Reassessing Twitter’s Agenda-Building Power

Abstract

This study examined the relationship between elite news media agendas and campaign agendas during the 2016 presidential primary season. Computer-assisted content analysis was used to assess issue emphasis within Twitter feeds of U.S. presidential primary candidates and their campaigns as well as the nation’s top newspapers. The relationship between the overall Twitter agenda and that of newspapers, as well as the influence of front-runners Clinton, Cruz, Sanders, and Trump, was investigated using time series analysis. Aggregate and candidate-specific findings reveal some reciprocal relationships, but overall greater influence of newspapers on the Twitter agenda was detected. Findings suggest that Twitter has the potential to break free from and influence traditional media gatekeeping.

Keywords

Twitter primaries intermedia agenda setting agenda building social media

Twitter regularly made headlines during the 2016 U.S. presidential election. From Donald Trump’s “tweetstorms” (Barbaro, Haberman, & Rappeport, 2016) to Hillary Clinton’s throwback photos (North, 2016), anecdotal evidence suggests elite news media followed candidate Twitter use. Previous tests of agenda-building theory suggest as much (Conway, Kenski, & Wang, 2015; Parmelee, 2013a). Despite the eccentricities of the 2016 election cycle, it is fertile ground for assessing the impact of social media agendas on elite news media agendas. Such replication is crucial as we contemplate the general impact of social media on political communication and elections (Towner & Dulio, 2012) and specifically on agenda-setting theory (Bennett & Iyengar, 2008).

The ramifications of new media for the political communication environment have been considered by scholars. Today, social media adoption is no longer at issue. “Politicians know that they have to be present in these [social media] venues” (Kenski & Conway, 2016, p. 203). The 2012 election saw practiced social media use by all presidential campaigns (Stromer-Galley, 2014). At this level of saturation, it is important to study the strategies used to exploit digital media on behalf of politicians and parties (Bimber, 2014; Kreiss, 2014). Campaign members are aware of the impact social media messaging has on the flow of information (Jamieson, 2013), something that can ultimately result in agenda-setting power (Kreiss, 2014).

Research not only suggests that a relationship exists between traditional and new media but also that this relationship may be reciprocal (Conway, Kenski, & Wang, 2015; Groshek & Groshek, 2013; Jacobson, 2013; Meraz, 2009; Sayre, Bode, Shah, Wilcox, & Shah, 2010; Wallsten, 2007). In a previous study, we examined issue emphasis on Twitter during the 2012 presidential primaries and found a symbiotic relationship between the two, with newspapers exhibiting greater influence than tweets (Conway, Kenski, & Wang, 2015). We replicated these findings in a recent chapter in which we again explored the Twitter-to-newspaper relationship in the 2016 presidential primaries (Conway-Silva, Filer, Kenski, & Tsetsi, 2017). The aforementioned studies focused on the issue ownership hypothesis—the extent to which issue emphasis and influence are driven by party-issue affiliations.¹ To add nuance to these findings, here we examined the aggregate relationship between the nation’s top newspapers and tweets by the 2016 presidential nominees and their campaigns as well as the power of front-runners in forecasting the press agenda. Did Twitter’s predictive power increase in 2016? To what extent do front-runners leverage the power of Twitter in the agenda-building process? Findings reveal reciprocal relationships, with the greatest predictive power lying with newspapers. Among the front-runners, Clinton appears to have harnessed Twitter’s potential to emphasize national political issues and predict the issue agenda of national newspapers to a greater extent than did other candidates.

Intermedia Agenda Setting and Social Media

The impact of campaign media on the news agenda is frequently studied under the banners of agenda building and intermedia agenda setting (Weaver & Choi, 2014). The original agenda-setting hypothesis suggests the news media agenda is adopted by the public (McCombs & Valenzuela, 2014). Agenda building and intermedia agenda setting focus on who/what sets the news media agenda. Generally, agenda building focuses on the macrolevel influences that determine the spread of issues among the three pillars of political communication: political elites, mass media, and the public (Lang & Lang, 1981). Underneath this label lies intermedia agenda setting (Denham, 2010), which focuses on the transference of issue salience across media (McCombs, 2004). Information subsidies disseminated by political campaigns and parties, including advertisements, websites, and blogs, fall under this heading, as do social media. Social networking sites (SNSs) can thus impact media agenda-building processes.

Social media have offered politicians a growing audience of voters and journalists since the 2008 election. As of the 2016 general election, 62% of Americans received news from SNSs (Gottfried & Shearer, 2016). When asked about information gathering on the election in the last week, 44% said they did so on SNSs (Gottfried, Barthel, Shearer, & Mitchell, 2016). Facebook remains the most popular SNS overall. However, of those adults who do use Twitter (about 24% of the Internet users), most do so for news and current events, and more than 60% of the content they see focuses on politics either “a lot” or “some” (Duggan & Smith, 2016; Greenwood, Perrin, & Duggan, 2016). Beyond voters, journalists also use Twitter for political information. Studies show journalists actively use Twitter to gather leads on breaking news and follow politicians, equating Twitter content with traditional information subsidies such as press releases and ads (Farhi, 2009; Parmelee, 2013b).

Candidates and their campaigns recognize this potential visibility and approach microblogging strategically. Studies across a number of countries indicate heavy emphasis on campaign activities (for an overview, see Jungherr, 2016). One notable finding is the lack of interactivity and the abundance of self-promotion or one-way message dissemination (Adams & McCorkindale, 2013; Golbeck, Grimes, & Rogers, 2010; Graham, Jackson, & Broersma, 2014; Grant, Moon, & Grant, 2010; Stromer-Galley, 2014). These findings, coupled with research showing that message output patterns on Twitter somewhat mirror those of traditional news (Jungherr, 2014), imply that for some users, Twitter serves a reactionary purpose rather than a place for interaction or dialogue with voters.

