A permanent campaign? Tweeting differences among Members of Congress between the campaign and routine periods before and after the 2016 US elections
Vidar Vasko 10876383 Master Thesis
Graduate School of Communication
Political Communication - Research Master dr. D.C. Trilling
Abstract
Agenda-setting research has traditionally postulated that mainstream media dominates the ability to influence the public agenda. With Social Networking Sites like Twitter increasing in
popularity, the possibilities for politicians to reverse this relationship abound. Twitter use by politicians has been extensively researched, but differences between campaign and routine periods have received less attention. In employing a large dataset of approximately 200,000
tweets by Members of Congress during and after the 2016 US elections, this study seeks to bridge this knowledge gap. In the routine period, politicians are found to focus more on hard
news, put more emphasis on domestic than foreign content on both national and state level, and publish more tweets, whereas in the campaign period positive and negative sentiment is higher. The results suggest that politicians do indeed exhibit differences in tweeting behavior
between campaign and routine periods, although some of them unexpected, and garner little support for the notion of a permanent campaign on Twitter.
Acknowledgements
This work was carried out on the Dutch national e-infrastructure with the support of SURF Cooperative. I would like to thank Damian Trilling for guidance, friends and family for support, and Thelma for encouraging purrs.
Introduction
The opening sentence of McCombs and Shaw’s (1972) seminal study on agenda setting reads: “In our day, more than ever before, candidates go before the people through the mass media rather than in person.” Almost half a century later, this relationship seems to be undergoing a reversal. As noted by Ott (2017), the agenda-setting function in politics previously dominated by television seems to be increasingly performed by Twitter. Hillary Clinton announced her presidential candidacy by tweeting: “I’m running for president. Everyday Americans need a champion, and I want to be that champion. – H” (Clinton, 12 April 2015). Similarly, Donald Trump uses Twitter to communicate with the public,
bypassing – and often harshly criticizing – traditional media: “The failing @nytimes is truly one of the worst newspapers. They knowingly write lies and never call to fact check. Really bad people!” (Trump, 13 March 2016). These examples seem to support the notion of a reversed logic of agenda setting. If this is indeed the case, there is a need to understand how tweeting politicians, as the new agenda setters, affect the public debate.
Influence of social media on political communication is an urgent matter in the study of political campaigns (Gueorguieva, 2008; Towner & Dulio, 2012). Extensive research has been conducted on the use of Twitter during election campaigns (for a literature review, see Jungherr, 2016). Other studies have focused on Twitter use outside campaign periods (e.g. Aharony, 2012; Peng, Liu, Wu & Shixia, 2016). There is, however, no research investigating the differences between politicians’ use of Twitter during election periods vis-à-vis routine periods. The question is whether an election campaign can be looked upon as a normal period, or does it present extraordinary phenomena when it comes to politicians’ Twitter activities? This study contributes to extant research by examining how agenda-setting and campaign versus routine period theories hold up in the modern political climate in which social networking has taken on a heightened role. With the potential for politicians on Twitter to
indirectly reach the public through setting the agenda of mainstream media, and through both direct and indirect communication with their followers (Vaccari & Valeriani, 2013), the SNS (Social Networking Site) constitutes an attractive outreach medium for politicians. Twitter is one of the most popular social media platforms: On May 15, 2017, Twitter ranked as the 11th most popular site in the world, and the 8th most popular site in the U.S. (Alexa.com, 2017). With an incredibly large potential audience and global reach, politicians’ tweets inevitably have implications for the public debate, and democracy at large. The present study seeks to remedy the above-mentioned knowledge gap by analyzing the differences in Twitter use among politicians during an election campaign period and a routine period. To examine this, I analyze a large dataset of US Members of Congress’ (MoCs) Twitter data from 137 days prior to and after the 2016 U.S. elections.
Theoretical Background Agenda setting in the age of social networking
A preamble to the inception of agenda-setting theory can be found in Cohen’s (1963) remark that the press “may not be successful much of the time in telling people what to think, but is stunningly successful in telling its readers what to think about.” (p. 13, emphasis in original). Explicitly introduced in the subsequent landmark study by McCombs and Shaw (1972), agenda-setting theory suggests that the media exerts considerable influence on
determining the important issues of the day. This notion is currently being dissected – even by founder McCombs himself – in search of potential new paradigms (Bennett & Iyengar, 2008; McCombs & Guo, 2014; Conway et al, 2015). The term reverse agenda setting, originally coined by McCombs (2004), suggests that the public, previously considered only a minor influence in agenda setting, could have the power to influence news media. This influence may be facilitated by the use of Internet technologies (Kim & Lee, 2006). Already for a decade, social media are considered indispensable for political campaigns (Perlmutter, 2008).
The interaction between the public and politician can be enhanced through an SNS (Towner & Dulio, 2012), considered unique because of its potential to allow users to transparently display their connection networks (boyd & Ellison, 2007).
