Who Tweets About Politics?

Abstract

Twitter has been lauded for its potential political value by academics, journalists, and politicians; yet, we know little about the citizenry’s use of Twitter to engage in politics. Under the backdrop of the 2011 gubernatorial elections, we observed Twitter users’ direct engagement in the electoral process by collecting usernames and tweets of anyone who mentioned a candidate. After the elections were called, we employed an original survey via Twitter of these political tweeters to answer the question, who tweets about politics? Unsurprisingly, the results of our logit analysis demonstrate that strong partisans and those exhibiting high levels of traditional political participation activity tweet about politics most often which supports those who argue that we can expect to find the same political activists online as offline. However, we also find evidence that racial minorities and secularists are engaged in the electoral conversation on Twitter suggesting that some marginalized groups may have found a political outlet.

Keywords

twitter Internet social media political participation behavior gubernatorial election voting

Americans’ use of social media has gone beyond casual socializing. Candidates engage with the electorate via Facebook and governmental agencies disseminate information on Twitter. Similarly, citizens use social networking sites (SNS) to engage in political behavior, such as asking their Facebook friends to vote or to keep informed about political candidates and elections by following politicians and journalists on Twitter. Although some early studies criticized the Internet’s alleged democratizing effects, most scholars now agree that it has affected the citizenry’s engagement in the political arena. Even skeptics have pronounced that the Internet has produced positive effects on political participation, though the effects thus far are small, contingent on various exogenous factors, and unstable over time (Bimber & Copeland, 2011).¹ Despite a growing consensus, we still have much to learn about whom social networking sites impact and how.

We try to untangle some of these complexities by employing an original survey of Twitter users to determine who participates in politics on Twitter. Under the backdrop of the 2011 gubernatorial elections in four states—Kentucky, Louisiana, Mississippi, and West Virginia—we observed citizens engaging in the electoral process on Twitter for a 30-day period prior to the elections. We collected the Twitter usernames and tweets of anyone who mentioned one of the gubernatorial candidates. After the winners of these races were declared, in typical Twitter fashion, we tweeted 1,448 politically minded users and asked them to respond to our survey about their Twitter usage during the campaigns. An impressive 21% of them obliged!²

This study answers two key questions for scholars interested in political behavior. Who participates in electoral politics on Twitter and are they the usual suspects (high education, highly engaged in politics, older)? Or does Twitter bring typically marginalized groups into the electoral process? If we think of posting about a gubernatorial election on Twitter as one type of political participation, we might expect to find the usual suspects doing most of the tweeting. The results show that people who tweet about the gubernatorial elections align with our expectations in terms of being strong partisans with a propensity to participate in politics through voting, volunteering, and other traditional participatory acts. However, the results also demonstrate that some politically marginalized groups—racial minorities and secularists—are tweeting about politics. This study should interest scholars of both conventional and nonconventional forms of political participation, since many of the activities that citizens engage in on Twitter are now intricately intertwined with politics offline. This research also has implications for electoral campaigns, as candidates who go online to reach potential voters may simply be tweeting to the choir.

The next section examines what we know about Twitter and what sets it apart from other SNS. The third section explores the relationship between the Internet, social media, and political participation. We use these theories to develop our hypotheses about who tweets about politics. The fourth section discusses the methodology for obtaining our data and details our hypotheses. Finally, we present the findings using both qualitative and quantitative analysis. This is significant because it allows us to detail Twitter users’ political behavior both online and offline.

Internet, Social Media, and Twitter

In less than 2 years since its initial launch in 2006, the microblogging site Twitter became one of the fastest growing SNSs in the United States. By 2008, politicians, candidates, and political campaigns widely used the site to connect with citizens and potential voters (Cogburn & Espinoza-Vasquez, 2011), including gubernatorial candidates (Johnson, 2011; Pole & Xenos, 2011). Scholars have studied the intersection between the Internet and politics for more than 20 years. Despite this, we have only recently focused on social media and have barely scratched the surface of Twitter.

Americans’ use of social media is no longer “new media.” Sixty-six percent of Americans, who use the Internet, use SNSs (A. Smith, 2011). Just as the Internet has various functions, not all SNS offer the same experience (Pasek, More, & Romer, 2009). Twitter users account for 15% of social media users, and the number of people who use Twitter daily has doubled in less than a year, making it the fastest growing SNS (A. Smith, 2011).³

Candidates quickly recognized that they could use the sites to reach out to voters (Gulati & Williams, 2009). Previous studies have examined candidates, parties, and campaigns using Twitter. The fact that campaigns have open and accessible profiles facilitates this type of research. Conversely, fewer studies examine individual citizens or potential voters on these same sites, because their profiles are private or hidden and because there is currently no publicly available database of users. Several large sample surveys are available to researchers, such as the Pew Internet and American Life Project. However, they do not allow us to directly observe user behavior during an election and, thus far, have only focused their questions on national elections. Other studies have been limited by their selection of specific individuals’ SNS behavior, such as college students (E. Hargittai & Litt, 2011, 2012), celebrities (Marwick & boyd, 2011), or academics (Focus, 2010). Our study observes citizens at the source of interest by examining actual Twitter users while they are engaged in political behavior. Indeed, our contacts were made as a result of their tweeting about politics.

Our examination of political behavior on Twitter allows us to address, at least partially, some of the limitations of prior studies. First, Twitter’s unique, open, and public platform allows us to directly observe users’ behavior while they are engaging in gubernatorial politics, unlike surveys that gauge individuals’ behavior during an entire electoral cycle. Second, we enhance existing studies of national elections by examining gubernatorial elections. Scholars have already documented important differences between these types of elections (Craig, Kane, & Gainous, 2005). For example, fewer people vote or pay attention during gubernatorial elections than during presidential or congressional elections. In addition, few studies to date examine politics on the Internet at the state level (Pole & Xenos, 2011). Thus, our research opens up a new avenue for comparing and contrasting state and national elections. Finally, though our sample is restricted to individuals who tweeted during a 30-day period before the four-state gubernatorial elections, we broadly look at Twitter users and do not limit our study to college students or any other relatively distinct group.

The Intersection of Participatory Politics and Twitter

The Internet, social media, and Twitter have all been lauded by proponents of democracy for being able to engage “average citizens” in the political process (Carpenter, 2010). As Internet access diffused and SNSs debuted, scholars and pundits alike anticipated that new and different types of people would be brought into the political process by arming users with information about the candidates and by linking voters with politicians (Gainous & Wagner, 2011; Johnson, 2011; McNeal, Tolbert, Mossberger, & Dotterweich, 2003; Panagopoulos, 2009). Once armed and linked, these users may be stimulated to vote or otherwise participate politically (Tolbert & McNeal, 2003). We put this assumption to task by observing who tweets about politics during the 2011 gubernatorial elections. Are these tweeters new or different entrants to the political arena, or are they the same people who are typically engaged in the political debate? We argue that individuals who are already participating in the electoral process and are interested in politics are most likely to be our political tweeters.

This research can be understood within the digital divide debate (Norris, 2001). Some argue that the Internet is a new avenue for participation by the same people (Hindman, 2009) and the technology actually exacerbates political inequalities (Schlozman, Verba, & Brady, 2010). Those who traditionally participate in the political process will adopt and use the most relevant technologies, maintaining or even expanding their political control. Thus, the participatory sphere is not broadened. As applied to Twitter, the theory suggests that those who are usually marginalized or unengaged politically are unlikely to tweet about politics, accept or respond to tweets about the political process, or otherwise engage in electoral politics on Twitter.

Those on the democratizing side of the digital divide debate argue that the Internet can bring new participants into the political system by removing barriers that favor traditional participants and conventional forms of participation (Mossberger, Tolbert, & McNeal, 2008). The Internet has the potential to broaden the sphere of democracy by promoting communication and providing information needed to participate. As applied to electoral politics on Twitter, the previously unengaged can communicate and exchange political views with other citizens, connect with politicians, and may be encouraged to engage in traditional forms of political participation reducing the digital divide.

The current challenge is determining how the Internet impacts participation and for whom it has an impact. This study mostly responds to the question of who is participating online, though we also provide some qualitative evidence for how citizens are engaged on Twitter (see Table 4). This study advances our understanding of the link between Twitter and participatory democracy by identifying who tweets about politics to determine if they are the typical participants. Candidates, too, might find the information useful in determining who they are reaching with social media campaigns.

Who Are Likely Political Tweeters?

Individuals with a personal interest in a specific topic, such as celebrities or politics, are more likely to adopt (E. Hargittai & Litt, 2011) and use Twitter for tweeting about those topics (E. Hargittai & Litt, 2012). For example, a study of tweeters during a 2009 German election found average citizens tweeting about the election, but a small percentage dominated most of the online debate (Tumasjan, Sprenger, Sandner, & Welpe, 2011). A similar result was found among Swedish election tweeters where political elites (politicians, journalists, and popular political bloggers) disseminated the most, if not the most significant, tweets (Larsson & Moe, 2011). These studies imply that our political tweeters are likely to be those already interested in politics.⁴ Others have shown that the Internet, blogs, and Twitter encourage clustering of like-minded individuals (Adamic & Glance, 2005; Conover et al., 2011; E. Hargittai, Gallo, & Kane, 2008; Yardi & boyd, 2010), whereby politically interested, partisan individuals might easily find each other on Twitter. Therefore, existing studies show that people who tweet about politics are likely to be politically interested partisans who tend to be older, educated, individuals who regularly cast a vote, or participate in politics in a variety of other ways. Thus, we expect them to be our political tweeters.

Although our Twitter users are caught in a political act online, we do not know whether they participate in politics offline? One of the most consistent findings is that the Internet has spurred more people to give politically affiliated donations including bringing in new donors, small donors, and younger donors (Bimber, 2001; Graf, Institute for Politics, & the, 2006; Panagopoulos & Bergan, 2009; Park & Perry, 2009). Some have found that gathering information online or being prompted by others to vote via e-mail, chat rooms, or text increases voter turnout including among younger voters (Bond et al., 2012; Dale & Strauss, 2009; Kirby, Marcelo, Gillerman, & Linkins, 2008; Mossberger et al., 2008; Tolbert & McNeal, 2003), while others did not find any significant effects of Internet use on voting (Bimber, 2001). E-mailing about politics, chatting online, and gathering political information online also increased traditional forms of participation such as attending political rallies, contacting politicians, joining a protest, and other civic activities (Dalton, 2008; Park & Perry, 2009; Shah, Kwak, & Holbert, 2001; Weber, Loumakis, & Bergman, 2003; Xenos & Moy, 2007; Zukin, 2006).

Another group of scholars report less positive results for participatory democracy. They argue that these are not new or different participants; rather they are the same people who were engaged in politics before the Internet (Schlozman et al., 2010; Weber et al., 2003), and ostensibly before Twitter. They lament that the Internet has not been the democratizing tool that many initially claimed it would be. To the contrary, it has reinforced, and in some cases intensified, typical participation patterns in which the young, the poor, and the uneducated are left behind (Schlozman et al., 2010).⁵ We asked our political tweeters if they had attended a political rally, volunteered time, donated money, talked about politics in order to change someone’s mind, put a campaign sign in their yard, wore a button, and/or voted. This line of questioning allows us to explore the issue of the digital divide by querying respondents about their involvement in politics.

Since research on the intersection of politics and the Internet is still somewhat limited and existing studies provide mixed results, it is a challenge to provide clear hypotheses on who our political tweeters might be and whether they are engaged citizens outside of Twitter. Our studies are likely at a crossroads. We now know that the Internet and social media have the capacity to engage new people in the political process, while simultaneously providing new ways for the standard bearers of participation to connect. Yet, we are far from understanding who, how, and why. Therefore, we base our hypotheses primarily upon our understanding of those we typically expect to see engaged in the political process. After all, tweeting about the electoral process encompasses several components of traditional participation including gathering information, engaging with the candidates, and talking about politics, to name a few. The next section describes our survey methods and details our specific hypotheses.

Data and Method

Twitter differs from Facebook because the content is, by default, publicly available to all of its users, not simply a user’s “friends” as is true of Facebook. Consequently, Twitter users are more likely to interact with people who are friends of friends, celebrities, politicians, and strangers, thereby providing a unique opportunity to determine who participates in politics on the second most popular SNS. Moreover, Twitter’s conspicuous characteristics proved advantageous in gathering survey responses, because users are accustomed to interacting with people whom they do not know.

We used the 2011 gubernatorial elections in four states—Kentucky, Louisiana, Mississippi, and West Virginia—as the backdrop for examining who tweets about politics.⁶ This allowed for a manageable, yet relatively higher profile election with which to gauge who participates in politics online.⁷ We have little reason to suspect that the results found here would be significantly different if our data were collected during gubernatorial elections held in different states. Nonetheless, it is worth noting that our data were collected in southern states, which might present some differences, especially for the church attendance variable discussed below. In addition, 36 states hold their gubernatorial elections during midterm election years, while the rest hold elections during presidential election years. We might also see differences between our data and that of states that hold elections during presidential years, because having the president on the ballot typically encourages participation.

Although Twitter provides a unique opportunity to study citizens’ political behavior on an SNS, it also has some challenges and limitations. First, there is no known method of obtaining a random and representative sample of all Twitter users beyond employing national surveys, which do not provide us with any of the qualitative data that many scholars are interested in, such as what people tweet about.⁸ At this point, our studies are limited by our selection of specific users or by our selection of a particular topic that can be searched. It is not unusual for researchers studying Twitter to focus on one group of users, one topic, or one case, such as an election (Golbeck, Grimes, & Rogers, 2010; Larsson & Moe, 2011; Lassen & Brown, 2010; Yardi & boyd, 2010). Thus, we collected Twitter usernames from accounts that mentioned one of the eight major party gubernatorial candidates during a 30-day period prior to the general election.⁹ Each of the major party’s candidates held a Twitter account that was used as a campaign tool. Thus, each candidate had an equal chance of being mentioned.

Following the end of the election, we collected 1,448 unique usernames who tweeted during the most political of all cases, an election. Not unlike exit pollsters surveying voters as they leave the polling station to understand voting behavior, we surveyed Twitter users after they tweeted about politics to understand more about this social media phenomenon. A link to our survey was publicly tweeted to each user from one author’s Twitter account that read, “I’m a professor studying tweeters & elections. Please take my 5 min survey. Thanks! (please don’t RT) [survey link].” The researcher also modified her own Twitter user profile to include her professional title, university name, and professional website address, so that curious or suspicious users could easily determine that she was an actual political scientist and the survey was legitimate research, rather than spam.¹⁰

Our study analyzes these data both qualitatively and quantitatively to try and flesh out who tweets about politics. Table 1 provides a summary graphic of the data we collected during the 30-day period leading up to the elections and isolates the total number of tweets about each gubernatorial candidate (n = 3,969). The second column shows the number of tweets sent by our Twitter users. Since each user could have sent more than one tweet, the third column shows the number of unique Twitter users who were sent our survey. The number of tweets sent by candidates is listed in the fourth column. All but one candidate sent numerous tweets though 78% of all tweets about a candidate came from someone other than the candidate himself or herself. Candidates sent an average of 3.5 tweets per day over the 30-day period, while each of our unique Twitter users posted a total of 2.15 tweets on average. Looking at the percentage of total tweets about each candidate in the fourth column, candidates, or their campaigns, do not appear to be dominating the conversation on Twitter.¹¹ However, some candidates were certainly more active than others. Most notably, Tara Hollis dominated the conversation accounting for 91% of all tweets about her race. She also received the lowest percentage of the vote of any of the candidates (17%). Jindal’s popularity and national name recognition probably had more to do with this rather than any Internet effect.¹²

Table 1.

Tweets by Candidate.

Gubernatorial Candidate (Party, State)	Tweets by Twitter Users^a	Unique Twitter Users	Tweets by Candidate (% of Total Tweets About Candidate)	Total Tweets About Candidate (Row Total)
Bobby Jindal (R, LA)	630	308	33 (5)	663
Tara Hollis (D, LA)	24	13	234 (91)	258
Phil Bryant (R, MS)	800	376	233 (23)	1,033
Johnny Dupree (D, MS)	499	231	71 (12)	570
David Williams (R, KY)	234	77	187 (44)	421
Steve Beshear (D, KY)	391	131	63 (19)	454
Bill Maloney (R, WV)	290	172	24 (8)	314
Earl Ray Tomblin (D, WV)	255	140	1 (0.003)	256
Column totals	3,123	1,448^b	846	3,969

^aOnly Twitter user tweets not including candidate tweets.

^bMany Twitter users mentioned a candidate’s name over several tweets. This is the number of unique tweeters. This also corresponds to the number of users who were sent the survey.

Since this is the first study to our knowledge that required interacting with users on Twitter, rather than secretly collecting their data, we took several precautions that also placed some limitations on our study. First, we knew we would need personal demographic information from our participants that surveyors have previously reported as difficult to obtain, even in an anonymous and confidential setting. Twitter is neither. Even though our survey instrument was not tied to a user’s Twitter ID, we essentially “played it safe” in the quantity and type of questions we asked. We limited our questions to the most basic demographics and focused our substantive questions on political participation. In order to limit nonresponses, we did not want the survey to take any more than 5 min to complete and we did not ask respondents to disclose their income. While we understand that an individual’s income may be an important component of whether someone tweets about politics or not, we believe that education is a more reliable predictor of social status, particularly since users of SNS tend to be younger and may likely be college students without a steady income. Ideally, we would have expanded our questionnaire to include more questions about why one tweets, including information about Internet skills and use of other SNSs. But given our desire to increase responses and all the unknowns surrounding a survey of Twitter users, we decided to keep the survey as short and simple as possible.

Finally, the lack of prior studies that contact strangers in an online environment left us wanting to protect ourselves from backlash on Twitter or in person, since one author was easily identified by her user profile. Interestingly, our cautious design was not completely unwarranted as we received several angry tweets in response to our survey. These ranged from name calling to threats of reporting us to Twitter for spam, which would have shut down our account. All in all, though, it was a positive experience that gained one author many new Twitter followers.

Measurement of Variables and Hypotheses

Our survey included questions about how often respondents used Twitter and their primary purpose for using the site which are used as the dependent variables (see the Appendix for key survey questions). First, we asked about general Twitter behavior and how often they used Twitter. Five response categories were used ranging from never (0) to everyday (4). This is used as the dependent variable General Twitter Use. Next, we asked how often respondents use Twitter to tweet about three specific topics: (1) to connect with family and friends, (2) to tweet about or with celebrities, and (3) to tweet about politics using five response categories (0 = never, 1 = rarely, 2 = some of the time, 3 = most of the time, 4 = all the time). The primary dependent variable in the ensuing discussion is the frequency of tweeting about politics, which is modeled along with the other ordinal response variables in Table 5 using ordinal logit.

A range of demographic and political participation questions serve as independent variables. Typically, we would expect older and more educated respondents to be more active politically, especially in terms of voting (Flanigan & Zingale, 2002; Rosenstone & Hansen, 1993) but also, by extension, engaged in other political activities, such as tweeting about politics. Age was indicated by a six-category measure (18–20, 21–29, 30–39, 40–49, 50–59, 60 or older). Education is indicated by a five-category measure ranging from lowest (high school diploma or equivalent) to the highest (graduate degree). However, if newly developed technologies, such as Twitter, are successful in attracting new participants to the political arena, we might expect younger (e.g., Park & Perry, 2009; Tolbert & McNeal, 2003), and ostensibly less educated, individuals to tweet about politics. Nevertheless, we maintain that older, educated users will be tweeting about politics.

We also included variables for gender and race. Women and minorities show few differences from men and Whites in their participation levels assuming equal socioeconomic status (Flanigan & Zingale, 2002); therefore, we do not expect any differences in gender or race among our political tweeters. However, two related factors might contribute to higher levels of Black participation in the current study: First, Black empowerment zones have been shown to develop in jurisdictions that have achieved significant levels of Black representation (Bobo & Gilliam, 1992). With the election of Barack Obama in 2008, the United States may have become an empowerment zone. Second, in the Mississippi governor’s race, one of the four governor’s races featured in the current research, one of the candidates in the general election was an African American, the first to be a candidate in the state since reconstruction. This may well contribute to Black interest in this election and to increased tweeting among African Americans. Gender is indicated by male (0) and female (1). Race was initially measured with six categories (White, African American, American Indian or Alaskan Native, Asian, Native Hawaiian or Pacific Islander, and multiple races). For our analysis, we recoded race as non-White (0) and White (1).

We also used church attendance as a measure of religious participation. Our initial interest in this variable came from our focus on southern states’ elections, but religious participation’s effect on participation is not regionally restricted (Verba, Schlozman, & Brady, 1995; Wald, Kellstedt, & Leege, 1993).¹³ Though church attendance does not cover the breadth of religiosity, frequency of attendance has been shown to correlate with political participation (Rosenstone & Hansen, 1993; Verba et al., 1995). We measure attendance with a simple frequency of attendance scale ranging between never attending (0) to attending more than once per week (4).

Party identification was measured on the typical 7-point scale from Strong Republican (1) to Strong Democrat (7). For our logit analysis, we created a new variable from party identification that reports party ID strength, while ignoring the direction using four categories: 1 = independent, 2 = leaning, 3 = weak, and 4 = strong. Existing studies show that Republicans have caught up to Democrats in their use of the web (Gainous & Wagner, 2011); therefore, we do not expect to see partisan differences for political tweeting, but we do hypothesize that strong partisans are more likely than weak partisans or Independents to tweet about politics.

Finally, we asked respondents a series of six questions about their political participation levels in the past 6 months: (1) had they voted in one of the four state’s gubernatorial elections (0 = no, 1 = yes); and how often had they engaged in a variety of political activities: (2) attended a political meeting, (3) volunteered for a campaign or candidate, or (4) talked to someone about politics in the past 6 months (four categories where 0 = never, 1 = 1–2 times, 2 = 3–4 times, 3 = 5 or more times—at least 30% of our respondents indicated that they had done at least one of the three activities 5 or more times). Using dummy variables, we also queried them about (5) donating money or (6) putting election signs in their yard (0 = no, 1 = yes). Ultimately, we created a general “political participation index,” aggregating the six questions above by recoding the frequency scaled measures as dummy variables to create a 7-point index ranging between 0 and 6 to measure political participation. Following others who demonstrate that those who participate in politics online are those who already vote and otherwise participate in politics (Norris, 2001; Schlozman et al., 2010), we hypothesize that higher levels of traditional participation will be associated with high levels of political tweeting.

Results

The intersection of Twitter and politics is relatively uncharted and untested research territory (Karpf, 2012), such that our scholarship would be best served by including a wide range of summary data tools in order to reflect our observations from different angles. As such, we provide a variety of summary statistics and descriptors, not only to be transparent about our findings but also to provide the scholarly community with clues toward an Internet paradigm (Gainous & Wagner, 2011). Table 2 reports characteristics of our sample of respondents who tweeted about one of the 2011 four-state gubernatorial candidates compared with national samples of Twitter users from two Pew Internet surveys taken in 2010 and 2012. However, the Pew data should also be viewed with caution. After isolating the Twitter users from their data, we were only left with 126 respondents in 2010 and 355 respondents in 2012, in comparison to our 306 users. In addition, the 2010 Pew data show only a small percentage of Americans, 8%, used Twitter; however, by 2012 that number had jumped to 15%, making Twitter the fastest growing SNS today (A. Smith & Brenner, 2012).

Table 2.

Comparison of 2011 Gubernatorial Election Tweeters With 2010 and 2012 Pew Sample of Twitter Users.

Variable	2010 Pew Data^a	2012 Pew Data^a	2011 Gubernatorial Election Data
Average Age	40	47	30–39
% College Degree	48	44	74
% White	62	60	80
% Male	37	47	71
% Democratic Party	42	40	47
% Ideology/Strong Party ID^b	10	14	57
% Pure Independents^c	15	13	7
	n = 126	n = 355	n = 306

^aData source: A. Smith (2011); Rainie and Smith (2012).

^bThe Pew survey does not have a question on partisan strength, so we used their question on ideology which states: “Would you describe your political views as: Very Conservative, Conservative, Moderate, Liberal, Very Liberal.” The percentages reported here are those who said they were very conservative or very liberal. Our gubernatorial data reports those who said they were strong Democrats or strong Republicans.

^cThe Pew survey does not include a question to gauge pure Independents. They ask, “As of today do you lean more to the Republican Party or more to the Democratic Party?” The only acceptable responses were lean Republican, lean Democrat, Don’t Know, or Refused to Answer. Here we report the percentage of respondents who said, “don’t know.”

Table 2 shows that our 2011 four-state gubernatorial election sample of political tweeters is younger, better educated, about evenly divided between Democrats and Republicans, and more heavily male and White than the Twitter Pew samples. One characteristic that is undoubtedly distinctive about the gubernatorial election sample is the high percentage of respondents who are strong partisans. Nearly 57% of our respondents said they were strong Democrats or Republicans compared with less than 15% in the Pew data. According to an analysis of party identification between 1952 and 2000 using American National Election Studies (ANES) data, the percentage of strong partisans in the American South, ignoring the direction of partisanship, varied between a low of 26% and a high of 42% (Flanigan & Zingale, 2002). In the year 2008, the most recent ANES data available, the percentage of strong partisans was 32%. Thus, by comparison, a larger than expected percentage of our tweeters are strongly committed to one party or the other.

Now, we turn to our first research question: Who participates in electoral politics on Twitter? White, middle-aged, college educated, males made up the majority of our respondents. Strong partisans and a few Independents were also represented. These descriptive data, especially when compared to the Pew data, provide one indicator that familiar political voices are speaking out on Twitter when it comes to gubernatorial elections.

Table 3 further details our data and makes comparisons on key variables between those who tweeted about politics and those who tweeted generally, which we previously noted as two of our dependent variables. By comparing those who said they tweet most often about politics with general tweeting behavior, we are likely tapping into the differences between those who mentioned a candidate daily with those who mentioned a candidate once or by chance. For example, Republicans, Independents, and Democrats reported similarly high levels of general Twitter use (86% of Democrats reported daily use, as did 95% of Independents, and 91% of Republicans; the differences are not statistically significant). Twitter use for political purposes, however, winnows out many of the Independents and some of the partisans. About 69% of Democrats and Republicans reported that they tweeted politically “all of the time” or “most of the time,” while only 37% of Independents gave this response. Partisan strength seems to be a major contributor to that relationship. For example, while 37% of Independents tweeted regularly about politics (combining “all of the time” and “most of the time” categories), 58% of “leaners,” 48% of weak partisans, and fully 77% of strong partisans were regular political tweeters. On the other hand, Independents were also more likely than any of the partisans to say they rarely or never tweeted about politics (31%) even though 94% of them said they tweeted everyday.

Table 3.

Political Tweeting and General Tweeting Behavior Compared (Select Variables Only).

	Tweeting About Politics					General Tweeting
	All Time	Most Time	Sometimes	Rarely	Never	Everyday	Near Everyday	1–2× Week	Rarely/Never
Republicans	46.38	22.46	20.29	9.42	1.45	91.37	8.63	0	0
Democrats	44.03	24.63	24.63	3.73	2.99	85.93	11.11	2.22	0.74
Independents	31.58	5.26	31.58	15.79	15.79	94.74	0	5.26	0
Strong Partisans	53.66	23.78	15.24	6.10	1.22	87.95	10.24	1.81	0
Weak Partisans	20.00	28.00	36.00	12.00	4.00	84.00	16.00	0	0
Leaners	35.80	22.22	32.10	6.17	3.70	91.36	7.41	0	1.23
White	42.17	23.04	23.47	8.23	3.04	90.52	8.62	0.86	0
Non-White	52.46	19.67	21.31	3.23	3.23	83.61	11.48	3.28	1.63

Note. All entries are percentages.

N = 291–293.

Non-Whites in this sample report a higher rate of regular use of Twitter (tweeting “all” or “most of the time”) for the purposes of communicating with family and friends and celebrities. They also reported that they were more likely to tweet politically. That is, among non-White respondents 72% say they tweet about politics “all” or “most of the time,” while 65% of Whites make similar reports. Men were slightly more likely to report that they followed politics “all” or “most of the time” than women in our sample (46% to 41% respectively). The preponderance of our respondents (66%) are in the age category 21–39. That being said, the proportion of those in that age group who tweeted about politics regularly (most or all the time) was comparatively low at 42% (n = 192).¹⁴

Although the summary statistics provide a starting point to understand who tweets about politics, a closer examination of the tweets supports evidence that certain types of users may dominate the conversation. We isolated the top 20 tweeters from our survey. The user who tweeted the most of any other user sent 62 tweets or 1.9% of all user tweets which is a smaller percentage than a previous study that performed a similar analysis of top tweeters (Larsson & Moe, 2011). Larsson and Moe’s (2011) top tweeter performed 3.2% of all “singleton” tweets in a national election. We also visited the pages of these top tweeters and 15 of the 20 top tweeters’ profiles noted that they were politically active in some way. Two were politicians, several others were heads of their local or state parties, and still others mentioned politics among their interests. Obviously, we cannot make any determinations regarding who read or received these users’ tweets, but individuals might have a better opportunity at having their voice heard on matters of state elections, than national elections, if there is less noise.

A closer look at the tweets also illustrates how the users think about political participation. Each tweet was coded to determine what, if any, political activities users said they were engaged in. We used the political participation index questions we asked in our survey as a starting point, but also looked for nontraditional acts.¹⁵ Table 4 demonstrates that although the political communication medium of Twitter is anything but traditional, our political tweeters mentioned and were engaged in traditional political participation activities such as voting (21.9% of all their tweets), attending a rally or other political event (12.9%), and campaigning or volunteering for a candidate (8%). Tweeters also asked others to get involved in these activities. We also noted some less conventional types of political participation such as signing online petitions and receiving text messages from candidates, though they were mentioned less than 1% of the time. Though receiving text messages may seem like a passive act, it is likely that individuals had to sign up to receive messages or to be contacted in the first place. However, it is also possible that some received the texts from candidates unwillingly. For example, several users posted complaints of one candidate sending out texts in the middle of the night. Others reported receiving “robocalls” from the candidates. Although we are unable to determine who responded to our tweeters by becoming politically involved themselves, we know that these, seemingly, small bits of political communication on Twitter have the potential to turn into acts of political participation beyond the initial user (Bond et al., 2012).¹⁶

Table 4.

Tweets by Type of Political Participation Mentioned by Users.^a

Vote	21.9%
Attend (also watch, listen)	12.9%
Campaign/Volunteer	8%
Donate money, talk about politics, post a yard sign or wear a button, other nontraditional participatory acts	1% or less of all tweets

Note. ^aEach of the six categories we coded for here roughly corresponds to one of the six political participation questions (independent variables), we asked on our survey as shown in the Appendix.

N = 3,123 tweets by users (not including tweets by candidates).

Recall that our primary dependent variable of interest, Tweeting About Politics, was measured with a 5-value frequency scale ranging from never (0) to all the time (4). We modeled our primary research question, who tweets about politics, using ordinal logit as shown in Table 5.¹⁷ We also estimated three other models of Twitter use; tweeting to connect with family and friends, tweeting about celebrities and general tweeting behavior, to make comparisons across models. Two thirds of our respondents (66.7%) indicated that they either tweeted about politics all of the time or most of the time. By way of comparison, 39.23% reported similar levels of tweeting to connect with family and friends, and 14.39% reported similarly when asked about how often they tweeted about celebrities.

Table 5.

General, Family and Friends, Celebrities, and Political Twitter Usage.

Independent variables	General		Family		Celebrities		Politics
Independent variables	Coefficients	Odds	Coefficients	Odds	Coefficients	Odds	Coefficients	Odds
Age	−.046 (.153)	.955	−.265 (.089)**	.767	−.257 (.091)**	.773	.139 (.091)	1.15
Gender	.164 (.446)	1.18	.250 (.244)	1.28	.425 (.253)	1.53	−.308 (.254)	.735
Education	.029 (.190)	1.03	.023 (.108)	1.02	.097 (.108)	1.10	.026 (.116)	1.03
Race	−.746 (.441)	.474	.177 (.291)	1.19	.660 (.287)*	1.94	.754 (.309)*	2.12
Church Attendance	.063 (.156)	1.07	.206 (.089)*	1.23	−.076 (.089)	.927	−.251 (.091)**	.778
Party ID Strength	−.376 (.220)	.687	.066 (.115)	1.07	.056 (.118)	1.06	.281 (.120)*	1.32
Participation	.150 (.116)	1.16	.124 (.069)	1.13	−.048 (.070)	.953	.329 (.074)***	1.39
Pseudo R²	.028		.029		.025		.060
	p = .485		p = .000		p = .008		p = .000
	n = 284		n = 279		n = 276		n = 282

Note. The first cell entries are coefficients with standard errors in parentheses. Odds ratios are reported in the second column.

*Significant at p = .05. **Significant at p = .01. ***Significant at p = .001.

Table 5 shows the results of our four ordinal logit models about tweeting using the odds ratio to interpret the results: general, family and friends, celebrities, and political tweeting.18 The first model, general Twitter behavior, does not have any predictive value. In two of the four models, age is significant. Older respondents tweet less often about family and friends and celebrities relative to younger respondents. However, age is not a significant predictor in the political tweeting model. We also found that church attendance has a positive and significant relationship with family and friends tweeting behavior, whereas as discussed below, it has a negative relationship with political tweeting. Race was only significant for the celebrity tweeting model, such that non-Whites were more likely than Whites to tweet about celebrities.

Our final and primary model of interest, tweeting about politics, included a mixed bag of expected and unexpected results. Not surprisingly, the measures of political interest (party ID strength and political participation) were both positively related to political tweeting and support our hypotheses. For a one-level increase in partisan strength (e.g., leaners vs. independents), the odds of tweeting about politics are 1.32 times higher, holding all else constant. In other words, as strength of partisanship increases so does political tweeting. In addition, respondents who participated in the electoral process, by voting, volunteering, attending political rallies and the like, were more likely to tweet about politics. As political participation increases by one level, the odds of tweeting about politics are 1.39 times higher, holding all else constant. Church attendance and race were also significant for political tweeting but were not in the expected direction. Church attendance yielded an odds ratio of .78 such that with a 1-unit increase in attending church, the odds of tweeting about politics are .78 times smaller. Those who attended church less often were more likely to tweet about politics. We expected for church attendance to have a positive relationship with political tweeting. Finally, our results show that the odds of tweeting about politics are 2.12 times higher for racial minorities than Whites. We did not expect to see any differences.

As a first cut at the question, “Is Twitter reaching out to the marginalized and bringing them into the political sphere?” we would have to answer that the data are mixed. The characteristics of our sample indicate that political tweeters are well educated and mostly male, and relatively few are over the age of 40. Young people as a group are less likely to vote, so the fact that they are more heavily represented in our sample is an indication that Twitter might be contributing to an expansion of the potential electorate. Even if they are not tweeting about politics as much as proponents of democracy would like, that they can be found on Twitter discussing gubernatorial politics at all is a positive outcome and evidence of one form of political engagement. This finding supports previous research that social media has played a role in pulling more youth into politics (Park & Perry, 2009; Tolbert & McNeal, 2003).

On the other hand, the fact that the majority of respondents are strong partisans and only a small percentage are pure independents offers support for the “business as usual” hypothesis; Twitter is just another “tool” in the toolbox of political activists. It might be that the politically sophisticated person (well educated and interested in politics) is attracted to Twitter and is also engaged in more traditional political activities. Our respondents, in other words, may be what political scientists have referred to in the past as “activists.” This line of thinking is reinforced by the positive association between regular types of electoral participation/strength of partisanship and political tweeting.

One of our most important findings for bringing diverse participants into the political sphere is the percentage of racial minorities that report tweeting about politics. This is made even more encouraging by the fact that the campaigns used as the basis for this research were gubernatorial rather than presidential since fewer people on average are involved in state-level politics. A recent survey found that racial minorities are more likely than Whites to use Twitter, so this finding may not be too surprising, but we are just beginning to learn how and if they use it for political purposes (Rainie & Smith, 2012). It is also notable that Mississippi’s gubernatorial election was an historic one with Democratic candidate Johnny DuPree being the first African American to be the nominee of a major party since reconstruction (Robertson, 2011). His presence on the ballot may have created the beginning of a state-level Black empowerment zone in Mississippi (Bennie Thompson, also an African American, represents the state’s Delta region in Congress) even though he was ultimately defeated. Though we cannot test the empowerment zone thesis in this report, it remains a potential topic for future researchers.¹⁹

Finally, we observed that church attendance was negatively related to tweeting about politics. While our measure of “religiosity” was obviously a simplified one, it is an initial step at understanding an important, if seldom studied component of politics and elections. In particular, if one objective is to determine who is participating in politics online and whether they are the usual suspects or not, our findings demonstrate that one of the most marginalized, yet growing, groups in American politics, the “none” religion group, may be discussing elections on Twitter (G. Smith & Funk, 2012). Although this result is new, unconfirmed by others, and not without its limitations as described above, it is not surprising since SNSs allow groups to come together, interact, and express themselves in ways that were unimaginable before.

Conclusion

Twitter and other social media sites have the potential to draw the unengaged into politics. Though this research demonstrates that some familiar faces—highly politically involved, partisan individuals—are extending their offline political reach to Twitter, we also find evidence that racial and secular minorities, too, are tweeting about politics. This is another indication that social media has increased participatory democracy, even if it is at the margins. These results are not unlike previous studies in that they are “contingent and ubiquitous” (Bimber & Copeland, 2011). Nevertheless, they enhance our understanding of the intersection of the Internet and politics at this unstable, albeit transitional, period in electoral politics in which elements of participatory democracy are emerging online.

Though this study did not attempt to explain who these Tweeters are reaching, if anyone, through their messages of support for the candidates or pleas for others to vote, it is probably safe to say that our highly involved partisans are paying attention to gubernatorial politics and how the candidates are presenting themselves on Twitter. Those who were already interested in campaigns and engaged with the workings of their government are likely using Twitter to find and communicate with those who share similar interests (Schlozman et al., 2010). Therefore, one implication of this study is for candidates and campaigns to understand who they may be targeting. It appears as if Twitter may be a good place to reach a highly engaged voting base who may, in turn, reach out to others (Bond et al., 2012). Campaigns should also consider its utility for reaching new or different participants, such as the racial and secular minorities we found tweeting about politics. Twitter’s public and engaging interface opens the possibility for users to unknowingly be brought into the political process.

There are many other key questions of political behavior research that we were unable to answer here, such as whether these Twitter users had an increased sense of political efficacy or trust in institutions as a result of their interactions, or lack thereof, with politicians. We limited our study to political participation as predictors of tweeting about politics, but our results suggest that we have only started to scratch the surface of understanding this new political behavior. Most obviously, future studies should include an income variable, but we should also examine other predictors outside of political participation and typical demographics that could cause a person to tweet about politics in a public forum. In particular, motivations of social capital and political trust might play a role, particularly since distrust in government and inefficacious feelings among citizens are high. Our finding that individuals who did not attend church are tweeting about politics is also an interesting one. While these results are tenuous, if we seek to understand marginalized groups and how they interact on Twitter then the growing atheist or agnostic populations should be examined. Like Internet studies before this (Bimber & Copeland, 2011; Lassen & Brown, 2010), we contend that more research needs to be done to better understand the who’s, how’s and why’s of tweeting about politics.

Finally, this study should be of interest to scholars beyond those interested in the Internet and its impact on politics. While SNS, even Twitter, may come and go, Internet technology is here to stay. As such, candidates and voters will continue to use the technology to communicate with one another. Studies such as this enable us to bridge the gap between “traditional” political science and the ways campaigns and politicians are connecting with voters online as we build new paradigms for the future (Gainous & Wagner, 2011). Although the methods may be imperfect, ignoring politics online or waiting for the perfect data will prevent us from accumulating knowledge and answering the larger more interesting questions. Today, parties serve as the primary link between elected officials and citizens; yet, the Internet, and certainly social media, has the power to serve that role.

Footnotes

Appendix

Acknowledgments

The authors would like to thank Amber Clark for her data collection assistance for this research. We would also like to thank Jason Gainous, Dave Karpf, Marcus McLean, Matt Stambaugh, Dave Garson, and the anonymous reviewers for their assistance during various stages of the process.

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.

Notes

References

Adamic

L. A.

Glance

(2005). The Political Blogosphere and the 2004 US Election. Paper presented at the Proceedings of the WWW-2005 Workshop on the Weblogging Ecosystem, Japan.

Bimber

(2001). Information and political engagement in America: The search for effects of Information Technology at the individual level. Political Research Quarterly, 54, 53–67.

Bimber

Copeland

(2011). Digital media and political participation over time in the US: Contingency and ubiquity. Annual Meeting of the European Consortium for Political Research, Reykjavik.

Bobo

Gilliam

F. D.

(1992). Race, sociopolitical participation, and black empowerment. American Political Science Review, 84, 377–393.

Bond

R. M.

Fariss

C. J.

Jones

J. J.

Kramer

A. D. I.

Marlow

Settle

J. E.

Fowler

J. H.

(2012). A 61-million-person experiment in social influence and political mobilization. Nature, 489, 295–298.

Carpenter

C. A.

(2010). The obamachine: Technopolitics 2.0. Journal of Information Technology & Politics, 7, 216–225.

Cogburn

Espinoza-Vasquez

(2011). From networked nominee to networked nation: Examining the impact of web 2.0 and social media on political participation and civic engagement in the 2008 Obama campaign. Journal of Political Marketing, 10, 189–213.

Conover

M. D.

Ratkiewicz

Francisco

Goncalves

Flammini

Menczer

(2011). Political polarization on Twitter. Paper presented at the Proceedings of the Fifth International AAAI Conference on Weblogs and Social Media, Indiana University, IN.

Craig

S. C.

Kane

J. G.

Gainous

(2005). Issue-related learning in a gubernatorial campaign: A panel study. Political Communication, 22, 483–503.

10.

Dale

Strauss

(Eds.). (2009). Mobilizing the mobiles: Text messaging and turnout. New Brunswick, NJ: Rutgers University Press.

11.

Dalton

R. J.

(2008). The good citizen: How a younger generation is reshaping American politics. Washington, DC: CQ Press.

12.

Flanigan

W. H.

Zingale

N. H.

(2002). Political behavior of the American electorate. Washington, DC: CQ Press.

13.

Focus

(2010). Twitter in higher education 2010: Usage habits and trends of today's college faculty. Special Report. Accessed on September 6, 2012, http://www.facultyfocus.com/free-reports/twitter-in-higher-education-2010-usage-habits-and-trends-of-todays-college-faculty/

14.

Gainous

Wagner

K. M.

(2011). Rebooting American politics: The Internet revolution. Lanham, MD: Rowman & Littlefield.

15.

Golbeck

Grimes

J. M.

Rogers

(2010). Twitter use by the U.S. Congress. Journal of the American Society for Information Science and Technology, 61, 1612–1621.

16.

Graf

, Institute for Politics, D., & the, I. (2006). Small donors and online giving: A study of donors to the 2004 presidential campaigns. Washington, DC: Institute for Politics, Democracy & the Internet, Graduate School of Political Management, George Washington University.

17.

Gulati

G. J. J.

Williams

C. B.

(2009). Closing gaps, moving hurdles: Candidate web site communication in the 2006 campaigns for congress. In Panagopoulos

(Ed.), Politicking online: The transformation of election campaign communications (pp. 48–76). New Brunswick, NJ: Rutgers University Press.

18.

Hampton

Sessions

Rainie

Purcell

(2011). Social networking sites and our lives: The Pew Internet & American Life Project (pp. 1–85). Accessed August 15, 2012, http://www.pewinternet.org/Reports/2011/Technology-and-social-networks.aspx

19.

Hargittai

Gallo

Kane

(2008). Cross-ideological discussions among conservative and liberal bloggers. Public Choice, 134, 1–2.

20.

Hargittai

Litt

(2011). The tweet smell of celebrity success: Explaining variation in Twitter adoption among a diverse group of young adults. New Media and Society New Media and Society, 13, 824–842.

21.

Hargittai

Litt

(2012). Becoming a tweep: How prior online experiences influence Twitter use. Information Communication and Society, 15, 680–702.

22.

Hindman

(2009). The myth of digital democracy. Princeton, NJ: Princeton University Press.

23.

Hoffman

L. H.

(2012). Participation or communication? An explication of political activity in the internet age. Journal of Information Technology & Politics, 9, 217–233.

24.

Johnson

D. W.

(2011). Campaigning in the twenty-first century: A whole new ballgame? New York, NY: Routledge.

25.

Karpf

(2012). Social science research methods in internet time. Information, Communication & Society, 15, 639–661.

26.

Kirby

E. H.

Marcelo

K. B.

Gillerman

Linkins

(2008, June). The Youth vote in the 2008 primaries and caucuses. CIRCLE Fact Sheet. Retrieved August 10, 2012, from www.civicyouth.org

27.

Larsson

A. O.

Moe

(2011). Studying political microblogging: Twitter users in the 2010 Swedish election campaign. New Media & Society, 13, 1180–1197.

28.

Lassen

D. S.

Brown

A. R.

(2010). Twitter: The electoral connection? Social Science Computer Review, 29, 420–421.

29.

Long

J. S.

Freese

(2005). Regression models for categorical dependent variables using stata. College Station, TX: Stata Press.

30.

Marwick

boyd

(2011). I tweet honestly, I tweet passionately: Twitter users, context collapse, and the imagined audience. New Media & Society, 13, 114–133.

31.

McNeal

R. S.

Tolbert

C. J.

Mossberger

Dotterweich

L. J.

(2003). Innovating in digital government in the American states. Social Science Quarterly, 84, 52–70.

32.

Mossberger

Tolbert

C. J.

McNeal

(2008). Digital citizenship: The internet, society and participation. Cambridge, MA: MIT Press.

33.

Norris

(2001). Digital divide: Civic engagement, information poverty, and the Internet worldwide. Cambridge, NY: Cambridge University Press.

34.

Panagopoulos

(Ed.). (2009). Politicking online: The transformation of election campaign communications. New Brunswick, NJ: Rutgers University Press.

35.

Panagopoulos

Bergan

(2009). Clicking for cash: Campaigns, donors, and the emergence of online fund-raising. In Panagopoulos

(Ed.), Politicking online: The transformation of election campaign communications (pp. 127–140). New Brunswick, NJ: Rutgers University Press.

36.

Park

H. M.

Perry

J. L.

(2009). Do campaign web sites really matter in electoral civic engagement? Empirical evidence from the 2004 and 2006 post-election internet tracking survey. In Panagopoulos

(Ed.), Politicking online: The transformation of election campaign communications (pp. 101–123). New Brunswick, NJ: Rutgers University Press.

37.

Pasek

Romer

(2009). Realizing the social internet? Online social networking meets offline civic engagement. Journal of Information Technology & Politics, 6, 197–215.

38.

Pole

Xenos

(2011). Like, comments and retweets: Facebooking and Tweeting on the 2010 Gubernatorial Campaign Trail. State Politics and Policy Conference, Hanover, NH.

39.

Rainie

Smith

(2012). Social networking sites and politics (Dataset). Pew Internet. Accessed September 13, 2012, http://pewinternet.org/Reports/2012/Social-networking-and-politics.aspx

40.

Robertson

(2011, October 2). For politics in South, race divide is defining. New York Times. Accessed October 10, 2012, http://www.nytimes.com/2011/10/03/us/for-politics-in-south-race-divide-is-defining.html?pagewanted=all&_r=1

41.

Rosenstone

Hansen

J. M.

(1993). Mobilization, participation, and democracy in America. In Niemi

(Ed.), Controversies in voting behavior (pp. 69–83). Washington, DC: Congressional Quarterly Press.

42.

Schlozman

K. L.

Verba

Brady

H. E.

(2010). Weapon of the strong? Participatory inequality and the internet. Perspectives on Politics, 8, 487–509.

43.

Shah

Kwak

Holbert

R. L.

(2001). “Connecting” and “Disconnecting” with civic life: Patterns of internet use and the production of social capital. Political Communication, 18, 141–162.

44.

Sippey

(2012). Changes coming in version 1.1 of the Twitter API. Twitter.com Announcements. Accessed on October 10, 2012, https://dev.twitter.com/blog/changes-coming-to-twitter-api

45.

Smith

(2011). Why Americans use social media. The Pew Internet & American Life Project. Accessed August 15, 2012, http://www.pewinternet.org/Reports/2011/Why-Americans-Use-Social-Media.aspx

46.

Smith

Brenner

. (2012). Twitter Use 2012. Pew Internet & American Life Project. Accessed September 6, 2012, http://www.pewinternet.org/Reports/2012/Twitter-Use-2012.aspx

47.

Smith

Funk

(2012). “Nones” on the rise. Pew Forum on Religion & Public Life. Accessed October 10, 2012, http://www.pewforum.org/Unaffiliated/nones-on-the-rise-preface.aspx

48.

Tolbert

C. J.

McNeal

R. S.

(2003). Unraveling the effects of the internet on political participation? Political Research Quarterly, 56, 175–185.

49.

Tumasjan

Sprenger

Sandner

Welpe

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

50.

Verba

Schlozman

K. L.

Brady

H. E.

(1995). Voice and equality: Civic voluntarism in American politics. Cambridge, MA: Harvard University Press.

51.

Wald

K. D.

Kellstedt

L. A.

Leege

D. C.

(1993). Church involvement and political behavior. In Leege

D. C.

Kellstedt

L. A.

(Eds.), Rediscovering the religious factor in American politics (pp. 121–138). New York, NY: M. E. Sharpe.

52.

Weber

Loumakis

Bergman

(2003). Who participates and why?: An analysis of citizens on the Internet and the mass public. Social Science Computer Review, 21, 26–42.

53.

Wolfe

Gould

(1998). An approximate likelihood-ratio test for ordinal response models. Stata Technical Bulletin, 7, 199–204.

54.

Xenos

Moy

(2007). Direct and differential effects of the internet on political and civic engagement. Journal of Communication, 57, 704–718.

55.

Yardi

boyd

(2010). Dynamic debates: An analysis of group polarization over time on Twitter. The Bulletin of Science, Technology & Society, 30, 316–327.

56.

Zukin

(2006). A new engagement?: Political participation, civic life, and the changing American citizen. Oxford, NY: Oxford University Press.