That politicians choose to use Twitter for one-way broadcasting also has implications for intermedia agenda setting. The goal of this study was to test agenda-building dynamics during the 2016 presidential nomination season, comparing these findings to previous results and documenting the developing role of Twitter in the political communication environment. Previous research on agenda building and agenda setting on Twitter reveals three potential outcomes: (1) top-down dominance of news media, (2) reciprocal relationships, or (3) “bottom-up” messaging in which SNSs lessen or even override the agenda-setting power of traditional media (Ceron, Curini, & Iacus, 2016; Meraz & Papacharissi, 2012). Most evidence aligns with the first two conclusions.

Supporting the top-down conceptualization, one study found that traditional media predicted the Twitter agenda with regard to the mortgage and housing crisis and the Deepwater Horizon/BP oil spill (Vargo, Basilaia, & Shaw, 2015). Further, Ceron and colleagues (2016) found that the agenda of online news on two Italian political debates preceded that of Twitter. In a study of Twitter and mainstream news agendas, Rogstad (2016) also detected greater influence on behalf of mainstream media than Twitter. Such findings echo previous research on blogs. Although bidirectional or reciprocal relationships exist, traditional media still hold greater agenda-setting power (Meraz, 2009; Wallsten, 2007).

Still, there is evidence that the bottom-up conceptualization can and does happen (Kenski & Jamieson, 2016). Social media are often used to push a unique agenda (Metzgar & Maruggi, 2009; Sayre et al., 2010). Overall, newspapers predicted Twitter feeds to a greater extent than the reverse in our previous study. However, Twitter did predict newspaper issue emphasis on certain issues, exhibiting greater power than newspapers within at least one reciprocal relationship (Conway, Kenski, & Wang, 2015). Such findings are attributable to the dynamics of agenda building during the campaign season. Overall, agenda-building research prior to the advent of SNSs suggests that during campaigns, “the media’s impact on candidates’ and parties’ agendas is limited or even absent” (Walgrave & Van Aelst, 2006, p. 96). Candidates and campaigns may capitalize on a unique agenda on Twitter that is then adopted by news media.

Putting forth a unique agenda on Twitter is increasingly likely if political tweets shape the issue agendas of political journalists (Parmelee, 2013a, 2013b) and journalists are intermingling with political elites on Twitter (Chadwick, 2013). A recent study on Trump’s Twitter use by Wells and colleagues (2016) suggests Trump may have successfully influenced news coverage. Their time series analysis incorporated various information subsidies, including press releases, events, and Twitter posts, and revealed that the volume of Trump’s tweets was not predicted by news stories and that his tweets positively predicted the amount of news coverage. Thus, preliminary findings for the 2016 presidential campaign align with the bottom-up hypothesis.

Because most findings suggest reciprocal relationships, with news media exhibiting greater influence than social media, we predicted that:

Hypothesis 1: Reciprocal relationships exist between the issue agendas of newspapers and the issue agendas of Twitter feeds.

Hypothesis 2: The issue agendas of newspapers predict the issue agendas of Twitter feeds to a greater extent than the reverse.

Wells and colleagues found a Trump effect; however, they did not analyze candidates besides Trump who was perhaps more influential in his Twitter messaging than were other candidates. They also focused on the frequency of media use, not issue agendas. Ergo we explore whether Trump, who led the Republican nomination polls from January 2016 to the end of the primary season, predicted the newspaper agenda to a greater extent than Hillary Clinton, who also led her party’s nomination polling during the time frame in question (realclearpolitics.com). For comparison, we also explored the predictive power of Ted Cruz and Bernie Sanders, who were second in the polls during this study’s time frame.

Research Question 1: Did tweets of the front-runners predict the newspaper agenda as much as they were influenced by it?

Method

This study used computer-assisted content analysis to determine issue emphasis within Twitter feeds of U.S. presidential primary candidates and their campaigns as well as the nation’s top newspapers. Time series analysis was then used to determine the strength of relationships between the overall Twitter agendas and those of newspapers and the influence of Republicans Cruz and Trump and Democrats Clinton and Sanders.

Data Set

Newspapers examined in this study were The Wall Street Journal, The New York Times, USA Today, The Los Angeles Times, and The Washington Post. The first four were chosen primarily because they are the largest circulation newspapers in the United States (Pew, 2013). Following Conway, Kenski, and Wang (2015), The Los Angeles Times was also included as representative of the West Coast. The Washington Post was included due to its emphasis on national politics. Articles published from January 1 to June 7 were collected using LexisNexis and ProQuest. The date January 1 was one month prior to the Iowa caucuses. On June 6, Hillary Clinton officially garnered enough delegates to secure the nomination (Chozik & Healy, 2016). Articles were retrieved using the term “president*” in combination with “Republican Primar*,” “Republican Caucus*,” “Republican nomination,” “Democratic Primar*,” “Democratic Caucus*,” “Democratic nomination,” or any of the names of the primary candidates. All news articles and editorials that mentioned any of the candidates were included. If any of the returned articles did not reference one of the nominees, they were discarded, resulting in a total of 6,434 relevant articles, of which The New York Times had 1,775, The Wall Street Journal had 1,623, USA Today had 593, The Los Angeles Times had 818, and The Washington Post had 1,625.²

All tweets posted on the feeds of major party candidates and their campaigns from January 1 to June 7 were included in this project to create a Twitter index (N = 27,845). This Twitter index was then used to establish the extent to which the Twitter agenda of presidential candidates and their campaigns preceded that of the news media. Included feeds are displayed in Table 1. For the purposes of this study, if a candidate also had an individual feed that was not necessarily linked to their campaign (e.g., Chris Christie had both a feed representing his presidential candidacy [@ChrisChristie] and a feed representing his role as governor of New Jersey [@GovChristie]), both were included. In the final data set, candidate and campaign tweets were included up to the day an individual dropped out of the race (e.g., the last tweet included for Ted Cruz occurred on May 3). Twitter feeds were gathered weekly using the statistical program R to query Twitter’s REST application programming interface (API). The REST API allows researchers to “query Twitter’s databases for data corresponding to specific parameters,” including tweets posted by specific users (Jungherr, 2016, p. 82). These data were exported into Excel files, which included the tweet, user name, and date posted.