Twitter contrasts many other SNSs in that its users commonly use public profiles that obviate the need to bidirectionally confirm connections (boyd & Ellison, 2007). The SNS, characterized by its 140 characters maximum tweets, is increasingly being used as a political platform to both disseminate information, demonstrated by Hillary Clinton announcing her presidency on Twitter (HillaryClinton, 12 April 2015), and facilitate dialogue between politicians and the public, as illustrated by the Obama administration’s Twitter Townhall (Office of the Press Secretary, 2011). Regarding agenda setting, Ott (2017) argues that televised news now follow the lead of Twitter. In contrast, Groshek & Groshek (2013) found Twitter to be more likely to follow traditional media than the other way around. Yet other studies point to a more complicated agenda-setting relationship between Twitter and
traditional media, characterized not by a simple one-way pattern but instead by a dynamic and complex interaction (Russel Neuman et al., 2014; Conway et al., 2015).
Campaign versus routine periods
While agenda-setting research has traditionally not focused on the political cycle, there are reasons to suspect that the relationship between politicians and the media is different between campaign and routine periods. Other streams of literature (see e.g. Van Aelst & De Swert, 2009) have analyzed such differences, and their possible contribution to the for this study important agenda-setting effects will be discussed in this section. First, however, a justification is needed for why we can look at media behavior literature when making hypotheses about politicians’ tweeting behavior.
In Blumler’s (1999) third age of political communication, politicians understand that carefully attending to communication through the media is an indispensable part of their
work. This professionalization of political communication leads politicians to increasingly adapt their behavior to media logic - the way media cover politics, including the presentation and organization of news material (Altheide, 2013). Politicians can thus be expected to conform to such logic when posting on Twitter, in order to maximize the chances of mainstream media picking up the content of their tweets for publication. This enables us to look for possible hypotheses on politicians’ Twitter behavior in literature on media behavior during election and routine periods in general.
The notion of “the permanent campaign” (Blumenthal, 1980) entails that politicians need to think about their daily endeavors in relation to media coverage as if the election campaign period never ends. It has been argued that the Clinton, W. Bush and Obama
administrations followed the theory of the permanent campaign, believing that this “provides the highest probability of political success” (Goidel, 2011, p. 138). If such a state of affairs is indeed in place, then political publicity should follow a uniform logic throughout both campaign and routine periods. Different forms of cyberdemocracy could lead to an
intensification of the permanent campaign (Ornstein, 2000). Most political news studies have traditionally focused on the election campaign period (Van Aelst & De Swert, 2009). This has generally been rationalized by the election campaign’s significance for democracy (Swanson & Mancini, 1996). Less political communication literature has investigated routine periods, and fewer studies yet exist on the differences between campaign and routine periods. There are, however, some exceptions, and they seem to point toward that there are indeed
differences between campaign and routine periods. Van Aelst and De Swert (2009) listed three reasons for why campaign and routine periods differ. First, the behavior of parties and candidates changes – they become more active in their pursuit of media attention. Second, rules and practices on balance and fairness (Semetko, 1996) influence the way media report during election times. Twitter, however, is not subject to the professional rules and practices
of regular journalism, and we should thus not expect any general tweeting pattern of fairness and balance. Third, ordinary citizens seem to become more interested in politics during campaign periods. In their analysis of Flemish television news broadcasts, two of Van Aelst and De Swert’s (2009) findings are relevant for this study. First, during election periods there are more hard than soft news. Economy, finance and foreign and domestic politics are usually considered to be hard news; news about celebrities, crime, royal families, service, scandals and sports are commonly referred to as soft news (Reinemann et al., 2012). Second, during election periods there are more domestic than foreign news. In the case of US Congress elections, in which politicians can only gain votes from citizens in their own state, such differences in focus on domestic and foreign news could also be extrapolated to a regional effect. Three hypotheses can be formulated:
H1: During the campaign period politicians will tweet more about hard news than during the routine period.
H2: During the campaign period politicians’ tweets will mention their own country more and foreign countries less than during the routine period.
H3: During the campaign period politicians will tweet more about their own state and less about other states than during the routine period.
In US presidential debates, approximately four out of every ten candidate statements are attacks (Airne & Benoit, 2004). The extensive journalistic coverage of negative campaign messages incentivizes candidates to employ a negative sentiment for more media attention (Hansen & Pedersen, 2008). Negativity aside, it is possible that also positivity can result in
increased attention during campaigns (Geer, 2009). It is important to note that positivity and negativity are not necessarily correlated, and that tweets may exhibit neither of the
sentiments. In election times, in which candidates fight for attention in order to gain votes, both positively and negatively emotional messages thus have potential to attract attention, leading to the following hypotheses:
H4: During the campaign period politicians’ tweets will exhibit a higher level of positive sentiment than during the routine period.
H5: During the campaign period politicians’ tweets will exhibit a higher level of negative sentiment than during the routine period.
Lassen & Brown (2011) found that the percentage of MoCs with Twitter accounts sharply increased slightly after the elections in 2008. Now, all MoCs have Twitter accounts, and we can assume that politicians and their media teams are more accustomed to using SNSs for outreach purposes. Because of the competitive nature of elections, it is expected that politicians more actively seek media attention (Van Aelst & De Swert, 2009) during
campaigns than during routine periods. In addition, in line with extant research on Twitter use in election campaigns we expect an increase in tweeting intensity as the campaign approaches election day (e.g. Ahmed & Skoric, 2014; Bruns & Highfield, 2013; Graham et al., 2014). There is also some evidence for an increase in positive sentiment as the campaign nears election day (Aragón et al., 2013), leading to tentatively hypothesizing a similar effect in the present study.
