The Rise of Twitter in the Political Campaign: Searching for Intermedia Agenda-Setting Effects in the Presidential Primary

Transcription

1 Journal of Computer-Mediated Communication The Rise of Twitter in the Political Campaign: Searching for Intermedia Agenda-Setting Effects in the Presidential Primary Bethany A. Conway Kate Kenski Department of Communication, University of Arizona, 1103 E University Blvd, Tucson, AZ Di Wang Department of Humanities and Arts, Macau University of Science and Technology, C209 Avenida Wai Long, Taipa, Macau Questions exist over the extent to which social media content may bypass, follow, or attract the attention of traditional media. This study sheds light on such dynamics by examining intermedia agenda-setting effects among the Twitter feeds of the 2012 presidential primary candidates, Twitter feeds of the Republican and Democratic parties, and articles published in the nation s top newspapers. Daily issue frequencies within media were analyzed using time series analysis. A symbiotic relationshipwasfoundbetweenagendasintwitterpostsandtraditionalnews,withvaryinglevelsofintensity and differential time lags by issue. While traditional media follow candidates on certain topics, on otherstheyareabletopredictthepoliticalagendaontwitter. Keywords: Twitter, Agenda Setting, Presidential Primary, Social Media, Campaigns, New Media. doi: /jcc The impact of social media on the political communication landscape is a pressing topic in the study of political campaigns (Gueorguieva, 2008; Towner & Dulio, 2012). Scholars are struggling to understand the effects of social media on the various types of agenda setting (Bennett & Iyengar, 2009). Much research has investigated the incorporation of social media into the campaign package (Adams & McCorkindale, 2013; Conway, Kenski, & Wang, 2013; Golbeck, Grimes, & Rogers, 2010; Graham, Broersma, Hzelhoff, & van t Harr, 2013; Grant, Moon, & Grant, 2010; Gueorguieva, 2008; Johnson & Perlmutter, 2010; Towner & Dulio, 2012). A growing line of research has also examined the complex relationship between topics trending in the Twittersphere and issues in traditional media (Groshek & Groshek, 2010; Metzgar & Maruggi, 2009; Sayre, Bode, Shah, Wilcox, & Shah, 2010). This study focuses on the relationship between the latter and the Twitter feeds of political candidates and parties. Editorial Record: First manuscript received on September 21, Revision received on January 14, Accepted by S. Shyam Sundar on March 4, Final manuscript received on March 12, First published online on May 11, Journal of Computer-Mediated Communication 20 (2015) International Communication Association 363

2 We examine the extent to which campaigns and parties are reliant on major news outlets to provide content for regular Twitter posts or choose to communicate a unique agenda that may then predict media content. We investigate this in the context of the 2012 presidential primary the first presidential election in which Twitter was heavily used. While the newspaper-to-twitter relationship suggests news media are influencing candidate and party reactions on Twitter, the reverse indicates Twitter may constitute a resource to bypass media gatekeepers, serving as a new source for journalistic content. Our results suggest both are taking place the influence between Twitter and the top newspapers in the U.S. is reciprocal, generally speaking. The high-speed nature of social media results in an interdependent relationship similar to that of blogs and websites. Politicians are still reliant on traditional media, but there are specific topics on which candidate Twitter feeds were able to predict the newspaper agenda to some extent. SNS Influence on Media Output Agenda-setting theory, which suggests that the media have the power to determine the important issues of the day (McCombs & Shaw, 1972), is currently undergoing scrutiny (Bennett & Iyengar, 2009). As we shift to an environment in which anyone can produce news and social media influence our understanding of the world (Mayer & Cornfield, 2008), scholars question whether the power of media gatekeepers has been overturned (Groshek & Groshek, 2013; Sayre et al., 2010). Predictions of how the abundance of information sources online might influence the impact of traditional media range from modification, to attenuation, to reversal of influence (Sayre et al., 2010). Reverse agenda setting brings us full circle and suggests the public, long seen as a minimal player in agenda setting, may influence news media (McCombs, 2004). New media could therefore impact agenda-building processes such as media agenda building and intermedia agenda setting. According to Denham (2010), agenda building more generally focuses on the macrolevelinfluences that determine the spread of issue publics among political elites, mass media, and the voting populace. Under this umbrella, media agenda building focuses on the influences facing media professionals, including extramedia influences like campaign information subsidies (Shoemaker & Reese, 1996), while intermedia agenda setting focuses on the transfer of issue salience across media (McCombs, 2004). In the communication literature, agenda building is most often applied to the influence of campaign press releases, while intermedia agenda setting is applied to advertisements, websites, and blogs. Recent studies examining social media influence on the news agenda also tend to apply an intermedia agenda-setting framework. The growth of online news sources makes this application all the more relevant (Denham, 2010). We, therefore, employ an intermedia agenda-setting framework while also taking into account findings on campaign press releases. 1 Press releases deliberately supplied to media practitioners often have agenda-building effects (Dunn, 2009; Kiousis, Mirook, Wu, & Seltzer, 2006; Lancendorfer & Lee, 2010; Tedesco, 2001; Tedesco, 2006). Thoughpressreleaseshaveastrongerinfluenceonthemediaagendathanthereverse,inmostcasesthese relationships are reciprocal. Both political advertising and campaign websites, approached through an intermedia agenda-setting framework, may also influence the television and newspaper agenda (Boyle, 2001; Lopez-Escobar, McCombs, & Lennon, 1998; Roberts & McCombs, 1994) although these relationships may also be reciprocal (Ku, Kaid, & Pfau, 2003). Research examining the relationship between candidate blogs and television news suggests campaigns may be more reliant on the media agenda than the reverse (Sweetser, Golan, & Wanta, 2008). Studies examining political discussions across the entire Twittersphere suggest the relationship between social and traditional media is generally reciprocal, with social media having limited 364 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