Table 1.

Twitter Feeds Included in Overall Twitter Index.

Twitter Handle	Total Tweets
Republicans
@CarlyFiorina	192
@ChrisChristie	546
@GovChristie	176
@gov_gilmore	389
@GovMikeHuckabee	360
@JebBush	742
@JohnKasich	2,607
@marcorubio	1,207
@RandPaul	1,248
@RealBenCarson	647
@RealDonaldTrump	2,131
@RickSantorum	128
@SenTedCruz	165
@TeamJohnKasich	627
@TeamMarco	1,828
@TeamSantorum	885
@TeamTedCruz	412
@TedCruz	4,205
Total	18,495
Democrats
@BernieSanders	3,423
@HFA	600
@HillaryClinton	2,663
@MartinOMalley	193
@SenSanders	1,170
@TheBriefing2016	1,301
Total	9,350

Coding

The coding program QDA Miner and its companion program WordStat were used to tally issue frequencies within Twitter feeds and newspapers. While QDA Miner is mainly used for qualitative coding, its quantitative component Wordstat allows researchers to examine the frequency of words and phrases and implement coding rules. For newspapers, the unit of analysis was all news content published in a newspaper that mentioned any of the candidates or the presidential nomination contests on a given day. For Twitter feeds, the unit of analysis was all tweets posted to an individual account on a given day.

Coding employed a combination of automated and manual analysis to ensure that the former was coded as intended by the researcher (Dang-Xuan, Stieglitz, Wladarsch, & Neuberger, 2013).³ First, we created a list of national political issues examined in previous studies (Conway, Kenski, & Wang, 2015; Tedesco, 2001). Based on these topics, thousands of words that appeared in news articles and tweets were analyzed in WordStat to detect additional categories and fully develop previously established categories. This resulted in a total of 27 issues.⁴ Words were included if they appeared 10 times or more. This included hashtags that were placed in the appropriate category based on their original intent. For example, in our data set, #FIXTHEDEBT was placed in the budget category. Replicating the steps of our previous study,

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. (Conway, Kenski, & Wang, 2015, p. 368)

If the placement of a word was unclear, the authors investigated the word in context to determine its use.

Overall, issue emphasis similarities across studies dominate (see Conway, Kenski, & Wang, 2015; Tedesco, 2001). For example, previous studies also included “abortion,” “affirmative action,” “agriculture,” “health care,” “economy,” and “immigration.” Nevertheless, issue emphasis does change across elections depending on historical events and campaign priorities—differences over time lie with narrow, emergent issues specific to a given time period. This election cycle, for example, required that we add “Clinton’s e-mail scandal” and “Zika” to the list. Agenda-setting studies should always include slightly different lists of issues if meant to represent the context they examine.

Analysis

To examine the overall relationship between Twitter feeds and news media, we first identified the top five issues that appeared within each medium: the newspaper index, the Twitter index, and the candidate/campaign feeds of Clinton, Cruz, Sanders, and Trump. This resulted in a total of 13 issues (see Table 2). To test Hypotheses 1 and 2, we focused on the top five issues within the newspaper index and the Twitter index for the sake of parsimony, resulting in a total of eight issues: the economy, education, employment, foreign policy, health care, immigration, national security/terrorism, and taxes. For the candidate-specific analyses, we focused on the top five issues in Trump’s feed, Sanders’ candidate and senator feeds, the combined candidate/campaign feeds for Clinton, and the combined candidate/senator/campaign feeds for Cruz, resulting in 11 issues: affirmative action, banking, education, employment, foreign policy, gun control, health care, immigration, military spending/veterans, national security/terrorism, and social welfare. We recognize that some issues are left out of this analysis but observe that those included are fairly representative of what the public considered to be the top issues in 2016 (Gallup, 2016). It is important to note that case occurrence was low across issues. For example, the issue that occurred the most (employment) within the Twitter index only occurred in 2.65% of tweets. Still, it occurred in 737 cases, a relatively large volume.

Table 2.

Total Case Occurrences by Issue Between January 1, 2016, and June 6, 2016.

Issue		Newspaper Index	Twitter Index	Hillary Clinton	Ted Cruz	Bernie Sanders	Donald Trump
Affirmative action	Total (%)	1,573 (24.49)	418 (1.50)	214 (4.69)	20 (0.42)	126 (2.74)	6 (0.28)
Affirmative action	Rank	7	10	1	12	11	15
Banking	Total (%)	1,447 (22.53)	446 (1.60)	73 (1.60)	10 (0.21)	300 (6.53)	18 (0.84)
Banking	Rank	10	8	11	16	2	6
Economy	Total (%)	2,134 (33.23)	394 (1.41)	62 (1.36)	25 (0.52)	164 (3.57)	12 (0.56)
Economy	Rank	2	11	12	11	9	9
Education	Total (%)	1,976 (30.77)	435 (1.56)	112 (2.45)	16 (0.33)	172 (3.74)	10 (0.47)
Education	Rank	5	9	6	14	5	12
Employment	Total (%)	2,114 (32.92)	737 (2.65)	130 (2.85%)	52 (1.09)	310 (6.75)	34 (1.60)
Employment	Rank	3	1	4	8	1	4
Foreign policy	Total (%)	3,018 (46.99)	669 (2.40)	80 (1.75)	113 (2.36)	120 (2.61)	60 (2.82)
Foreign policy	Rank	1	3	9	2	12	2
Gun control	Total (%)	634 (9.87)	383 (1.38)	173 (3.79)	54 (1.13)	13 (0.28)	10 (0.47)
Gun control	Rank	16	12	2	7	21	11
Health care	Total (%)	1,530 (23.82)	612 (2.20)	142 (3.11)	56 (1.17)	292 (6.38)	15 (0.70)
Health care	Rank	8	4	3	6	3	8
Immigration	Total (%)	2,056 (32.01)	469 (1.68)	125 (2.74)	88 (1.84)	97 (2.11)	47 (2.21)
Immigration	Rank	4	7	5	5	14	3
Military spending	Total (%)	1,451 (22.59)	489 (1.76)	52 (1.14)	89 (1.86)	62 (1.35)	34 (1.60)
Military spending	Rank	9	6	14	4	17	5
National security	Total (%)	1,884 (29.34)	705 (2.53)	88 (1.93)	97 (2.03)	63 (1.37)	64 (3.00)
National security	Rank	6	2	8	3	16	1
Social welfare	Total (%)	931 (14.50)	345 (1.24)	41 (0.90)	3 (0.06)	236 (5.14)	6 (0.28)
Social welfare	Rank	15	13	18	20	4	16
Taxes	Total (%)	1,431 (22.28)	554 (1.99)	93 (2.04)	119 (2.49)	156 (3.40)	17 (0.80)
Taxes	Rank	11	5	7	1	6	7