H6: During the campaign period politicians will publish more tweets per day than during the routine period.
H7: At the end of the election campaign, there will be an increase in number of tweets published per politician per day.
H8: At the end of the election campaign, there will be an increase in the level of positive sentiment exhibited in the politicians’ tweets.
Candidates from highly competitive districts or states have previously been found to have the most solid online presence (Esterling et al., 2005). In their investigation of Twitter use among U.S. members of Congress, Lassen and Brown (2011) found no indication of any effect of being in a competitive district on high activity on Twitter or on having a high “Twitter Influence” – actual audience size, number of posts reacted to by other Twitter users, and “network score”, a measure which increases if a person’s followers also are influential users. Evans et al. (2014), similarly, did not find House of Representatives candidates from competitive districts to tweet more than others, but did find them to significantly oftener display an “Attack” style of tweeting. This latter effect is only expected to be present during the campaign period, when candidates are in competition with each other.
H9: Competitive districts or states will display a higher increase in negative sentiment during the campaign period as opposed to the routine period than less competitive districts or states.
Members of the House of Representatives are elected every two years. Senators, however, serve terms of six years. Senators’ seats are divided up into three different classes,
so that approximately one third of the Senate is up for election every two years. In 2016, all Class 3 Senators were up for election. It is reasonable to assume that the rest of the Senate is less affected in their Twitter activities during the campaign – they do not have to gain the trust of voters to the same extent as the ones up for election. The following hypothesis is
constructed:
H10: The effects in hypotheses 1-8 will be higher for members of the House of Representatives and Class 3 Senators than for Class 1 and 2 Senators.
Data and method
The case under study is the 2016 US House of Representatives and Class 3 Senate elections, 137 days prior to and after election day November 8 2016. The full period is thus from June 25 2016 to and including March 25 2017. To be able to compare differences between campaigning and non-campaigning politicians, Class 1 and 2 Senators were included in the sample. The data only contained MoCs who were present in Congress during the entire period. C-SPAN’s compilation of MoCs’ Twitter accounts (C-SPAN, 2016) was used to obtain a list of Twitter accounts. The list was manually checked to make sure all relevant politicians were included, resulting in the addition of several accounts. Any MoC with less than 15 tweets in either the period prior to or after election day November 8 was excluded, meaning that the final dataset only contained politicians who tweeted at least once in most of the weeks. Any account created after June 25 2016 was excluded. Since MoCs have a varying numbers of Twitter accounts (generally 1-3), the list was reduced to only one per person. Accounts typically belonged to one of three categories: campaign, official and personal. Many accounts had no information on type of account. In presentation these often resembled the official account type. Since the purpose of this study was to investigate differences between campaign and routine periods, all explicit campaign accounts were excluded. All MoCs with a
personal account also had an official account. Thus, the final dataset consisted of either explicitly official accounts or accounts with no information on type, except for Lacy Clay, whose only account was explicitly described as campaign account.
The tweets were collected with the help of a Python script written by Trilling (2015). Since Twitter’s API only allows for the retrieval of the last 3200 tweets per account, data collection was performed on two different occasions (November 25 2016 and May 15 2017). All retweets were removed to assure that the analysis only considered content endorsed by the politician. This is illustrated by the fact that some account profiles even feature a disclaimer stating that not all retweets are endorsements. The final sample consisted of 209,915 tweets by 432 individuals. In the analysis, the total N is 864, since each politician’s values in the campaign and routine period are represented by an individual case.
Preprocessing. Python scripts were written to analyze the tweets. For all hypotheses,
tweets were prepared by making all characters lowercase and removing punctuation. In the case of sentiment analysis, stopword removal (Boumans and Trilling, 2016) was attempted, but finally left unperformed since the removed stopwords were deemed to contribute to a more realistic evaluation of sentiment. Stemming – cutting off parts of a word to recognize common stems, like thank and thanks – was initially tried with the external module NLTK SnowballStemmer (Bird et al., 2009). This, however, led to the altering of words that should have been left untouched, distorting their meaning, and was finally not performed.
Hard versus soft news. A list of the 2000 most common words used by all politicians
throughout both periods was compiled, out of which 112 words suggesting a topic of hard news according to Reinemann et al. (2012) were selected (Table A1). All instances of hard news words in politicians’ tweets were counted. To avoid frequent tweeters skewing the data, the count of hard news mentions was divided by the total number of tweets per politician, resulting in a ratio ranging from .09 to 1.62 (M = .56, SD = .17).
Domestic versus foreign emphasis. For this hypothesis, a list of all countries and
nationalities in the world was used to measure in how many tweets they were mentioned at least once. The acronyms UK and US, with potential punctuation, were also included. America and American were also included as indicators of the US, with the requirement that they were preceded by neither South, North nor Latin. Having the counts of tweets with domestic mentions and foreign mentions for each politician, a ratio between the two variables was calculated. Since some politicians had either no domestic or no foreign mentions, to avoid division by zero both variables were increased with 1 before computing the ratio. The variable had a minimum of .09 and maximum of 17.00 (M = 1.72, SD = 1.62). With strong skew (3.68) and kurtosis (22.36), the variable was transformed to its natural logarithm before testing (min = -2.36, max = 2.83, M = .24, SD = .77, skew = .04, kurtosis = .23).