3 to moderate influence on news (Groshek & Groshek, 2010; Metzgar & Marrugi, 2009; Neuman, Guggenheim, Jang, & Bae, 2014). Teasing out such influences is challenging, but there is evidence that social media are being used to circulate information without having to rely on traditional media to act as gatekeeper or fact-checker or moderator (Metzgar & Marrugi, 2009, p. 152). Groshek and Groshek (2013) found that Twitter was more likely to follow traditional media than the reverse. In a recent study, focused on the top political issues of the day, Neuman and colleagues (2014) not only found that social media communicated a distinct agenda compared to that of traditional media, it was a better predictor of traditional media than the reverse. To understand social media influence, there appears to be a need for more focused examinations of big data (Groshek & Groshek, 2013). For example, in a study of Proposition 8 in California, Sayre et al. (2010) found a reciprocal relationship between YouTube and California newspapers. This demonstrates social media s potential to lead and follow traditional media. The former effect is important because it opens the possibility of social media putting forth an entirely unique agenda, which could influence public reaction. Microblogging outlets similar to Twitter, such as China s Sina Weibo, have also served as a place for citizens to discuss emergent issues, in turn influencing traditional media (Wu, Atkin, Mou, Lin, & Lau, 2013). Jungherr s (2014) findings during the 2009 German election suggest that overall Twitter and traditional media follow a different logic, but this was not the case when it came to the leading candidates. With high-profile individuals and events, Twitter may play a role similar to that of traditional news media, functioning more like a news outlets than a social one (Kwak, Lee, Park, & Moon, 2010). Moving toward a focused examination of big data, we investigate the impact of political candidate and party Twitter feeds on news media output. Twitter as a Political Platform Social media are now essentials of the political campaign (Perlmutter, 2008; Towner & Dulio, 2012). Among the different types of social media, the social networking site (SNS) is considered unique because it allows users to display a network of connections openly (boyd & Ellison, 2007) and has been touted for enhanced facilitation of politician-public interaction (Towner & Dulio, 2012). Unlike SNSs that emphasize privacy controls, Twitter users have mainly public profiles that do not require bidirectional confirmation of connections (boyd & Ellison, 2007) and are limited in the characters allowed in user posts. Today, Twitter is being adopted as a new political podium to diffuse information to a growing audience. Political events such as the Twitter Townhall held by the Obama administration (Office of the Press Secretary, 2011) and the Twitter debate held by the 2012 GOP candidates (Khandaroo, 2011) demonstrate its power to facilitate politician-constituent dialogue and, effectively, create news. Research thus far reveals candidates and campaigns are struggling to harness SNSs, as revealed through disparate activity levels and a penchant for unidirectional messaging (Aparaschivei, 2011; Conway et al., 2013; Golbeck et al., 2010; Graham et al., 2013; Klinger, 2013; Stromer-Galley, 2014). Some candidates exploit its interactive features, but many choose to use Twitter as a broadcast medium rather than a venue for discussion with voters (Stromer-Galley, 2014). In 2012, U.S. presidential primary candidates also used links extensively in tweets (Conway et al., 2013). It is not clear whether unidirectional messaging and use of hyperlinks indicate a reliance on information supplied by traditional media. A New Avenue for Influence As candidates already used websites and blogs to draw-in journalists (Lawson-Borders, & Kirk, 2005; Lipinski & Neddenriep, 2004), it is likely that SNSs serve a similar purpose. Research is needed, however, to understand how candidate and party Twitter feeds influence media coverage of elections. Journal of Computer-Mediated Communication 20 (2015) International Communication Association 365