Note. Newspaper index represents overall issue emphasis in newspapers. Twitter index represents overall issue emphasis in primary candidate/campaign/party Twitter feeds. Percentages represent the percent of case occurrences.

Before examining the relationship between the two media, we first inspected the data for linear and quadratic trends (underlying patterns across time series data). When working with data that include a temporal element, patterns that are associated with time within each series suggest that a variable’s error terms are correlated. The newspaper index, Twitter index, and the tweets of Clinton, Cruz, Sanders, and Trump were then detrended if the linear or quadratic trends were statistically significant, allowing the relationship between the days to be analyzed without concerns of autocorrelation (Romer, 2006). Then, the cross-lagged correlations between the detrended time series were calculated to evaluate the strength of the relationship (i.e., predictive value) among the sources of interest. To test Hypothesis 2, we evaluated the number of lags and leads. This test of relationships should not be equated with causality, but such patterns, especially consistent patterns over time, do suggest a nonrandom relationship (Sayre et al., 2010).

Results

After creating newspaper and Twitter indices, daily frequencies were examined for linear and quadratic trends within each series. Trends were detected in newspaper indices for the economy (quadratic function with R² = .09, p < .01), education (quadratic function with R² = .08, p < .01), employment (quadratic function with R² = .13, p < .01), foreign policy (quadratic function with R² = .10, p < .01), health care (linear function with R² = .09, p < .01), immigration (linear function with R² = .06, p < .01), and national security/terrorism (linear function with R² = .13, p < .01). Trends were detected in Twitter indices for the economy (linear function with R² = .16, p < .01), education (linear function with R² = .23, p < .01), foreign policy (linear function with R² = .09, p < .01), health care (linear function with R² = .19, p < .01), immigration (linear function with R² = .16, p < .01), national security/terrorism (linear function with R² = .18, p < .01), and taxes (linear function with R² = .05, p < .01). These trends were removed prior to analysis.⁵

Hypotheses 1 and 2 focus on the overall relationships between newspapers and Twitter, testing whether the relationships were reciprocal and if one medium exhibited greater influence than did the other. Leads shown in Table 3 indicate that newspapers predicted emphasis on Twitter on six of the eight issues (education, employment, health care, foreign policy, immigration, and national security/terrorism). For example, newspapers led on education 2, 5, and 7 days prior. Still, tweets predicted newspapers 1–7 days prior on four of the eight issues (the economy, employment, foreign policy, and taxes), with reciprocal relationships on employment and foreign policy supporting Hypothesis 1. There were also many contemporaneous relationships, indicating that issues appeared simultaneously in newspapers and tweets. Based on the number of issues on which newspapers led, as well as the number of leads for newspapers (nine) and Twitter (five), overall newspapers had more predictive power than Twitter, supporting Hypothesis 2.

Table 3.

Significant Cross-Correlations Between the Newspaper Index and the Primary Candidate/Campaign Twitter Index.

	Economy	Education	Employment	Foreign Policy	Health Care	Immigration	National Security	Taxes
Twitter index	Lag 1: 0.16 Lag 2: 0.17	Lead 7: 0.21 Lead 5: 0.22 Lead 2: 0.23 Lag 4: −0.15 Lag 7: −0.17	Lead 7: 0.18 Lag 0: 0.34 Lag 1: 0.21	Lead 7: 0.18 Lag 0: 0.23 Lag 1: 0.23	Lead 6: 0.19 Lag 0: 0.23 Lag 4: −0.16	Lead 7: 0.16 Lead 6: 0.24 Lag 0: 0.39	Lead 6: 0.16 Lag 0: 0.22	Lag 0: 0.16 Lag 2: 0.20

Note. All cross-lagged correlations shown here are significant, p < .05. Leads shown indicate that newspapers predicted tweet emphasis 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 contemporary frequencies. Lag 0 indicates a contemporaneous relationship. Negative correlations indicate low incidence on Twitter.