Home versus other states emphasis. To investigate the emphasis MoCs placed on their
home as opposed to other US states, an identical script as in the previous hypothesis was used, but with states instead of countries. Here, no acronyms but only full state names were used, as manual checks of the data gave no sign of prevalence of the former. The preparation for this hypothesis is equal to that of the domestic versus foreign mentions hypothesis, with the ratio between domestic and foreign states mentions ranging from .06 to 37.50 (M = 3.26, SD = 4.17). The variable exhibited strong skew (3.64) and kurtosis (18.63) and was transformed to its natural logarithm prior to testing (min = 2.83, max = 3.62, M = .63, SD = 1.07, skew = -.08, kurtosis = -.02).
Sentiment. For the hypotheses on sentiment, the external module SentiStrength
(Thelwall et al., 2012) was used. The SentiStrength algorithm rates a text on its positivity and negativity, enabling a comparison of the sentiment in different politicians’ texts. The
sentiment score was divided by number of tweets per politician. The negative sentiment score was reversed for ease of interpretation. Positive (min = .20, max = 2.87, M = 1.78, SD = .29),
negative (min = .10, max = 2.42, M = 1.52, SD = .24) and average (min = -.94, max = 1.48, M = .26, SD = .39) sentiment were recorded.
Tweet frequency. A simple count of tweets was performed to investigate the
hypothesis on tweet frequency in the campaign versus routine period, which ranged from 16 to 1496 (M = 243.96, SD = 216.22).
End of the campaign (H7 & H8). Binary variables Last week (1 = last week, 0 = rest
of period) and Last month (1 = last month, 0 = rest of period) were used as independent variables to distinguish between politicians’ scores on the relevant dependent variables. For the hypothesis on tweet frequency at the end of the campaign (H7), the tweet count was divided by number of days in the corresponding period. For the hypothesis on positive sentiment at the end of the campaign (H8), some of the politicians published no tweets in either the last week or last month, and were thus excluded from the respective analysis. See Table B1 for a summary of minimum, maximum, mean and standard deviation of all end of the campaign variables.
Competitiveness. To construct a measure for competitiveness, scores from four indices
(Cook, RealClearPolitics, Rothenberg and Sabato) from October 31 to November 7 were merged to create an absolute scale of competitiveness ranging from 0 (all indices classifying a seat as “safe” for either Democrats or Republicans) to 12 (all indices classifying a seat as “tossup”), with M = .71 and SD = 2.33 for an N of 742 campaigning politicians
(non-campaigning politicians were not interesting for this analysis). This was an improvement from previous studies that used the losing major-party presidential candidate’s share of votes in the previous election in (Lassen & Brown, 2011) and scores from a single index (Evans et al., 2014).
Campaigning and non-campaigning politicians. There were 371 campaigning
(Class 1 and 2 Senators) politicians. Independent variable Campaigning was coded 1 for campaigning and 0 for non-campaigning politicians.
Control variables. Previous studies have found age, gender, party affiliation and
chamber (Lassen & Brown, 2010; Hemphill et al., 2013) to aid explain Twitter use by Members of Congress. These variables were thus included as control variables in the analyses. Lassen and Brown (2010) found no significant effect of members’ tenure in
Congress on probability of using Twitter. With this study having other dependent variables, it remained an interesting explanatory variable, and Years in office was included as control variable. Years in office was calculated by subtracting the year a politician assumed office from 2017, and ranged from 2 to 52 (M = 10.61, SD = 8.98). The variable exhibited
significant skew (1.44), and had its natural logarithm computed, resulting in a new variable with min = .69, max = 3.95, M = 2.02, SD = .84 and skew = .05. Age was calculated by
subtracting year of birth from 2017, and ranged from 33 to 88 (M = 60.19, SD = 10.73). There were 84 women and 348 men in the sample, and the variable Gender was coded with 1 = Female and 0 = Male. There were 88 Senators and 344 members of the House of
Representatives in the sample, and the binary variable Chamber was coded in the following manner: 1 = House of Representatives, 0 = Senate. There were 195 Democrats, 235
Republicans and 2 Independents in the dataset. Democrat was used as baseline and dummy variables were included for Independent and Republican.
Analysis. The variable Period (1 = campaign, 0 = routine) was used to distinguish
between the campaign and routine periods. To test the hypotheses, regression models were estimated. Some dependent variables showed significantly non-normal, overdispersed distributions. In those cases, for count variables, negative binomial regressions were used, whereas for interval variables simple linear regressions were run on their natural logarithm. In these analyses, incidence rate ratios (IRRs) are given in the corresponding table in Appendix
B. IRRs are simple to interpret: For example, an IRR of 1.20 means that a one-unit increase in the independent variable leads to a 20% increase in the dependent variable, whereas an IRR of .80 means that a one-unit increase in the independent variable leads to an 80% decrease in the dependent variable.