4 Qualitatively, one can see why candidates on Twitter should outperform the entire Twittersphere when it comes to influencing the media. Barack Obama is currently the third most followed member of the Twitter community ( In terms of retweets, politicians and political parties have been labeled influentials on Twitter (Dang-Xuan, Stieglitz, Wladarsch, & Neuberger, 2013). Further, journalists report using Twitter to find story leads, follow politicians, and equate Twitter content with other information subsidies such as campaign press releases (Farhi, 2009; Parmelee, 2013a; Parmelee, 2013b). Reporters describe Twitter as a resource to collect data, enhance audience reach, and follow the activities of high-profile news sources, the latter working to fulfill an agenda-building function (Parmelee, 2013b). Beyondstudiesofreportingpractices,theinfluenceofcandidatesocialmediaontraditionalmedia agendas is to a certain extent backed by intermedia agenda-setting studies on candidate websites and blogs, though such research is limited and largely based on cross-lagged correlations of ranked issues. Relationships between candidate websites and traditional media appear reciprocal; however, websites are stronger predictors of media coverage than the reverse, with the incumbent holding sway (Ku et al., 2003). A study of television news, advertising, and blog content during the 2004 presidential election found that candidate blog agendas were also correlated with subsequent media coverage (Sweetser et al., 2008). We, therefore, propose: H1: Twitter feeds will predict the agenda of traditional news media. Yet, in that same study Sweetser and colleagues (2008) found that blogs reacted to the media agenda to a greater extent than the reverse. Heavy use of media content within blogs may be attributed to its low-cost nature, the need for regular updates, and the avoidance of potential backlash rationales that could all be applied to candidate Twitter feeds. Research shows presidential primary candidates make heavy use of attack appeals in blogs (Trammell, Williams, Postelnicu, & Landreville, 2006), as well as heavy use of external media hyperlinks in general election blogposts (Williams, Trammell, Postelnicu, Landreville, & Martin, 2006) and Twitter feeds (Conway et al., 2013). Because direct communication by candidates, especially attacks, may be perceived as lacking legitimacy, candidates may look to news coverage to provide validity (Lipinski & Neddenriep, 2004). Due to the frequent nature of social media posts, Marwick and boyd (2011) suggest self-censorship and attribution to traditional media as mechanisms for message control. This rationale and findings on candidate blogs result in the following hypothesis: H2: Traditional media will predict the agenda of Twitter feeds. Similar to Sayre and colleges (2010), we do not consider these hypotheses contradictory. The relationship between candidate media and news coverage has demonstrated reciprocal time and again and may run in both directions at different points in time. As new media forms grow in popularity, we shift from their ability to set the agenda to their power to set the agenda the crux of the current agenda-setting debate. Therefore, we ask the following research question: RQ1: Did Twitter feeds predict the newspaper agenda to a greater extent than the reverse? Issue Ownership Issue choice is an important determinant in the agenda-building process (Walgrave & Van Aelst, 2006). For example, the media are expected to have greater agenda-setting power when they act as primary 366 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

5 sources, while politicians are expected to hold sway when they own or take responsibility for the topic at hand. Issue ownership is defined as a party s reputation for handling a given issue based on historic concern (or lack there of) for that issue and innovative approaches toward its treatment (Petrocik, 1996). Voters then associate the parties with various issues based on management by incumbents and makeup of party constituencies. This foundation is fairly stable,but scholars recognize that parties canestablish a lease on non-owned issues based on perceived incumbent inadequacies (Petrocik, 1996; Petrocik, Benoit, & Hansen, 2004). Presidential candidates focus on party-owned issues with notable, strategic exceptions. For example, Republicans place more emphasis on party-owned issues than do Democrats (Benoit & Hansen, 2004; Petrocik et al., 2004). Democratic emphasis on Republican issues may be attributable to the impossibility of avoiding GOP issues such as taxation, limited government, and national defense (Benoit & Hansen, 2004; Petrocik et al., 2004). Politicians are also more willing to trespass on issues when they are trailing in the polls or see an issue as important to the public (Damore, 2004). Recent research further suggests thatissueconvergencemaybethenorm,asbothsidesgainfromstressingthoseissuesseenasimportant to the public (Sigelman & Buell, 2004). Overall, candidate emphasis on issues is dynamic and dependent on myriad factors. Regardless of possible convergence, candidates traditionally see more favorable media coverage on owned issues, with defense, taxes, and spending benefitting Republicans and social welfare benefitting Democrats (Hayes, 2008). High discrepancy has been found between media and campaign emphasis on issues, with the U.S. media linking candidates to party-owned issues even when campaign agendas do not reflect this divide (Petrocik et al., 2004). This may not only dissuade candidates from so-called issue trespassing, but also suggests the media play a pivotal role in the maintenance of issue ownership. Such emphasis on issue ownership by the media makes it important that we analyze the issues on which candidates are the most influential. Because of media emphasis on owned issues, we suggest the following: H3A: Twitter feeds affiliated with the Democratic Party will have greater success at setting the media agenda when it comes to Democratic-owned issues. H3B:TwitterfeedsaffiliatedwiththeRepublicanPartywillhavegreatersuccessatsettingthemedia agenda when it comes to Republican-owned issues. Method To test this intermedia agenda-setting relationship, newspaper articles from the nation s top newspapers and tweets from presidential primary candidate Twitter feeds, campaign feeds, and those of the Republican and Democratic parties were analyzed for issue mentions over the course of the 2012 Republican presidential nomination campaign. Computer-assisted content analysis and time series analysis, ideal for testing the application of theory to big data (Neuman et al. 2014; Parks, 2014), were used to analyze the relationship between media. Dataset Newspapers included The Wall Street Journal, The New York Times, USA Today, The Los Angeles Times, and The Washington Post. The first three were chosen as the largest circulation newspapers in the country (stateofthenewsmedia.org, 2012). Though The New York Daily News has a higher circulation than The Los Angeles Times, the latter paper was chosen as representative of the western United States. Finally, The Washington Post, the eighth-largest circulation newspaper in the US, was chosen due to its emphasis on national politics. Journal of Computer-Mediated Communication 20 (2015) International Communication Association 367