Research Question 1 asked whether the front-runners’ Twitter agendas influenced the newspaper agenda (and vice versa). First, we investigated trends on issues within newspapers that were not included in the previous analysis. Trends were detected for affirmative action (quadratic function with R² = .06, p < .01), banking (quadratic function with R² = .07, p < .01), gun control (quadratic function with R² = .23, p < .01), and social welfare (linear function with R² = .03, p < .05). Trends were detected for Clinton’s tweets on banking (quadratic function with R² = .07, p < .01), employment (linear function with R² = .03, p < .05), gun control (linear function with R² = .08, p < .05), heath care (linear function with R² = .12, p < .01), and social welfare (quadratic function with R² = .04, p < .05). Trends were detected for Cruz on affirmative action (linear function with R² = .05, p < .05), employment (linear function with R² = .25, p < .01), gun control (linear function with R² = .04, p < .05), and military spending (linear function with R² = .12, p < .05). Trends were detected for Sanders’ tweets on banking (linear function with R² = .04, p < .01), education (linear function with R² = .02, p = .05), foreign policy (quadratic function with R² = .04, p < .05), gun control (linear function with R² = .04, p < .05), health care (linear function with R² = .07, p < .01), and immigration (linear function with R² = .02, p = .05). Trends were detected for Trump’s tweets on banking (quadratic function with R² = .04, p < .05) and education (linear function with R² = .03, p < .05). These trends were removed prior to analysis to prevent autocorrelation.

As shown in Table 4, Clinton’s tweets predicted newspaper emphasis anywhere from 2 to 7 days on four of the 11 issues (banking, gun control, immigration, and social welfare)—more than any other candidate. Trump’s tweets predicted newspapers 3 to 6 days prior on three issues (banking, foreign policy, and health care). The tweets of Sanders (affirmative action and gun control) and Cruz (employment and social welfare) predicted newspaper emphasis on two issues each from 3 to 6 days prior. On the other hand, newspapers led Clinton on seven issues, Cruz on four issues, Sanders on five issues, and Trump on four issues anywhere from 1 to 7 days prior. Therefore, newspapers out-predicted all Twitter sources in terms of leads. Some of these relationships were also reciprocal (e.g., banking and gun control for Clinton and health care for Trump). Cruz and Sanders had one issue each on which they predicted newspapers and the reverse did not occur (bottom-up relationships), while Trump had two. But there were 17 instances in which newspaper issue emphasis predicted candidate emphasis and not the reverse (top-down relationships). Contemporaneous relationships were also discovered on several issues.

Table 4.

Significant Cross-Correlations Between the Newspaper Index and the Front-runners.

Issue	Hillary Clinton	Ted Cruz	Bernie Sanders	Donald Trump
Affirmative action	Lead 2: 0.17		Lag 6: 0.21	Lead 3: 0.16 Lead 1: 0.19
Banking	Lead 1: 0.25 Lag 0: 0.34 Lag 7: 0.21			Lag 3: 0.18
Education	Lead 2: 0.15		Lead 6: 0.20 Lead 5: 0.18 Lag 7: −0.19	Lead 5: 0.17 Lead 2: 0.19
Employment	Lead 2: 0.21 Lead 1: 0.20 Lag 0: 0.28	Lead 7: 0.18 Lag 1: 0.18 Lag 3: −0.20 Lag 4: −0.18	Lead 2: 0.18	Lag 2: −0.17
Foreign policy	Lag 0: 0.26		Lead 7: 0.19	Lag 3: 0.21
Gun control	Lead 2: 0.20 Lag 0: 0.24 Lag 6: 0.19	Lead 6: 0.22 Lead 3: 0.22	Lead 3: 0.23 Lag 0: 0.26 Lag 3: 0.16	Lead 3: 0.18 Lead 2: 0.26
Health care	Lead 6: 0.17 Lag 0: 0.21 Lag 3: −0.22 Lag 4: −0.26		Lead 1: 0.19	Lead 1: 0.16 Lag 0: 0.20 Lag 5: 0.16 Lag 6: 0.22
Immigration	Lead 7: 0.24 Lead 2: 0.18 Lead 1: 0.20 Lag 0: 0.36 Lag 6: 0.20	Lead 4: 0.29 Lag 0: 0.18	Lead 1: 0.16
Military spending		Lag 0: 0.27	Lead 6: 0.16
National security		Lead 2: 0.22 Lag 0: 0.27	Lead 1: 0.21	Lag 0: 0.16
Social welfare	Lead 5: −0.23 Lag 0: 0.19 Lag 2: 0.19 Lag 5: 0.16	Lag 7: 0.18	Lead 6: 0.21	Lag 6: −0.17

Discussion

This study builds on previous intermedia agenda-setting work by examining relationships among the issue agendas of the nation’s top newspapers and the Twitter feeds of 2016 presidential candidates and their campaigns. Focusing specifically on national political issues, findings suggest a symbiotic relationship between the two. This was also the case in our study of the 2012 election, as was the finding that newspapers predicted the Twitter agenda to a greater extent than the reverse (Conway, Kenski, & Wang, 2015). Still, candidate Twitter feeds predicted the news agenda on a variety of issues, some of which were not owned by their party (e.g., Clinton on immigration and Trump on affirmative action). This was also the case in 2012. Our candidate-centered analyses further reveal that the agenda-building relationship is likely candidate specific.

Focusing on the overall analysis, there are interesting patterns when we look at the length of leads. Based on the number of leads, newspapers were somewhat successful in predicting issue emphasis on Twitter. They often did so 1 to 2 days prior, but there are also 6 and 7 day leads. We know that newspapers are confined to a certain extent by their publication cycle and that they can set the agenda for other news media, such as television news (Roberts & McCombs, 1994). If this is the case, longer lag times could be related to a buildup of coverage across different news media over time. This finding might also support the argument of scholars who believe the time frame for agenda setting by online media is swifter than that of traditional media (Jacobson, 2013; Lee, Lancendorfer, & Lee, 2005; Roberts, Wanta, & Dzwo, 2002).

That said, large contemporaneous relationships on all issues, which were also highly prevalent in our 2012 analysis, speak to the similarities of Twitter and traditional newspapers. Both are commenting on the issues of the day in a rapid fashion. Such contemporaneous relationships were also prevalent in our analysis of the front-runners, suggesting that candidates and their campaigns are mindful of the event-driven nature of Twitter. We cannot mistake correlations for causation, but findings suggest issues may be transferred from Twitter to other media at a quicker rate than that of newspapers to Twitter. This could be not only due to monitoring and quick response by newspaper journalists to candidate/campaign tweets but also to worldwide events—events on which candidates/campaigns commented on the previous day.

On candidate-specific analyses, compared to others, Clinton may have done a slightly better job using Twitter to emphasize national political issues that were in turn covered by the press agenda. The Trump effect suggested by Wells and colleagues (2016) did not emerge in our examination of issue emphasis. Of course, Trump was criticized for a lack of issue knowledge and vague or false statements on the issues (Kopan, 2016; Politifact, 2017). Across candidates, average case occurrence on the 13 issues in question was higher for Clinton (2.33%) and Sanders (3.54%) than for Cruz (1.19%) and Trump (1.20%). It is possible that Trump’s influence manifested in nonpolicy-related statements, which is beyond the scope of the present study. Future research should investigate what agendas, other than issue agendas, might be important in the Twitter to newspaper relationship.

Issue mentions are theorized to be less frequent during nomination campaigns than in the general election period (Kendall, 2016), but issue emphasis was not overwhelmingly absent on Twitter during the primaries. Across candidates, the percentages of tweets that included at least one issue mention were Trump = 14.45%, Cruz = 15.72%, Clinton = 32.32%, and Sanders = 48.65%. These numbers are not inconsequential, nor are issues in a political campaign. As we contemplate the ramifications of new media for democratic outcomes, issue emphasis (as opposed to entertainment-laden content) begs future research (Van Aelst et al., 2017).

This study does not account for the tone with which candidates were portrayed in news media. The ability to predict the news media agenda does not necessarily mean that the news content was positive. Clinton, for example, “had by far the most negative coverage of any candidate” in 2015 (Patterson, 2016, para. 3). While Clinton operated within existing political structures, helping build news media agendas in terms of issue focus, this tactic did not necessarily result in a major advantage for her campaign. Tweets by Clinton and her campaign were also predicted by newspapers to a greater extent than the other three candidates. Thus, her campaign may have used Twitter in an effort to reframe news media coverage.

Trump’s rhetorical style on social media about issues was also focused in part on reframing news stories and encouraging news media distrust after stories had been published. It is clear looking back on the 2016 election cycle that Trump was heavily invested in being the consummate political outsider, which included attacking news media credibility and breaking free from traditional news outlets as gatekeepers. “Though he regularly bashes the media as dishonest, scum and the ‘absolute worst,’ Donald Trump disproportionately benefited from the Fourth Estate’s coverage” (Gass, 2016, para. 1). Future research should investigate how news coverage and candidate tweets, both issue focused and not, are taken up and interpreted. For example, investigating news coverage of Trump’s tweets about Ted Cruz’s wife Heidi Cruz and other incendiary comments would be enlightening. Future research should also investigate different news media as well as local newspapers covering state primaries.

The largest limitation of this study is that we cannot suggest causality. Time series analysis has a temporal/predictive component, but we are limited in the strength with which we can argue that these relationships are Twitter specific. These findings are based on Twitter and newspaper indices and do not take into account the various other types of campaign media that can influence the press (ads, press releases, rallies, and so on). Of note, our study of candidate tweets and campaign ads in the 2012 presidential campaign detected issue overlap (Conway, Filer, & Kenski, 2015). One should also ask the extent to which Twitter and newspaper agendas are driven by public concern. Arguably, much work is also needed on the impact of social media use by candidates on the public, although excellent research is already being done. Another limitation of this study is the reliance on dictionary work, which may result in false negatives (Guo, Vargo, Pan, Ding, & Ishwar, 2016). Our study employed a more succinct dictionary compared to Guo and colleagues (2016) in their test of dictionary coding versus LDA. We included all relevant words that appeared 10 times or more (see Note 3). Given that dictionary work is more conservative than LDA, our approach is on the conservative side and relationships may be larger than those reported here.

This study has limitations but is an important step in the way of replication. Candidates can and do use Twitter as an agenda-building tool. Three studies now suggest that candidate/campaign Twitter feeds can predict newspaper emphasis and vice versa during American presidential primary campaigns. That the news agenda predicted that of Twitter to a greater extent than the reverse is a positive indicator to those who believe agenda setting exists in the new media environment. Although scholars anticipated that new media would enhance the agenda-building process at the expense of the agenda-setting function of news media (Chaffee & Metzger, 2001), the results of this study, combined with findings from 2012 (Conway, Kenski, & Wang, 2015), suggest that traditional news gatekeepers continue to shape the political campaign environment by setting the agenda.

Footnotes

Declaration of Conflicting Interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Adams

McCorkindale

(2013). Dialogue and transparency: A content analysis of how the 2012 presidential candidates used Twitter. Public Relations Review, 39, 357–359. doi:10.1016/j.pubrev.2013.07.016

Barbaro

Haberman

Rappeport

(2016, 9 30). As America sleeps, Donald Trump seethes on Twitter. New York Times. Retrieved from http://www.nytimes.com/2016/10/01/us/politics/donald-trump-alicia-machado.html

Bennett

W. L.

Iyengar

(2008). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58, 707–731. doi:10.1111/j.1460-2466.2008. 00410.x

Bimber

(2014). Digital media in the Obama campaigns of 2008 and 2012: Adaptation to the personalized political communication environment. Journal of Information Technology & Politics, 11, 130–150. doi:10.1080/19331681.2014.895691

Ceron

Curini

Iacus

S. M.

(2016). First- and second-level agenda building in the Twittersphere: An application to the Italian political debate. Journal of Information Technology & Politics, 13, 159–174. doi:10.1080/19331681.2016.1160266

Chadwick

(2013). The hybrid media system: Politics and power. Oxford, England: Oxford University Press.

Chaffee

S. H.

Metzger

M. J.

(2001). The end of mass communication? Mass Communication and Society, 4, 365–379. doi:10.1207/ S15327825MCS0404_3

Chozik

Healy

(2016, 6 6). Hillary Clinton has clinched Democratic nomination, survey reports. The New York Times. Retrieved from http://www.nytimes.com/2016/06/07/us/politics/hillary-clinton-presidential-race.html

Conway

B. A.

Filer

C. R.

Kenski

(2015). Campaign media congruence: Issues in 2012 presidential election tweets and ads. Paper presented at the American Political Science Association, San Francisco, CA.

10.

Conway-Silva

B. A.

Filer

C. R.

Kenski

Tsetsi

(2017). Issue emphasis and agenda building on twitter during the 2016 presidential primary season. In Baumgartner

Towner

(Eds.), The Internet and the 2016 Campaign (pp. 241–262). Lantham, MD: Rowman & Littlefield.

11.

Conway

B. A.

Kenski

Wang

(2015). The rise of Twitter in the political campaign: Searching for intermedia agenda-setting effects in the presidential primary. Journal of Computer Mediated Communication, 20, 363–380. doi:10.1111/jcc4.12124

12.

Dang-Xuan

Stieglitz

Wladarsch

Neuberger

(2013). An investigation of influentials and the role of sentiment in political communication on Twitter during election periods. Information, Communication & Society, 16, 795–825. doi:10.1080/1369118X.2013.783608

13.

Denham

B. E.

(2010). Toward conceptual consistency in studies of agenda-building processes: A scholarly review. The Review of Communication, 10, 306–323. doi:10.1080/15358593.2010.502593

14.

Duggan

Smith

(2016, 10 25). The political environment on social media. Pew. Retrieved from http://www.pewinternet.org/2016/10/25/the-political-environment-on-social-media/

15.

Farhi

(2009). The Twitter explosion. American Journalism Review. Retrieved from http://www.ajr.org/Article.asp?id=4756

16.

Gallup. (2016). Most important problem. Gallup. Retrieved from http://www.gallup.com/poll/1675/most-important-problem.aspx

17.

Gass

(2016, 6 14). Study: Trump boosted, Clinton hurt by primary coverage. Politico.com. Retrieved from http://www.politico.com/story/2016/06/media-study-trump-helped-clinton-hurt-224300

18.

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. doi:10.1002/asi.21344

19.

Gottfried

Barthel

Shearer

Mitchell

(2016). The 2016 presidential campaign—A news event that’s hard to miss. Pew Research Center. Retrieved from http://www.journalism.org/2016/02/04/the-2016-presidential-campaign-a-news-event-thats-hard-to-miss/

20.

Gottfried

Shearer

(2016). News use across social media platforms 2016. Pew Research Center. Retrieved from http://www.journalism.org/2016/05/26/news-use-across-social-media-platforms-2016/

21.

Graham

Jackson

Broersma

(2014). New platform, old habits? Candidates’ use of Twitter during the 2010 British and Dutch general election campaigns. New Media & Society, 18, 765–783. doi:10.1177/ 1461444814546728

22.

Grant

W. J.

Moon

Grant

J. B.

(2010). Digital dialogue? Australian politicians’ use of the social network tool Twitter. Australian Journal of Political Science, 45, 579–604. doi:10.1080/10361146.2010.517176

23.

Greenwood

Perrin

Duggan

(2016). Social media update 2016. Pew Research Center. Retrieved from http://www.pewinternet.org/2016/11/11/social-media-update-2016/

24.

Groshek

M. C.

(2013). Agenda trending: Reciprocity and the predictive capacity of social networking sites in intermedia agenda setting across topics over time. Media and Communication, 1, 15–27. doi:10.12924/mac2013.01010015

25.

Guo

Vargo

C. J.

Pan

Ding

Ishwar

(2016). Big social data analytics in journalism and mass communication: Comparing dictionary-based text analysis and unsupervised topic modeling. Journalism & Mass Communication Quarterly, 93, 332–359. doi:10.1177/1077699016639231

26.

Jacobson

(2013). Does audience participation on Facebook influence the news agenda? A case study of the Rachel Maddow Show. Journal of Broadcasting & Electronic Media, 57, 338–355. doi:10.1080/08838151.2013.816706

27.

Jamieson

K. H.

(2013). Electing the President, 2012: The insiders’ view. Philadelphia: University of Pennsylvania Press.

28.

Jungherr

(2014). The logic of political coverage on Twitter: Temporal dynamics and content. Journal of Communication, 64, 239–259. doi:10.1111/jcom.12087

29.

Jungherr

(2016). Twitter use in election campaigns: A systematic literature review. Journal of Information Technology & Politics, 13, 72–91. doi:10.1080/19331681.2015.1132401

30.

Kendall

K. E.

(2016). Presidential primaries and general election campaigns: A comparison. In Benoit

W. L.

(Ed.), Praeger handbook of political campaigning in the United States (pp. 31–60). Santa Barbara, CA: Praeger.

31.

Kenski

Conway

B. A.

(2016). Social media and elections. In Benoit

W. L.

(Ed.), Praeger handbook of political campaigning in the United States (pp. 191–208). Santa Barbara, CA: Praeger.

32.

Kenski

Jamieson

K. H.

(2016). Political communication: Looking ahead. In Kenski

Jamieson

K. H.

(Eds.), The Oxford handbook of political communication [online version]. Oxford, England: Oxford University Press. doi:10.1093/oxfordhb/9780199793471.013.86

33.

Kopan

(2016, 6 10). McConnell: “Obvious” Trump doesn’t know issues. CNN. Retrieved from http://www.cnn.com/2016/06/10/politics/mitch-mcconnell-donald-trump-issues/

34.

Kreiss

(2014). Seizing the moment: The presidential campaigns’ use of Twitter during the 2012 electoral cycle. New Media & Society, 18, 1473–1490. doi:10.1177/1461444814562445

35.

Lang

E. G.

Lang

(1981). Watergate: An exploration of the agenda-building process. Mass Communication Review Yearbook, 2, 447–468.

36.

Lee

Lancendorfer

Lee

(2005). Agenda-setting and the Internet: The intermedia influence of Internet bulletin boards on newspaper coverage of the 2000 general election in South Korea. Asian Journal of Communication, 15, 57–71. doi:10.1080/01292980420003 29793

37.

McCombs

(2004). Setting the agenda: The mass media and public opinion. Malden, MA: Blackwell.

38.

McCombs

Valenzuela

(2014). Agenda-setting theory: The frontier research questions. In Kenski

Jamieson

K. H.

(Eds.), The Oxford handbook of political communication [online version]. Oxford, England: Oxford University Press. doi:10.1093/oxfordhb/9780199793471.013.48

39.

Meraz

(2009). Is there an elite hold? Traditional media to social media agenda setting influence in blog networks. Journal of Computer-Mediated Communication, 14, 682–707. doi:10.1111/j.1083-6101.2009.01458.x

40.

Meraz

Papacharissi

(2012). Networked gatekeeping and networked framing on #Egypt. The International Journal of Press/Politics, 18, 138–166. doi:10.1177/1940161212474472

41.

Metzgar

Maruggi

(2009). Social media and the 2008 U.S. Presidential Election. Journal of New Communications Research, 4, 141–165.

42.

North

(2016, 9 27). The power of Hillary Clinton’s throwback photos. The New York Times. Retrieved from http://takingnote.blogs.nytimes.com/2016/09/27/the-power-of-hillary-clintons-throwback-photos/

43.

Parmelee

J. H.

(2013a). The agenda-building function of political tweets. New Media & Society, 16, 434–450. doi:10.1177/1461444813487955

44.

Parmelee

J. H.

(2013b). Political journalists and Twitter: Influences on norms and practices. Journal of Media Practice, 14, 291–305. doi:10.1386/jmpr.14.4.291_1

45.

Patterson

T. E.

(2016, 6 13). Pre-primary news coverage of the 2016 presidential race: Trump’s rise, Sanders’ emergence, Clinton’s struggle. Harvard Kennedy School’s Shorenstein Center on Media, Politics and Public Policy. Retrieved from https://shorensteincenter.org/pre-primary-news-coverage-2016-trump-clinton-sanders/

46.

Pew. (2013). Top 25 daily newspapers. Pew’s Project for Excellence in Journalism. Retrieved from http://www.stateofthemedia.org/2013/newspapers-stabilizing-but-still-threatened/22-top-25-daily-newspapers/

47.

Politifact. (2017). All false statements involving Donald Trump. Politifact. Retrieved from http://www.politifact.com/personalities/donald-trump/statements/byruling/false/?page=2

48.

Roberts

McCombs

(1994). Agenda setting and political advertising: Origins of the news agenda. Political Communication, 11, 249–262. doi:10.1080/10584609.1994.9963030

49.

Roberts

Wanta

Dzwo

(2002). Agenda-setting and issue salience online. Communication Research, 29, 452–465.

50.

Rogstad

(2016). Is Twitter just rehashing? Intermedia agenda setting between Twitter and mainstream media. Journal of Information Technology & Politics, 13, 142–158. doi:10.1080/19331681.2016.1160263

51.

Romer

(2006). Time series models. In Romer

Kenski

Winneg

Adasiewicz

Jamieson

K. H.

(Eds.), Capturing campaign dynamics 2000 & 2004 (pp. 165–243). Philadelphia: University of Pennsylvania Press.

52.

Sayre

Bode

Shah

Wilcox

Shah

(2010). Agenda setting in a digital age. Policy & Internet, 2, 7–32. doi:10.2202/1944-2866.1040

53.

Stromer-Galley

(2014). Presidential campaigning in the Internet age. New York, NY: Oxford.

54.

Tedesco

J. C.

(2001). Issue and strategy agenda-setting in the 2000 presidential primaries. American Behavioral Scientist, 44, 2048–2067. doi:10.1177/00027640121958483

55.

Towner

T. L.

Dulio

D. A.

(2012). New media and political marketing in the United States: 2012 and beyond. Journal of Political Marketing, 11, 95–119. doi:10.1080/15377857.2012.642748

56.

Van Aelst

Strombäck

Aalberg

Esser

de Vresse

Matthes

Hopmann

Salgado

Hubé

Stępińska

Papathanassopoulos

Berganza

Legnante

Reinemann

Shaefer

Stanyer

(2017). Political communication in a high choice media environment: A challenge for democracy? Annals of the International Communication Association, 41, 3–27. doi: 10.1080/23808985.2017.1288551

57.

Vargo

C. J.

Basilaia

Shaw

D. L.

(2015). Event versus issue: Twitter reflections of major news—A case study. Communication and Information Technologies Annual—Studies in Media and Communications, 9, 215–239. doi:10.1108/S2050-206020150000009009

58.

Walgrave

Van Aelst

(2006). The contingency of the mass media’s political agenda setting power: Toward a preliminary theory. Journal of Communication, 56, 88–109. doi:10.1111/j.1460-2466.2006.00005.x

59.

Wallsten

(2007). Agenda setting and the blogosphere: An analysis of the relationship between mainstream media and political blogs. Review of Policy Research, 24, 567–587. doi:10.1111/j.1541-1338.2007.00300.x

60.

Weaver

D. H.

Choi

(2014). The Media agenda: Who (or what) sets it? In Kenski

Jamieson

K. H.

(Eds.), The Oxford handbook of political communication [online version]. Oxford, England: Oxford University Press. doi:10.1093/oxfordhb/9780199793471.013.37

61.

Wells

Shah

D. V.

Pevehouse

J. C.

Yang

Pelled

Boehm

Schmidt

J. L.

(2016). How Trump drove coverage to the nomination: Hybrid media campaigning. Political Communication, 33, 669–676. doi:10.1080/10584609.2016.1224416