Results
H1: Hard versus soft news. A simple linear regression was calculated to predict hard
news mentions over number of tweets based on Period and control variables. A significant regression equation was found, F(7, 856) = 15.92, p < .001, with an R2 of .12 (Model 1, Table C1). In the campaign period, politicians’ ratio of hard mentions over number of tweets was .10 lower than in the routine period. This is the opposite effect of what was predicted, and Hypothesis 1 is thus rejected. For each 1 year increase in a politician’s age, the ratio of hard mentions over number of tweets increased by .003, and being in the House of Representatives increased the ratio of hard mentions of number of tweets by .03.
Including Campaigning and the Period x Campaign interaction variable barely increased model fit, F(9, 854) = 12.51, p < .001, R2 = .12. The results (Model 2, Table C1) showed that whether politicians were campaigning or not had no significant effect on tweeting hard news.
H2: Domestic versus foreign emphasis. To predict the ratio of domestic and foreign
mentions based on Period and the control variables, a simple linear regression was calculated (Model 1, Table C2). A significant regression equation was found, F(7, 856) = 2.73, p = .008, with an R2 of .02. In the campaign period, politicians’ ratio of domestic over foreign mentions was 86% of that in the routine period. This is the opposite of what was predicted, and
Hypothesis 2 is thus rejected. Members of the House of Representatives scored 17% higher on the dependent variable than Senators. Independents had a 97% higher score on the ratio
between domestic and foreign mentions than Democrats, but the difference was only significant at the .10-level.
Including Campaigning and the Period x Campaigning interaction variable resulted in a similar model fit, F(9, 854) = 2.25, p = .017, R2 = .02. The results (Model 2, Table C2) showed that whether politicians were campaigning or not had no significant effect on domestic versus foreign emphasis.
H3: Domestic versus foreign state emphasis. To predict the ratio of domestic and
foreign states mentions based on Period and the control variables, a simple linear regression was calculated (Model 1, Table C3), resulting in a significant regression equation, F(7, 856) = 8.94, p < .001, R2 = .07. Period had no significant effect on domestic over foreign states mentions, rejecting Hypothesis 3. Members of the House of Representatives scored 54% lower on the dependent variable than Senators, and women had a 37% higher ratio of domestic over foreign states mentions than men. Republicans scored 25% higher on the dependent variable than Democrats.
A model including Campaigning and the Period x Campaigning interaction variable had a similar model fit, F(9, 854) = 6.95, p < .001, R2 = .07. The results (Model 2, Table C3) showed that whether politicians were campaigning or not had no significant effect on
domestic versus foreign states emphasis.
H4 & H5: Sentiment. To predict the differences in sentiment between the campaign
and routine periods based on Period and control variables, three simple linear regressions were calculated. For positive sentiment, a significant regression equation was found (Model 1, Table C4), F(7, 856) = 10.76, p < .001, R2 = .08. Politicians had a .06 higher value of positive sentiment during the campaign period than the routine period, supporting Hypothesis 4. Members of the House of Representatives exhibited a .06 higher value of positive sentiment than Senators, and Republicans had a .14 higher value of positive sentiment than Democrats.
While women had a .05 higher score on positive sentiment than men, the difference was not statistically significant.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 854) = 8.37, p < .001, R2 = .08. The results (Model 2, Table C4) showed that whether politicians were campaigning or not had no significant effect on positive sentiment.
For negative sentiment, a significant regression equation was found, F(7, 856) = 26.21, p < .001, R2 = .18 (Model 1, Table C5). Politicians exhibited a .03 higher value of negative sentiment during the campaign than routine period, supporting hypothesis 5. Women had a .05 higher negative sentiment value than men, and Republicans a .17 lower value of negative sentiment than Democrats. A one percent increase in years in office resulted in a .03 increase in negative sentiment.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 854) = 21.21, p < .001, R2 = .18. The results (Model 2, Table C5) showed that whether politicians were campaigning or not did not significantly affect negative sentiment.
With both higher positive and negative sentiment in the campaign period, it was interesting to check whether there was more of either sentiment. To investigate this, a simple linear regression based on Period and the control variables was calculated for average
sentiment. A significant regression equation was found, F(7, 856) = 28.65, p < .001, R2 = .19 (Model 1, Table C6). The results suggested that the level of average sentiment stayed the same during both the campaign and routine period. Republicans had a .31 higher average sentiment value was than Democrats, and one percent increase in years in office resulted in a .05 decrease in level of average sentiment.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 854) = 22.56, p < .001, R2 = .19. The results (Model 2, Table C6) showed that whether politicians were campaigning or not had no significant effect on average sentiment.
H6: Tweet frequency. With the number of tweets being strongly overdispersed, a
negative binomial regression was calculated to predict number of tweets per politician based on Period and control variables. The model explained the variance in the data significantly better than the null hypothesis, χ2 (7) = 134.13, p < .001 (Model 1, Table C7). In the campaign
period, politicians published 17% less tweets than in the routine period. This is the opposite effect as suggested, thus rejecting Hypothesis 6. Women published 32% more tweets than men, members of the House of Representatives 64% less than Senators, and Republicans 22% less tweets than Democrats. One year increase in age resulted in 1% less tweets.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, χ2 (9) = 134.00, p < .001. The results (Model 2, Table C7) showed that whether
politicians were campaigning or not did not significantly affect tweet frequency.
H7: Tweet frequency end of the campaign. Binary variables Last week and Last month
were used, together with control variables, to predict the differences in tweet frequency between the end and the rest of the campaign. For the last week, a significant simple linear regression equation was found, F(7, 856) = 16.14, p < .001, R2 = .12 (Model 1, Table C8). In the last week politicians published an average of .35 tweets less per day than during the rest of the campaign period.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 854) = 14.56, p < .001, R2 = .12. The results (Model 2, Table C8) showed that whether politicians were campaigning or not did not significantly affect whether they tweeted more in the last week of the campaign than in the rest of the period.
For the last month, the regression equation was also significant, F(7, 856) = .133, p < .001, R2 = .13 (Model 1, Table C9). In the last month politicians published an average of .29 tweets less per day than during the rest of the campaign period.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 854) = 14.56, p < .001, R2 = .13. The results (Model 2, Table C9) showed that whether politicians were campaigning or not did not significantly affect whether they tweeted more in the last month of the campaign than in the rest of the period.
In both the last week and last month regressions, the opposite effect as the one
hypothesized was found. Politicians tweeted less in the end of the campaign than in the rest of the period, rejecting Hypothesis 7.
Figure 1 displays the average number of tweets per day for all politicians during the entire period. The most striking feature is that weekdays have a visibly higher average of tweets (Monday: 1.64, Tuesday: 2.28, Wednesday: 2.56, Thursday: 2.43, Friday: 2.02) than weekends (Saturday: .84, Sunday: .67). There are also spikes around specific dates.
Furthermore, the level of tweeting is visibly higher in the period after the 115th congress convened for the first time on January 3 2017 than during the campaign.
H8: Sentiment end of the campaign. To predict the differences in positive sentiment
between the end of the campaign and the rest of the campaign period, simple linear
regressions were calculated based on control variables and either the Last week or Last month dummy variable. For the last week, a significant regression equation was found, F(7, 807) = 3.40, p < .001, R2 = .03 (Model 1, Table C10). In the last week politicians exhibited a .11 higher level of positive sentiment than in the rest of the campaign period.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 805) = 2.68, p < .001, R2 = .03. The results (Model 2, Table C10) showed that whether politicians were campaigning or not did not significantly affect whether they
exhibited more positive sentiment in the last week of the campaign than in the rest of the period.
Figure 1. Average number of tweets per day for Members of Congress.
.03 (Model 1, Table C11). In the last month politicians posted tweets with a .07 higher level of positive sentiment than during the rest of the period.
Including Campaigning and the Period x Campaigning interaction resulted in a similar model fit, F(9, 852) = 3.26, p < .001, R2 = .03. The results (Model 2, Table C11) showed that whether politicians were campaigning or not did not significantly affect whether they
exhibited more positive sentiment in the last month than in the rest of the campaign.
The results showed that politicians indeed posted tweets with a higher level of positive sentiment in the end of the campaign than in the rest of the period, supporting Hypothesis 8.
As shown in the above analyses, neither Campaign nor the interaction Period x Campaigning had any effect on any of the dependent variables in hypotheses 1 through 8, firmly rejecting Hypothesis 10.
H9: Competitive districts. To predict the differences in negative sentiment between the
campaign and routine period based on Period, Competitiveness and the interaction between Period and Competitiveness, a simple linear regression was calculated. A significant regression equation was found, F(6, 735) = 27.85, p < .001, R2 = .19 (Table C12). The
coefficient of the interaction between Period and Competitiveness, however, was
insignificant. Competitive districts did not display a higher increase in negative sentiment during the campaign period as opposed to the routine period than less competitive districts, leading to the rejection of Hypothesis 9.
Discussion and Conclusion
Political communication theory has historically taken for granted the monopoly of large media outlets as the most influential agenda-setters. While McCombs (2004) himself has noted the potential of citizens reversing the agenda-setting relationship, Internet technologies undoubtedly stimulate this on an unprecedented level. In a time with quick development and growing popularity of social networking sites (SNSs) like Twitter, politicians increasingly encounter possibilities to influence this relationship and set the agenda themselves, directly by engaging with citizens and indirectly by media covering their online activities. It is in such a context that this study investigated the differences between politicians’ campaign and routine activities before and after the 2016 US elections. The results do little to reinforce the idea of a permanent campaign (Blumenthal, 1980). Ornstein & Mann (2000) posited that cyberdemocracy could lead to an intensification of the permanent campaign, but it seems like on Twitter differences abound. The effects, however, are not always in the directions posited in the hypotheses (for an overview, see Table 1). Topics, sentiment and tweeting frequency lay on different levels throughout the two periods.
Members of Congress tweeted less about hard news during the campaign than routine period, contrasting previous research on Flemish news broadcasts by Van Aelst and de Swert (2009). In an election campaign, there might be more to gain for a politician from focusing more on other topics, perhaps connecting with voters on an emotional level, than discussing hard news. Supporting this conclusion, both the negative and positive sentiment expressed in politicians’ tweets were higher during the campaign than routine period. This is in line with previous
Table 1
Overview of hypotheses
Confirmed? Note
H1 During the campaign period politicians will
tweet more about hard news than during the routine period.
-
Opposite effect
H2 During the campaign period politicians’ tweets
will mention their own country more and foreign countries less than during the routine period.
- Opposite effect
H3 During the campaign period politicians will
tweet more about their own state and less about other states than during the routine period
- Opposite effect
H4 During the campaign period politicians’ tweets
will exhibit a higher level of positive sentiment than during the routine period.
+
H5 During the campaign period politicians’ tweets
will exhibit a higher level of negative sentiment than during the routine period.
+
H6 During the campaign period politicians will
publish more tweets per day than during the routine period.
- Opposite effect
H7 At the end of the election campaign, there will be an increase in number of tweets published per
politician per day.
- Opposite effect H8 At the end of the election campaign, there will be
an increase in the level of positive sentiment exhibited in the politicians’ tweets.
+ H9 Competitive districts or states will display higher
increase in negative sentiment during the campaign period as opposed to the routine period than non-competitive districts or states.
- No effect
H10 The effects in hypotheses 1-8 will be higher for House of Representatives and Class 3 Senators
than for Class 1 and 2 Senators.
- No effect
research finding that higher levels of sentiment resulted in increased media attention.
Furthermore, in line with Aragón et al. (2013), the end of the campaign saw an increase in the level of positive sentiment expressed by tweeting politicians. Taken together, these findings fit the idea of tabloidization (see e.g. Esser, 1999; Sparks, 2000), pushing aside hard news topics in favor of a more popularized and personalized style. In communicating in such a
manner, especially with the potential to through SNSs like Twitter bypass traditional media and directly reach voters, politicians may hope to garner more support than by talking about for example hard news in an objective manner. As Skovsgaard (2014) argues, however, such a focus may not just be a threat to but also carry benefits for democracy, in that it could increase citizens’ attention to politics. Future research will surely employ such theories in conducting research on Twitter.
Politicians’ ratio between both domestic and foreign mentions and domestic and foreign states mentions were higher in the routine than campaign period. These results are somewhat surprising, as it would be expected that politicians mainly focus their attention on areas where their potential voters reside. Another surprising result is that politicians published more tweets per day during the routine than campaign period. Lassen & Brown (2011) found a similar result in that the percentage of Members of Congress with Twitter accounts
substantially increased shortly after the 2008 elections, but concluded that the future may present spikes in Twitter usage during campaign periods. It seems as though either is this scenario not quite reached yet, or there are other forces at work. Perhaps there is a spike in tweeting the months after a new congress convenes, because of new energy in the beginning of a period or because of restructuring of social media teams. Another possibility is that Members of Congress mainly used their campaign accounts during the campaign, after which they switched to their official or another type of account. Since this study only included one campaign account (Lacy Clay), these differences would not be considered. A call for future research to investigate differences between types of accounts seems to naturally follow the discovery of this knowledge gap, but these studies are currently difficult to undertake, since type of account is not included in most Twitter account descriptions. If politicians further professionalize their SNS activities, such research will be made possible. Looking at the past, professionalization seems possible: In January 2010, only 35% of Members of Congress had
an account (Lassen & Brown, 2010), while at the time of this study, all Members of Congress of interest had Twitter accounts, and only one had no published tweets. Another possible explanation for the surprising results may be that individual events or the congressional schedule skew the data. In focusing on differences between campaign and routine periods, attention to events is beyond the scope on this study, but the results on domestic versus foreign emphasis and tweet frequency invite for future Twitter research to focus on individual events for a more detailed picture. As a side note, politicians tweeted notably more on
weekdays than weekends, something that is easily explained by the five-day workweek. It was hypothesized that politicians would increase the number of tweets published per day toward the end of the campaign. The data revealed the opposite – that they in fact publish less in the last month and week than during the rest of the period. This is an interesting result, perhaps pointing toward that at the end of the campaign, politicians’ teams focus their efforts on other campaign activities than social media. We might see a reversal of such an effect in the future, should the importance of SNSs relative to other online and offline campaign efforts rise.
Even though other studies have found no impact of competitiveness on tweet frequency and influence on Twitter, the result that competitiveness did not affect level of negative sentiment contrasts Evans et al. (2014), who found that competitiveness influenced the amount of “Attack” tweets. An improved technique to automatically detect nuances or manual content analysis would be necessary to investigate the differences between the influence of competitiveness on “Attack” tweets and negativity.
While there were several differences in Twitter behavior between the campaign and routine periods, no statistically significant differences could be found
between campaigning and non-campaigning politicians. This is a quite unexpected and interesting result, that suggests that election campaigns have broader effects on tweeting
behavior than on just the campaigning politicians. Indeed, the context of this study is the US, specifically a Senate and House of Representatives election that also coincides with a
presidential election. Undoubtedly, the election of one of – if not the most – important politicians in the world will have a ripple effect on the behavior of many politicians and others. Future studies incorporating a broader set of actors such as journalists, politicians from other countries and citizens, and studies from other countries, could shed light on the extent to which campaigns affect people’s Twitter behavior.
Some effects of control variables (Table D1) are worth mentioning. First, contrary to Hemphill et al. (2013), but in line with Evans et al. (2014), this study found that women tweeted more than men. Second, women express more sentiment and have a larger focus on domestic than other states, lending support to the notion presented by Evans et al. (2014) that on Twitter men and women seem to employ different campaign strategies. Third, this study echoes previous research in finding that Senators are more active tweeters than members of the House of Representatives (e.g Lassen & Brown, 2011), but finds other differences: House members published more hard news tweets, showed a more higher level of domestic versus foreign emphasis and expressed more positive sentiment than senators, whereas the latter displayed a larger focus on domestic versus foreign states. In line with Jungherr (2016), older politicians tweet more, but also tweeted more hard news than their younger counterparts. Republicans expressed more positive sentiment, had a more positive average level of
sentiment, and had a higher ratio of domestic versus foreign states than Democrats, while the latter displayed a higher level of negative sentiment. Independents focused more on domestic versus foreign issues than their counterparts, but this result should be interpreted with care, since only two Independents were included in the dataset.
There are limitations to the design of this study that should be discussed. First, as mentioned, it is not always evident what type of account a politician holds or who is
publishing the tweets. Some descriptions feature indicators of type of account or whether it is the politician themselves or staff tweeting. Other descriptions state that some tweets are signed to distinguish between staff and politician. Most accounts, however, do not contain any information on type of account or who publishes the content. Having access to this type of information would enable a more profound analysis. Second, as always with automated content analysis, there are nuances that will not be picked up by algorithms. The
operationalization of hard news in this study is certainly no absolute indicator of the type of content in tweets, but rather an approximation. As technology progresses, future research will be able to more intricately investigate such questions. Furthermore, some of the online
communication is done not in text but in images (sometimes with embedded text), videos or emoticons. The analysis in this study does not consider the meaning of such content. Third, the competitiveness index is based on reports released between a day and approximately a week before election day. If competitiveness has any influence on politicians’ social media activities, it would be ideal to incorporate such measures from earlier on in the campaign. Fourth, there may exist cross-country differences impeding generalizations to politicians in general. The kind of electoral system and technological adaption are two likely influential factors. Future research on other countries will strengthen our understanding of the subject. Limitations notwithstanding, this study contributed to extant research by investigating the previously unstudied differences in politicians’ behavior between campaign and routine periods on Twitter. As online information becomes more transparent, and if we improve our abilities to analyze large amounts of data, our knowledge of political communication will reach new levels.
References
Airne, D., & Benoit W. L. (2005). 2004 Illinois U.S. Senate debates. Keyes vs. Obama.
American Behavioral Scientist, 49(2), 343–352.
Aharony, N. (2012). Twitter use by three political leaders: an exploratory analysis. Online
Information Review, 36(4), 587-603.
Ahmed, S., & Skoric, M. M. (2014). My name is Khan: The use of Twitter in the campaign for 2013 Pakistan general election. In R. H. Sprague Jr. (Ed.), HICSS 2014: Proceedings of
the 47th Hawaii International Conference on System Science, 2242–2251, Washington,
DC: IEEE Computer Society.
Alexa. (2017, May 15). twitter.com traffic statistics. Retrieved from: http://www.alexa.com/siteinfo/twitter.com
Altheide D. L. (2013) Media logic, social control, and fear. Communication Theory, 23, 223 238.
Aragón, P., Kappler, K. E., Kaltenbrunner, A., Laniado, D., & Volkovich, Y. (2013). Communication dynamics in Twitter during political campaigns: The case of the 2011 Spanish national election. Policy & Internet, 5(2), 183–206.
Bennett, W. L., & Iyengar, S. (2009). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58, 707–731. Bird, S., Klein, E., & Loper, E. (2009). Natural language processing with Python (1st ed.).
O'Reilly Media.
Blumenthal, S. (1980). The Permanent Campaign: Inside the World of Elite Political
Operatives. Boston, MA: Beacon Press.
boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, 210–230.
automated content analysis approaches and techniques for digital journalism scholars,
Digital Journalism, 4(1), 8–23.
Bruns, A., & Highfield, T. (2013). Political networks on Twitter: Tweeting the Queensland state election. Information, Communication & Society, 16(5), 667–691.
C-SPAN. (2016, October 15). Members of Congress. Retrieved from: https://twitter.com/cspan/lists/members-of-congress?lang=en
Clinton, C. [HillaryClinton]. (2015, April 12). I'm running for president. Everyday Americans need a champion, and I want to be that champion. –H. Retrieved from: https://twitter.com/hillaryclinton/status/587336319321407488?lang=en
Cohen, B. C. (1963). The press and foreign policy. Berkeley, CA: University of California Institute of Governmental Studies.
Conway, B. A., Kenski, K., & Wang, D. (2013). Twitter Use by Presidential Primary
Candidates During the 2012 Campaign. American Behavioral Scientist, 57(11), 1596-1610 Esser, F. (1999). ‘Tabloidization’ of news: a comparative analysis of Anglo-American and
German press journalism. European Journal of Communication, 14(3), 291-324.
Esterling, K., Neblo, M., & Lazer, D. (2005). Home(page) style. International Journal of E-
government Research, 2, 48-62.
Evans, H. K., Cordova, V., & Sipole, S. (2014). