed on the live-tweeting stream again. This allows users to quickly and easily spread an original tweet from another source. In the case of a live-tweeting event, a tweet that is retweeted is more likely to be seen by more users. Because there are too many tweets occurring too quickly in a popular live-tweeting event (like a political debate) it is impossible for all users to see every tweet. So a tweet that is retweeted has a larger likelihood of being seen by more people; that likelihood grows the more it is retweeted again. This function corresponds to Entman’s (2003, 2004) concept of magnitude, which suggests that through repetition a frame is more likely to be adopted. One would expect that a retweet would have more of an effect on the larger frame that emerges than an individual tweet that is not retweeted. To determine differences between the two groups in this regard, the following research question is asked:

Research Question 1b: Are media and political elite sources or nonelite sources retweeted with greater frequency?

Mentions of candidates are an important aspect of debate coverage and would likely be an important factor in the Twitter conversation. A particular candidate as focus of a tweet suggests that they are being emphasized in the frame. Candidate mentions correspond to Entman’s (2003, 2004) concept of cultural resonance. More mentions of a candidate may make that candidate more salient in the minds of the audience. It is also important to examine how candidates are discussed, because evaluations of those candidates in the Twitter conversation are likely major components associated with the frames emerging from each group. This aspect of the debate coverage on Twitter is explored through the following research questions:

Research Question 2a: How frequently were each of the candidates mentioned in Twitter debate dialogue?

Research Question 2b: Were the candidates mentioned differently in tweets from media and political elite sources and nonelite sources?

Research Question 2c: What was the tone of tweets associated with candidate mentions from media and political elites sources and nonelite sources?

One final aspect of the twitter debate dialogue explored in this project is issue mentions. While users are tweeting about a debate, they may also be engaging in substantive discussion of issues. Thus, the Twitter conversation during the live event may indicate which issues are important to citizens, the media, and political elites (McKinney, Houston, & Hawthorne, 2012) and mentions of issues may in turn influence the salience of the issue to viewers of and participants in the live-tweeting event. This leads to our final set of research questions:

Research Question 3a: What issues and how frequently are these issues mentioned in Twitter debate dialogue?

Research Question 3b: Are there differences in the issues mentioned and the frequency of issue tweets authored by media and political elite and nonelite sources?

Method

The Florida Republican Primary Debate on January 26, 2012, at the University of Northern Florida was utilized to explore the research questions posed by this project. The debate was sponsored by CNN, which offered national coverage of the event. The small field of potential Republican presidential candidates appearing on stage included Mitt Romney, Newt Gingrich, Rick Santorum, and Ron Paul. The debate drew 5.4 million viewers (Seidman, 2012). Tweets from the January 26, 2012, CNN Florida Republican primary debate were collected using the DataSift web service, which provides access to the full stream of tweets and captures tweets posted using designated search terms during a specified time period. CNN publicized the hashtag “#CNNDebate” as the official debate hashtag for use by Twitter users both prior to the debate and during the 90-min broadcast itself. All tweets that contained that hashtag were collected for this analysis. The final database contained a total of 181,803 tweets from 42,476 individual user accounts, a relatively small subset of the 200 million active monthly Twitter users (Fiegerman, 2012). The unit of analysis was an individual tweet, and the database included (1) the twitter message, (2) the username of the poster, (3) whether the tweet was a retweet, and (4) the username of the original retweeted user. Only publicly accessible and viewable tweets were collected using this method.

It is important to note that this specific debate was the 25th in the 2012 Republican Primary Debate series. It occurred relatively late in the cycle and will likely be remembered as just another debate in a long series of debates. The media narrative regarding this debate series may have already been established by the time debate occurred, implying that preexisting frames could be affecting individual views, elite views, and frames emergent within the social media conversation.

To measure the status of users as media/political elites or nonelite users, we first determined which users were media and political elite users. This was accomplished by comparing the Twitter username associated with each captured tweet with a list of media and political elite Twitter accounts. The list of elite accounts was created by going to the main and subsidiary Twitter feeds of every national news network (MSNBC/NBC, FOX News, CNN, CBS, and ABC) and collecting the usernames of all accounts operated by individuals in the employ of those networks as indicated in the user biography of the account (e.g., several people indicated in their user biography that they were editors for CNN, so they were considered media elites) as well as any political figures that were followed by those accounts. These included all Republican presidential candidates as well as many members of Congress and other nationally prominent political office holders that have active Twitter accounts. The final list contained a total of 1,055 elite accounts.

Metadata for each tweet indicated if the tweet was a retweet or not. In addition, we calculated a metric that allowed us to compare the distribution of each group’s tweets within the conversation. By calculating the number of tweets that were retweeted per the number of tweets posted by each group, we can see which group’s tweets were propagated more by other users. The higher the number, the more often the group was retweeted per each tweet posted by the group. This metric was calculated by finding the number of tweets posted by the group that were retweeted and dividing it by the total number of tweets posted by each group.

Tone for each tweet was coded using the Affective Norms for English Language Words (ANEW), which is a list containing 2,476 words that have been assigned emotion and tone scores by human subjects (Bradley & Lang, 1999). Each word in the ANEW list is coded on three dimensions on a scale from 1 to 9. These dimensions include valence, indicating that the word is associated with happiness at the high end of the scale and sadness at the low end of the scale; arousal, indicating that the word is associated with excitement at the high end of the scale and with boredom at the low end of the scale; and dominance, indicating that the word is associated with feelings of being in control at the high end of the scale and with feelings of being submissive at the low end of the scale.

The system of coding each tweet for an emotional score is similar to the method employed by González-Bailón, Banchs, and Kaltenbrunner (2012). Tweets were searched for any words that were in the ANEW list and the emotional dimension scores of each word were saved along with the total frequency of ANEW words that occurred in each tweet. Mean and standard deviation of each emotional variable were calculated for each tweet which established a level of valence, arousal, and domination associated with each.

Candidate mentions were measured by coding each tweet for whether it included a candidate’s name (including uses of first, last, or both names). A single tweet could include more than one candidate’s name, and in that case the tweet was counted as mentioning multiple candidates.

Issue mentions were measured by examining each tweet for whether it included words associated with different issues. Based on work from Petrocik, Benoit, and Hansen (2003), a dictionary of issues and words associated with those issues was used to find specific issues mentioned in the Twitter conversation. This list was updated to include terms relevant to the 2012 political context and to remove concatenation from the word list (Cole & Hawthorne, 2013). To establish the number of issue mentions, the body of tweets for each group was searched for each issue word in the list. This produced a frequency count of the words associated with each issue mentioned by each group.

To analyze the data associated with differences in frequency or proportions between elites and nonelites, χ² tests were used, with the null hypothesis being that the frequency or proportions between groups should be equal. To analyze the differences between elites and nonelites on the continuous variables related to tone, independent sample t-tests were employed.

Results

Research Question 1 focused on the number of tweets posted by elites compared to nonelite users as well as what group was retweeted more. A chi-square difference test compared the frequency of tweets to determine differences between the two groups. Results can be seen in Table 1. There were stark differences in the frequencies of tweets and retweets in the two groups, with nonelite users posting and being retweeted significantly more than elite users. However, when the amount of retweets per tweet was examined, elite users were found to be responsible for more.

Table 1.

Frequency and Percentage of Tweets and Retweets and Retweet/Tweet Ratio of Both Groups.

	Total	Elites	Nonelites	χ² Difference Test
All tweets	181,803 (100%)	1,152 (0.63%)	180,651 (99.4%)	177,224.20***
Retweets authored	64,344 (35.4%)	9,141 (14.2%)	55,203 (85.8%)	32,974.45***
Retweets/tweet	0.35	7.93	0.31	7.05***

Note. All χ² difference test df = 1.

*p < .05. **p < .01. ***p < .001.

Research Question 2 focused on the specific candidates emphasized in the Twitter conversation. Table 2 presents the results of this analysis. Because the frequency of tweets posted between the two groups was vastly different, a chi-square difference test on the proportions of posts associated with each candidate was used to measure the differences between elite and nonelite groups. Here we found no significant differences between the proportions of elite and nonelite user mentions of any candidate.

Table 2.

Frequency and Percentage of Tweets That Reference the Candidates From Each Group.

	Total	Elites	Nonelites	χ²Difference Test
Romney mentions	39,923 (21.95%)	336 (29.17%)	39,587 (21.91%)	1.03
Santorum mentions	22,777 (12.53%)	137 (11.89%)	22,640 (12.53%)	0.02
Gingrich mentions	39,254 (21.59%)	292 (25.35%)	38,962 (21.57%)	0.30
Paul mentions	30,650 (16.85%)	185 (16.06%)	30,465 (16.86%)	0.02

Note. All χ² difference test df = 1; Percentages in the total column are the percentage of the total tweets, and in the elite column it is the percentage of elite authored tweets, with the same applying to the nonelite column.

*p < .05. **p < .01. ***p < .001.

The differences in tone associated with mentions of candidates for each group was also explored. There were no significant differences between the two groups on levels of tone associated with mentions of Mitt Romney, Rick Santorum, or Ron Paul. When examining tone associated with mentions of Newt Gingrich, however, significant differences were observed on levels of valence, arousal, and dominance. Elites communicated with more happiness, excitement, and language of control than did nonelites (see Table 3).

Table 3.

Tone Scores Associated With All Candidates.

	Elite Score, M (SD)	Nonelite Score, M (SD)	t-Statistic (df)
Romney mentions
Valence	.16 (1.00)	.13 (0.86)	0.93 (39,921)
Arousal	.16 (0.99)	.11 (0.77)	1.00 (338.52)
Dominance	.15 (0.92)	.11 (0.78)	0.94 (39,921)
Santorum mentions
Valence	.14 (0.86)	.14 (0.95)	−0.02 (22,775)
Arousal	.16 (0.94)	.12 (0.82)	0.58 (22,775)
Dominance	.13 (0.74)	.12 (0.84)	0.06 (22,775)
Gingrich mentions
Valence	.26 (1.23)	.12 (0.85)	2.01 (293.10)*
Arousal	.25 (1.16)	.11 (0.79)	2.00 (293.03)*
Dominance	.25 (1.16)	.11 (0.76)	2.06 (292.86)*
Paul mentions
Valence	.15 (0.99)	.10 (0.83)	0.79 (30,648)
Arousal	.17 (1.01)	.09 (0.73)	0.99 (185.16)
Dominance	.13 (0.80)	.09 (0.71)	0.69 (30,648)

Note. *p < .05. **p < .01. ***p < .001.

Research Question 3 asked what issues were emphasized in the Twitter conversation and if different issues were emphasized in tweets authored by media and political elites compared to nonelite users. Table 4 shows that the most mentioned issue in the Twitter conversation was “foreign policy,” with 20,224 mentions (28.2% of total issue mentions). This issue was mentioned significantly more than the next most mentioned issue (the economy, with 12,217, or 17.0% of the total), χ²(1) = 1,976.27, p < .001. When examining the differences between the two groups, a chi-square test on the proportions of mentions of each issue from each group was used to determine whether there were differences. Each issue was examined, and no significant differences between the two groups were found.

Table 4.

Frequency of Issue Mentions in Tweets.

Issue	All Mentions (%)	Elite Mentions (%)	Nonelite Mentions (%)	χ²Difference in Proportions
Education	490 (0.68)	0 (0)	490 (0.69)	.69
Health	10,625 (14.82)	62 (16.53)	10,563 (13.81)	.24
Poverty	1,296 (1.81)	5 (1.33)	1,291 (1.81)	.07
Environment	409 (0.57)	0 (0)	409 (0.57)	.57
Elderly	2,565 (3.58)	12 (3.2)	2,553 (3.58)	.02
National defense	9,132 (12.73)	37 (9.87)	9,095 (12.75)	.37
Government spending	3,728 (5.20)	21 (5.6)	3,707 (5.20)	.01
Taxes	3,249 (4.53)	19 (5.07)	3,230 (4.53)	.03
Crime	4,924 (6.87)	25 (6.67)	4,924 (6.90)	.00
Economy	12,217 (17.03)	67 (17.87)	12,150 (17.03)	.02
Foreign policy	20,224 (28.20)	124 (33.07)	20,100 (28.27)	.37
Total	71,718 (100)	375 (100)	71,343 (100)

Note. *p < .05. **p < .01. ***p < .001.

Discussion

The goal of this project was to determine whether there were significant differences between the conversation posted on Twitter by media/political elite users and nonelite users during a presidential debate. The results from this analysis show that nonelite users posted significantly more and were retweeted significantly more than elite users when considering frequency. However, elite users’ comments were retweeted significantly more often per tweet than were the comments of nonelite users. When examining the mentions of specific candidates, the tone associated with those candidates, and the mentions of issues between the Twitter comments of elites and nonelites, there were few significant differences between the two groups. The main difference between the two groups occurred on the levels of tone associated with mentions of Newt Gingrich, who was discussed with more happiness, excitement, and language of control by elite users.

These results imply that while nonelite users posted significantly more than elite users, the conversations were not substantially different. The difference in frequency of tweets and retweets between the two groups is primarily a function of the difference between the sizes of the two groups. However, views of elite users were substantially favored compared with nonelite users as demonstrated by the retweet per tweet ratio. This implies that people spread the posts of elite users more than nonelite users. Based on the retweet per tweet metric and the lack of differences in the elite and nonelite tweet content, we do not find support for the notion that unique citizen frames are being developed on Twitter. For competing frames to emerge, the two content frames should have demonstrated greater differences. In addition, the fact that elites were retweeted more per tweet than nonelite users show that the posts of media and political elites spread farther throughout the discussion and may have been seen by more users. This in turn suggests that their interpretations may have been more influential than the opinions of nonelite users. Together, these results provide some evidence that media and political elites seemed to have a greater influence on the conversation occurring on Twitter than the much larger group of nonelite tweeters.

As previously discussed, past research shows that nonelite citizens and the media/political elites may have differing issue agendas (Jackson-Beeck & Meadow, 1979; McKinney, 2005). Twitter is a space where those citizens have the potential to discuss issues and which candidate might better serve to meet their needs. Our analysis suggests that citizens are not taking full advantage of this developing technology in this regard.

Limitations of This Study and Future Research Implications

This study is not without its limitations. The research is somewhat limited in that it occurs in a relatively nongeneralizable context. The Twitter conversation in this study occurred in reference to a primary debate that was only broadcast on one cable news channel. A live-tweeted primary debate, for example, is much different from the general election debates, as candidates in those events include representatives of both major political parties and those debates are broadcast on all major cable and network news outlets. The audience who watches the general election debates is much larger and more diverse than those who watch primary debates. The audience of the primary debate may also be more politically engaged than other individuals. Results of the current investigation, therefore, may be a function of the homogeneity of the audience attracted to the debate examined or the result of the increased attention to political media that this audience likely paid prior to the debate. However, these results still paint a portrait of the Twitter discussion that occurs within a single-party debate during a period of the campaign in which the party platform is still under construction.

Further examination of what makes a specific Twitter account an elite or nonelite should be explored. For example, should a blogger who runs a popular political blog and live tweets during the debate be considered a political elite? Similarly, should a person who has a significant online following, such as one who operates a satirical Twitter account (e.g., @TeaPartyCat), be considered a media elite? In this project, such individual accounts were not among the elite subset of users. However, as social media continues to transform the media landscape, the notion of what constitutes elite users may also need to change.

Another limitation of this project is with the operationalization of all content analysis features, specifically tone associated with the different political candidates, the candidate mentions, and issue mentions. The operationalization of tone or sentiment in this project is limited as the analytic measure used is based on the mention of specific words within each tweet, the operationalization of candidate mentions was based on uses of the candidate’s first and/or last names, and the operationalization of issue mentions focuses on the use of specific issue words. This semantic approach, unfortunately, cannot account for such linguistic features as irony or sarcasm, or include in its analysis the surrounding verbal context of a full tweet comment. This approach will also miss any misspellings of the words searched for. Since a single tweet message is relatively small, 140 characters or less, the overall context of the tweet is likely necessary—rather than analysis of single words—to determine the true meaning behind a particular expression. Also, to save space for more content, some users concatenate words, which would lead this method to missing those terms. Future research should address this problem by creating statistical models that allow for a prediction of the context to aid tone or sentiment analysis and that can capture misspellings or concatenation of terms in a social media context.

In conclusion, this project did not find many meaningful differences between the tweet conversations posted by elite and nonelite viewers of a nationally broadcast presidential primary debate. The only differences observed were found in the frequency of tweets and retweets authored by the two groups and the levels of tone associated with one politician. These results suggest that citizen deliberation is not a feature of social media dialogue in the debate live-tweeting context and may have implications for representation.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Ampofo

O’Loughlin

Anstead

(2011). Trust, confidence, credibility: Citizen responses on Twitter to opinion polls during the 2010 UK general election. Information, Communication and Society, 14, 850–871.

Benoit

W. L.

Hansen

G. J.

Verser

R. M.

(2003). A meta-analysis of the effects of viewing U.S. presidential debates. Communication Monographs, 70, 335–350.

Berkowitz

(1984). Some effects of thoughts on anti- and prosocial influences of media events: A cognitive-neoassociation analysis. Psychological Bulletin, 95, 410–427.

Borah

(2011). Conceptual issues in framing theory: A systematic examination of a decade’s literature. Journal of Communication, 61, 246–263.

Bradley

M. M.

Lang

P. J.

(1999). Affective norms for English words (ANEW): Stimuli, instruction manual and affective ratings (Tech. Rep. No. C-1). Gainesville: The Center for Research in Psychophysiology, University of Florida.

Chaffee

S. H.

Dennis

(1979). Presidential debates: An empirical assessment. In Ranney

(Ed.), The past and future of presidential debates (pp. 75–106). Washington, DC: American Enterprise Institute for Public Policy Research.

Cole

Hawthorne

(2013). Issue ownership in the 2008 presidential election acceptance speeches and debates. Presented at the Annual Convention of the Central States Communication Association, Kansas City, MO.

Entman

R. M.

(2003). Cascading activation: Contesting the White House’s frame after 9/11. Political Communication, 20, 415–432.

Entman

R. M.

(2004). Projections of power framing news, public opinion, and U.S. foreign policy. Chicago, IL: University of Chicago Press.

10.

Fiegerman

(2012, December 18). Twitter now has more than 200 million monthly active users. Mashable. Retrieved March 15, 2013, from http://mashable.com/2012/12/18/twitter-200-million-active-users/

11.

Fridkin

K. L.

Kenney

P. J.

Gershon

S. A.

Shafer

Woodall

G. S.

(2007). Capturing the power of a campaign event: The 2004 Presidential debate in Tempe. Journal of Politics, 69, 770–785.

12.

Gaffney

(2010). #iranElection: Quantifying online activism. Presented at the Websci10: Extending the Frontiers of Society Online, Raleigh, NC. Retrieved from http://journal.webscience.org/295/

13.

González-Bailón

Banchs

R. E.

Kaltenbrunner

(2012). Emotions, public opinion, and U.S. presidential approval rates: A 5-year analysis of online political discussions. Human Communication Research, 38, 121–143.

14.

Hamdy

Gomaa

E. H.

(2012). Framing the Egyptian uprising in Arabic language newspapers and social media. Journal of Communication, 62, 195–211.

15.

Hellweg

S. A.

Pfau

Brydon

S. R.

(1992). Televised presidential debates: Advocacy in contemporary America. New York, NY: Praeger.

16.

Jackson-Beeck

Meadow

R. G.

(1979). The triple agenda of presidential debates. Public Opinion Quarterly, 5, 173–180.

17.

Larsson

A. O.

Moe

(2011). Studying political microblogging: Twitter users in the 2010 Swedish election campaign. New Media & Society, 5, 729–747.

18.

McKinney

M. S.

(2005). Let the people speak: The public’s agenda and presidential town hall debates. American Behavioral Scientist, 49, 1–15.

19.

McKinney

M. S.

Carlin

D. B.

(2004). Political campaign debates. In Kaid

L. L.

(Ed.), Handbook of political communication research (pp. 203–234). Mahwah, NJ: Lawrence Erlbaum.

20.

McKinney

M. S.

Chattopadhyay

(2007). Political engagement through debates: Young citizens’ reactions to the 2004 presidential debates. American Behavioral Scientist, 50, 1169–1182.

21.

McKinney

M. S.

Houston

J. B.

Hawthorne

(2012). Live-tweeting the presidential debates: Using social media to process political campaign communication. Presented at the Annual Convention of the Association for Education in Journalism and Mass Communications, Chicago, IL.

22.

Morozov

(2009). Iran: Downside to the “Twitter revolution”. Dissent, 56, 10–14.

23.

Morozov

(2012). The net delusion: The dark side of internet freedom (Reprint.). New York, NY: PublicAffairs.

24.

Petrocik

Benoit

W. L.

Hansen

(2003). Issue ownership and presidential campaigning, 1952-2000. Political Science Quarterly, 118, 599–626.

25.

Rahimi

(2011). The agonistic social media: Cyberspace in the formation of dissent and consolidation of state power in postelection Iran. Communication Review, 14, 158–178.

26.

Rusack

R. J.

(2012, October 17). Dallas debate tweeters rock: Nearly 1,000 tweets logged on #dmndebate during the second Obama-Romney showdown. Trail Blazers Blog. Retrieved October 22, 2012, from http://trailblazersblog.dallasnews.com/2012/10/dallas-debate-tweeters-rock-nearly-1000-tweets-logged-on-dmndebate-during-the-second-obama-romney-showdown.html/

27.

Seidman

(2012, January 27). Updated: Florida GOP Debate Averages 5.4 Million Viewers on CNN. TVbytheNumbers. Retrieved November 27, 2012, from http://tvbythenumbers.zap2it.com/2012/01/27/florida-gop-debate-averages-5-4-million-viewers-on-cnn/117850/

28.

Steeper

F. T.

(1978). Public response to Gerald Ford’s statements on eastern Europe in the second debate. In Bishop

G. F.

Meadow

R. G.

Jackson-Beeck

(Eds.), The presidential debates: Media, electoral, and policy perspectives (pp. 81–101). New York, NY: Praeger.

29.

The Times-Union. (2012, October 17). Tweeters using #jaxdebate favor Obama in debate. jacksonville.com. Retrieved October 22, 2012, from http://jacksonville.com/news/national/2012-10-17/story/tweeters-using-jaxdebate-favor-obama-debate

30.

Tsfati

(2003). Debating the debate. The International Journal of Press/Politics, 8, 70–86.

31.

Tumasjan

Sprenger

T. O.

Sandner

P. G.

Welpe

I. M.

(2011). Election forecasts with Twitter: How 140 characters reflect the political landscape. Social Science Computer Review, 29, 402–418.

32.

Twindex: Inside the political Twitterverse. (2012, August 10). The War Room with Jennifer Graham. Current TV. Retrieved from http://current.com/shows/the-war-room/videos/twindex-inside-the-political-twitterverse