6 All articles published from 26 December 2011 through 3 May 2012, were collected using Lexis-Nexis and Pro Quest. 26 December marked one week prior to the Iowa primary on 3 January. 3 May marked the resignation of Gingrich s campaign, thus cementing Romney s bid for the Republican nomination (Shear, 2012). Articles were retrieved using the term "president * " in combination with "Republican Primar *," "Republican Caucus *," "Republican nomination," or any of the names of the primary candidates. All full news articles and editorials focusing on the primary election or the candidates were included, resulting in an entire population of articles. If any of the returned articles was irrelevant or did not focus on the primary election, they were discarded, resulting in a total of 2,884 relevant articles. 2 Tweets posted on individual candidate Twitter feeds, candidate campaign feeds, and the feeds operated by the Republican National Committee (RNC) and Democratic National Committee (DNC) were gathered over the same time period. All those identified as Republican candidates at the start of the collection period had verified Twitter feeds. They included Michele Bachmann, Herman Cain, Newt Gingrich, Jon Huntsman, Fred Karger, Ron Paul, Tim Pawlenty, Rick Perry, Mitt Romney, and Rick Santorum. Data for campaign feeds were also included. Not all candidates had verified campaign feeds. Thus, candidate campaign feeds only included Gingrich2012HQ, Huntsman2012HQ, TeamBachmann, and TeamRomney. Tweets for Barack Obama s individual Twitter feed were gathered for comparison. Tweets were gathered by regularly querying Twitter s API (N = 11,322). These data were then entered into an excel spreadsheet, displaying the user name, the date a tweet occurred, and its content. Coding Procedures QDA Miner, a computer-assisted qualitative coding program, and its quantitative component WordStat, a text mining software tool, were used to code and analyze issue frequencies within news coverage and Twitter data. Such computer-assisted analyses are typical of projects using big data (Parks, 2014). For newspapers, the unit of analysis was all news content concerning the primary election published in a newspaperonagivenday.fortwitterfeeds,theunitofanalysiswasalltweetspostedtoanindividual account on a given day. Coding of the project data was conducted through WordStat, but the researchers manually created the project dictionary that facilitated the coding. Automated coding is necessary with such large datasets; however, a combination of automated and manual analysis ensures that the former is coded as intended by the researcher (Dang-Xuan et al., 2013). Because issue priorities differ across elections, the program WordStat was used to develop a specific dictionary for the current project. First, we harvested issue topics used by Tedesco (2001) in a previous agenda-setting study that was also performed using computer-assisted coding. The dictionary included 21 issues with subsequent key word groupings. Using these as a starting point, all news and Twitter data were entered into QDA Miner, and words and word frequencies were analyzed in Wordstat to create a complete list of issue categories and develop each category with all related words. This required manual examination of thousands of words. In order to appear under a given category, a word had to appear 10 times or more. Hashtags employed in Twitter feeds were also analyzed and placed into the appropriate categories. For example, the hashtag #BUFFETTRULE, a plan for a minimum 30% tax rate on those making more than $1 million, was placed in the Taxes category. After adding words to the project dictionary based on this initial run through, the data were analyzed using the issue categories developed thus far, and leftover words (those not included in the dictionary) were examined to make sure none were left out of the proposed categories. The authors and two communication researchers unaffiliated with the project then went through the dictionary to ensure that words funneled into each category were truly representative. Any question as to the use of a word was resolved by looking at that specific word in context, both in tweets and in news articles. The final dictionary included 21 policy issues. 3 As expected, some of Tedesco s (2001) categories were not included in the current study. This process resulted in a dictionary that we believe fully captures issue use. 368 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

7 With computer-assisted content analysis, we assume the computer is 100% reliable with the indices it is given; however, word indices may or may not be valid they may not represent what the researchers intend. Combining human and computer coding, with computers providing the reliability and humans providing the validity check, gives confidence in the patterns analyzed. To accomplish this, a researcher unaffiliated with the project performed validity checks on 100 random references to the seven categories used in the current analysis (see below) from each medium. Agreement across all categories was α =.82 or above. 4 Analysis In order to examine the relationship between issue emphasis in news articles and Twitter content, the top seven issues across sources were identified. These seven issues collectively constituted approximately two thirds of issue mentions. 5 These issues were: budget, economy, employment, energy, foreign policy/trade, health care, and taxes. Due to limited tweeting by Republican candidates and their campaigns, Republican candidate and campaign tweets were summed together to form an index. Mentions across the five newspapers were also summed to form a newspaper index. Thus, we analyzed the relationship between newspapers and four Twitter sources: Barack Obama (personal feed), the Republican candidates (personal and campaign feeds), the DNC, and the RNC. Time series of the frequencies of mentions from all domains were examined for linear and quadratic trends. They were then de-trended, if the linear or quadratic trend was statistically significant, so that the relationship between the days could be analyzed without concerns over autocorrelation (Romer, 2006). Then, the correlations between the de-trended time series were calculated and used to evaluate the strength of relationship among the sources of interest. In other words, we rigorously tested the relationships among Twitter indices/feeds and the traditional news index by evaluating the correlations as discovered through content analysis. It is important to note here that we do not equate correlation with causality. Correlations in this case, as with all statistical findings, simply suggest a nonrandom relationship (Sayre et al., 2010). Results As shown in Table 1, newspapers mentioned issues much more frequently than did the Twitter feeds. Considering the character limit in Twitter feeds, this finding is not surprising. It is, however, different from research on the entire Twittersphere, where frequency of issue mentions in social media are significantly more numerous than in traditional media (Neuman et al., 2014). Candidates also mentioned topics to varying degrees. For example, the Republican candidates made energy their number 1 topic with 466 mentions during the 130 days analyzed. In comparison, these candidates only tweeted about the economy 67 times. For Obama and the DNC, taxes lead the list of issues over the time period. Table 2 provides the Spearman rank correlations for the 21 issues under investigation by the five sources examined. Significant positive correlations were found among all five sources, with the lowest occurring between the two parties and the highest between Obama and the Republican primary candidates. When theanalysisfocusedonthetop-sevenissues,however,therankingswerenotconsistentacrosssources. 6 The sources appeared to agree on which issues were important broadly, but they had different priorities on what was most important. Before analyzing cross-correlations among the measures for each of the top-seven issues, we examined the daily frequencies for trends, specifically looking for linear and quadratic relationships that time with each series. Trends were detected in newspaper indices for the issues of economy (linear function with R 2 =.056, p <.01), employment (linear function with R 2 =.059, p <.01), energy (quadratic Journal of Computer-Mediated Communication 20 (2015) International Communication Association 369

8 Table 1 Average Daily Frequencies by Topic and Source Between 26 December 2011 and 3 May 2012 Newspaper Index Barack Obama Tweets DNC Tweets Republican Candidate Tweet Index RNC Tweets Budget Mean (SD) (27.55) 0.20 (0.53) 0.33 (0.99) 1.44 (3.16) 0.92 (1.74) Total Issue Rank Economy Mean (SD) (26.56) 0.32 (0.84) 0.26 (0.68) 0.73 (1.82) 0.52 (.082) Total Issue Rank Employment Mean (SD) (27.33) 0.39 (0.89) 0.75 (1.78) 0.85 (1.41) 0.92 (1.46) Total Issue Rank Energy Mean(SD) (32.04) 0.73 (1.73) 0.30 (1.17) 3.58 (5.54) 5.59 (8.64) Total Issue Rank Foreign Policy Mean (SD) (48.82) 0.22 (0.57) 0.39 (1.18) 0.71 (1.11) 0.18 (0.54) Total Issue Rank Healthcare Mean (SD) (24.89) 0.65 (1.77) 0.79 (1.84) 0.93 (2.43) 1.18 (2.61) Total Issue Rank Taxes Mean (SD) (62.82) 0.95 (2.44) 1.01 (2.31) 1.04 (2.05) 0.82 (2.50) Total Issue Rank Note: The top-seven issues were identified by aggregating issue mentions across media. The top-seven issues collectively constitute 36,389 issue mentions of the total 54,853 issue mentions in the study. 370 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

9 Table 2 Rank-Order Correlations of Issue Priorities Between Sources Rank-Order Correlations for all 21 Issues Newspaper Index Obama Tweets DNC Tweets Republican Candidate Tweets RNC Tweets Newspaper Index Obama Tweets DNC Tweets Republican Candidate Tweets RNC Tweets Rank-Order Correlations for Top-seven Issues Newspaper Index Obama Tweets DNC Tweets Republican Candidate Tweets RNC Tweets Newspaper Index n.s. n.s. n.s Obama Tweets n.s n.s n.s. DNC Tweets n.s. n.s n.s. n.s. Republican Candidate Tweets n.s n.s n.s. RNC Tweets n.s. n.s. n.s p <.05 p <.001 Journal of Computer-Mediated Communication 20 (2015) International Communication Association 371

10 function with R 2 =.094, p <.01), and health care (quadratic function with R 2 =.180, p <.001). For the Republican candidate tweet indices, trends were found for economy (quadratic function with R 2 =.107, p <.001), employment (quadratic function with R 2 =.101, p <.01), energy (quadratic function with R 2 =.315, p <.001), and foreign policy (linear function with R 2 =.031, p <.05). For Barack Obama s tweet indices, trends were found for budget (linear function with R 2 =.032, p <.05), energy (quadratic function with R 2 =.052, p <.05), health care (quadratic function with R 2 =.059, p <.05), and taxes (linear function with R 2 =.065, p <.01). For RNC tweets, trends were found on energy (quadratic function with R 2 =.208, p <.001), health care (quadratic function with R 2 =.059, p <.05), and taxes (quadratic function with R 2 =.059, p <.05). For DNC tweets, a trend was found on employment (quadratic function with R 2 =.101, p <.01). To prevent problems with autocorrelation, the trends were then removed from these variables. 7 H1 and H2 focus on the relationship between newspaper coverage and tweets. We examined these relationships for the top-seven issues. Leads shown in Table 3 indicate that newspapers predicted issue emphasis in Twitter a given number of days prior to the contemporary frequencies. Lags indicate that tweets predicted newspaper mentions a given number of days prior to the frequencies in newspapers. For example, the 1-day lead on taxes for Republican candidate tweets indicates that newspaper frequencies the day before were correlated with the current Republican candidate tweets. In addition, a Lag 0 means that there was a same-day association between two variables. Lags shown in Table 3 indicate that tweets predicted newspaper mentions on six of the top-seven issues, offering support for H1. The Republican candidate tweets, for example, had predictive value when it came to employment, energy, foreign policy, health care, and taxes. On the economy, only the DNC tweets had a significant lag. Contemporaneous associations were found for the budget, economy, employment, energy, foreign policy, health care, and taxes, although this varied for the different tweet sources. As shown in Table 3, newspapers influenced Twitter feeds of the candidates and parties, supporting H2. Leads of newspaper mentions predicted tweets by the Republican candidates, the RNC, Obama, and the DNC on the budget, employment, health care, and taxes. Newspaper mentions also predicted tweets when it came to the economy for Obama and the DNC, and energy for the Republican candidates, the RNC, and Obama. On foreign policy, influence was apparent on tweet frequencies by the Republican candidates and Obama. When it came to the issue of the budget, newspaper mentions lead Twitter feeds but not the reverse across all four tweet sources. That newspapers lead candidates on an issue and not the reverse already suggests that RQ1 should be answered in the affirmative for newspapers.the only issue for which influence occurred solely with Twitter feeds was energy for the DNC. The number and strength of lags and leads suggests that Twitter feeds did not overpower the influence of newspapers. Looking at those issues for which mutual influence was found, leads outnumbered lags overall and the strength of influence was often greater among newspapers, suggesting newspapers did most of the leading. On those issues that were top priorities for the two camps, this relationship was especially pronounced. For example, with the issue of energy, leads for newspapers ranged in magnitude from r =.18 to r =.46. For Twitter, the range was r =.18 to r =.38. The same can be demonstrated for taxes where the strength of newspaper influence (r =.18 to r =.63) was above that of Twitter (r =.19 to r =.21). Moving to H3A and H3B, the results do not suggest that candidates from either party or the parties themselves (1) emphasized owned issues or (2) set the newspaper agenda to a greater extent with owned issues, undermining both hypotheses. First, Spearman s rank correlations already suggest the use of issues was similar across Twitter feeds. Some issues owned by either party,such as social welfare for Democrats and national defense for Republicans, did not rank high on their lists. Further, among the seven issues made salient within the dataset, both parties made high use of issues traditionally owned 372 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

11 Table 3 Significant Cross-Correlations Between Newspaper Index to Candidate and Party Tweets Budget Economy Employment Energy Foreign Policy Health care Taxes Republican Candidate Tweets Lag 0: Lead 1: Lead 5: RNC Tweets Lag 0: Lead 1: Barack Obama Tweets DNC Tweets Lag 0: Lead 1: Lag 2: Lead 1: Lag 6: Lag 5: Lag 1: Lead 1: Lead 7: Lag 6: Lag 5: Lag 4: Lag 1: Lag 0: Lead 1: Lead 2: Lead 5: Lead 6: Lead 7: Lag 7: Lag 6: Lag 5: Lag 4: Lag 0: Lead 1: Lead 2: Lead 7: Lead 1: Lead 1: Lead 1: Lag 6: Lag 3: Lag 0: Lead 1: Lag 4: Lag 0: Lead 1: Lag 1: Lag 0: Lead 1: Lag 5: Lead 1: Lead 4: Lead 1: Lead 6: Lag 2: Lag 1: Lag 0: Lead 1: Lead 4: Lag 5: Lag 0: Lead 2: Lead 3: Lead 4: Lead 7: Lag 2: Lag 0: Lead 1: Lead 2: Lead 4: Lead 7: Lag 6: Lag 0: Lag 5: Lag 0: Lead 4: Lag 2: Lag 1: Lag 0: Lead 1: Lag 1: Lag 0: Lead 1: Lead 7: Lag 1: Lag 0: Lead 1: Lead 2: Lag 0: Lead 1: Lead 2: Lead 6: Lead 7: Note: Leads shown in Table 3 indicate that newspapers predicted issue emphasis in Twitter a given number of days prior to the contemporary frequencies. Lags indicate that tweets predicted newspaper mentions a given number of days prior to the frequencies in newspapers. A lag of 0 indicates that suchinfluences were contemporaneous. Journal of Computer-Mediated Communication 20 (2015) International Communication Association 373

12 by the other party. Though Republicans saw a greater number of lags, Democrats also emphasized and exhibitedinfluenceontaxes,atraditionallyrepublican- owned issue.energy,atopicoftenlinkedtothe environment and Democrats (Petrocik, 1996) was the focus of Republicans and an issue on which they demonstrated greater influence than Democrats. On employment, influence appeared on both sides, perhapsduetorepublicanstoutinglaggingemploymentgrowthandtheobamaadministrationtouting job growth. Both sides also exhibited influence on health care. Discussion With the advent of new channels of political communication comes the need to re-evaluate the agenda-setting power of the news media (Bennett & Iyengar, 2009; McCombs, 2004). This study looks at how candidates and political parties on Twitter may influence or be influenced by national newspapers. Time series analysis shows Twitter and traditional news media appear to have a symbiotic relationship that varies in intensity and duration depending on the issues being analyzed. The reciprocal relationship between political Twitter feeds and news media is similar not only to previous intermedia agenda-setting studies (Dunn, 2009; Ku et al., 2003; Lancendorfer & Lee, 2010; Sweetser et al., 2008; Walgrave & Van Aelst, 2006), but also a growing line of research examining the relationship between social and traditional media (Groshek & Groshek, 2010; Metzgar & Maruggi, 2009; Sayre et al., 2010). It suggests traditional media and Twitter feeds of politicians, campaigns, and parties are involved in a source cycle. Contemporaneous relationships and minimal time lags also suggest that agenda setting is expedient in the online arena (Denham, 2010; Sweetser et al., 2008). Twitter s influence on newspaper coverage of six of the seven issues supports the notion that, during election time, politicians and their parties,throughtheir energy and dollars, can influence the media and the public (Dalton, Beck, Huckfeldt, & Koetzle, 1998). Social media can be used as a tool by politicians and the public to communicate an agenda that, in turn, shapes the media agenda(metzgar& Maruggi, 2008; Neuman et al., 2014; Sayre et al., 2010). Our findings align with those of issue convergence across camps (Sigelman & Buell, 2004). Positive correlations between issue ranks in news coverage and Twitter feeds suggest newspapers, candidates, and parties placed similar emphasis on issues. Newspapers lead on the budget when it came to all tweet sources, but influenced and were influenced by posts from candidates and parties on both sides of the aisle on most issues. OntheissueofenergywheretherewasreciprocalinfluencebytheRepublicancandidates,theRNC, and Obama, it is possible that, rather than an owned issue, it may be one Republicans were trying to co-opt. Republicans may have been establishing a lease on a non-owned issue based on perceived incumbent inadequacies (Petrocik, 1996; Petrocik et al., 2004). Energy is often associated with environmental concerns such as carbon emissions and renewable resources. The sheer amount of energy references and the findings from a qualitative examination of the energy category, however, suggests Republicans took a different approach. Much of the talk within Republican Twitter feeds was centered on words such as Keystone, pipeline, and gas prices, as well as the botched funding of Solyndra through the American Recovery and Reinvestment Act. In other words, their motivation was economic, a trend common among both Republicans and national news outlets that is likely to arise during economic downturns and times of high gas prices (Nisbet, Maibach, & Leiserowitz, 2011). Past research also suggests that issues of energy security and environmental policy are often potentially at loggerheads (Peterson & Rose, 2006, p. 620). Although the DNC Twitter feed was not affected by the newspaper agenda on energy in 2012, it seems Republicans were trying to take control of the category of energy through a focus on economic security. 374 Journal of Computer-Mediated Communication 20 (2015) International Communication Association

13 Have traditional media completely lost their agenda-setting power? Our study suggests that the answer is no. Candidates are still going to look to the media for legitimacy, just as social media users are going to look to the media for information on unobtrusive issues. Traditional media demonstrated leads on the top-seven issues analyzed in this study. Along with an effort to co-opt the energy issue, RepublicansonTwittermayalsohaveemphasizedcoverageofPresidentObama srejectingapipeline permit application in January of 2012 (Eilperin & Mufson, 2012). On health care, Republicans, the RNC, Obama, and the DNC all looked to the media for legitimacy. Republicans may have done so because they were trespassing on a Democratic-owned issue. Democrats may have done so because it was an issue on which they felt vulnerable a result of tensions over Obamacare. Politicians still rely on traditional media for legitimacy on contested or controversial issues, a trend that will likely persist into the future. Limitations and Future Directions Duetodifferenttimelagsandthepossibleinfluencesofothercampaignmedia,weareuncertainthat candidate Twitter feeds directly influenced traditional media. If Twitter feeds reflect the agenda of other campaign media (e.g. advertisements or websites), this influence may not be Twitter specific. We also know that offline events strongly influence Twitter content during the election season (Aragon, Kappler, Kaltenbrunner, Laniado, & Volkovich, 2013; Bruns & Burgess, 2011). Such events may also be responsible for some of the shared variance in the Twitter-newspaper relationship. As stated previously, what time series analyses suggest is that a nonrandom relationship exists. We cannot exclude the possibility that Twitter feeds are indeed directly influencing the media agenda, especially with journalists increasingly using Twitter feeds as a source and specifically following politicians (Parmelee, 2013b). We also cannot generalize our findings to all politicians or all elections (boyd & Crawford, 2012; Parks, 2014). Future research should examine whether content in social media mirrors that of traditional campaign media. Research is currently underway to determine the extent to which candidate Twitter agendas and those of advertisement actually align. Past research has found differential effects for various campaign media, for example ads and press releases (Kiousis, Kim, McDevitt, & Ostrowski, 2009). Twitter dynamics, including frequency, interactivity, and low cost, present campaigns with an animal quite different from other candidate media (Graham et al., 2013; Jungherr, 2014). It is, therefore, quite possible that the content of such media do not align as Twitter is reactive and influenced by outside occurrences. Thoughwehavenoreasontosuggestthatdifferentmediawouldbefocusingondifferentaspectsof salient issues, future studies should also break down the broad categories used in the current study to discover the extent of second-level agenda-building effects. Future election research should continue to examine use of various campaign platforms by politicians and parties, how such content relates to external events, as well as the use of social media by journalists in order to tease out the various relationships involved in the agenda-building process. As they hone their social media skills and continue to utilize social media as a new tool in the campaign arsenal, and as social media audiences continue to grow and change, we may find that Twitter and other socialnetworkingsitesplayauniqueroleinpoliticalcampaigns.themechanismsbehindtheuseof social media are also missing from the current study. Future research should examine the extent to which thesefactorsareinfluencedbythegrowingroleofmobiledevicesandthemobilenatureoftwitteruse. Notes 1 We thank our reviewers for pointing out that the line between media agenda building and intermedia agenda-setting is a difficult line to draw in social media. While we choose to take an Journal of Computer-Mediated Communication 20 (2015) International Communication Association 375

14 intermedia agenda-setting approach, we agree with previous scholars that altered conceptions of what constitutes a news source in the social media world blur these lines (see, for example, Sayre, Bode, Shah, Wilcox, & Shah, 2010; Groshek & Groshek, 2013). 2 Of the articles discarded, The New York Times had 93, The Wall Street Journal had 182, USA Today had 15, The Los Angeles Times had 93, and The Washington Post had Issues included: Abortion, affirmative action, agriculture, banking, budget, campaign finance reform, crime, economy, education, employment, energy, environment, ethics, foreign policy/trade, guns, health care, immigration, military/defense, taxes, terrorism, and welfare. 4 For newspapers issue references, reliability statistics for the issues used in analysis were as follows: Budget, α =.86; Economy, α =.88; Employment, α =.90; Energy, α =.97; Foreign Policy, α =.96; Health care, α = 1; Taxes, α =.93. For Twitter issue references, reliability statistics for the issues used in analysis were as follows: Budget, α =.92; Economy, α =.93; Employment, α = 1; Energy, α =.83; Foreign Policy, α =.96; Health care, α = 1; Taxes, α = We narrowed the number of issues analyzed for the cross-correlation analysis to seven. Other issues were analyzed but not included in this article. Issues that were not mentioned frequently did not yield enough data to provide the variance required for statistical analyses. For example, there were 2,366 campaign finance reform mentions in the newspapers but only 10 mentions across the Republican candidates (five from Fred Karger, four divided between Newt Gingrich s two feeds, and one from TeamBachmann). Obama mentioned the campaign finance reform topic only five times over the 130 days analyzed. Consequently, we were able to include campaign finance in our rank-order correlations of 21 issues but were not able to use the data for time series analysis specifically. 6 It should be noted that correlations among only the top-seven issues (rather than all 21) supported the incumbent party bonus and even suggested the priorities of the newspaper agenda were at odds with that of the Republican National Committee. 7 When the R 2 statistics were similar in size, we chose the less complex model (i.e., linear over quadratic) as the detrending tool. References Adams, A., & McCorkindale, T. (2013). Dialogue and transparency: A content analysis of how the 2012 presidential candidates used Twitter. Public Relations Review, 39, doi: /j.pubrev Aparaschivei, P. A. (2011). The use of new media in electoral campaigns: Analysis of the use of blogs, Facebook, Twitter, and YouTube in the 2009 Romanian presidential campaign. Journal of Media Research, 2, Bennett, W. L., & Iyengar, S. (2009). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58, doi: /j x Benoit, W. L., & Hansen, G. J. (2004). Issue ownership in primary and general presidential debates. Argumentation and Advocacy, 40, boyd, d., & Crawford, K. (2012). Critical questions for big data. Information, Communication & Society, 15, doi: / X boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, doi: /j x Boyle, T. P. (2001). Intermedia agenda setting in the 1996 presidential election. Journalism & Mass Communication Quarterly, 78, Journal of Computer-Mediated Communication 20 (2015) International Communication Association

15 Bruns, A., & Burgess, J. E. (2011). #ausvotes: How Twitter covered the2010australianfederal election. Communication, Politics and Culture, 44, Conway, B. A., Kenski, K., & Wang, D. (2013). Twitter use by the presidential primary candidates during the 2012 campaign. American Behavioral Scientist, 57, doi: / Dalton, R., Beck, P., Huckfeldt, R., & Koetzle, W. (1998). A test of the media-centered agenda-setting: Newspaper content and public interests in a presidential election. Political Communication, 15, doi: / Damore, D. F. (2004). The dynamics of issues ownership in presidential campaigns. Political Research Quarterly, 57, doi: / Dang-Xuan, L., Stieglitz, S., Wladarsch, J., & Neuberger, C. (2013). An investigation of influentials and the role of sentiment in political communication on Twitter during election periods. Information, Communication & Society, 16, doi: / X Denham, B. E. (2010). Toward conceptual consistency in studies of agenda-building processes: A scholarly review. The Review of Communication, 10, doi: / Dunn, S. W. (2009). Candidate and media agenda setting in the 2005 Virginia gubernatorial election. Journal of Communication, 59, doi: /j x Eilperin, J., &Mufson, S. (2012,January 18). Obama administration rejects Keystone XL pipeline. The Washington Post. Online edition. Retrieved from Farhi, P. (2009). The Twitter explosion. American Journalism Review. Retrieved from org/article.asp?id=4756 Golbeck, J., Grimes, J. M., & Rogers, A. (2010). Twitter use by the U.S. Congress. Journal of the American Society for Information Science and Technology, 61, doi: /asi.2134 Graham, T., Broersma, M., Hazelhoff, K., & van t Haar, G. (2013). Between broadcasting political messages and interacting with voters. Information, Communication & Society, 16, doi: / X Grant, W. J., Moon, B., & Grant, J. B. (2010). Digital dialogue? Australian politicians use of the social network tool Twitter. Australian Journal of Political Science, 45, doi: / Groshek, J., & Groshek, M. C. (2010). Agenda trending: Reciprocity and the predictive capacity of social networking sites in intermedia agenda setting across topics over time. Media and Communication, 1, Gueorguieva, V. (2008). Voters, MySpace, and YouTube: The impact of alternative communication channels on the 2006 election cycle and beyond. Social Science Computer Review, 26, doi: / Hayes,D. (2008).Partyreputation, journalistic expectations: How issue ownership influences election news. Political Communication, 25, doi: / Johnson, T. J., & Perlmutter, D. D. (2010). Introduction: The Facebook election. Mass Communication and Society, 13, doi: / Jungherr, A. (2014). The logic of political coverage on Twitter: Temporal dynamics and content. Journal of Communication, 64, doi: /jcom Khandaroo, S. T. (2011, July 20). Six GOP candidates to hold Twitter debate. Christian Science Monitor. Online edition. Retrieved from Six-GOP-candidates-to-hold-Twitter-debate.-Here-s-how-to-see-it?cmpid=addthis_ #. TxSZayGi7ek. . Journal of Computer-Mediated Communication 20 (2015) International Communication Association 377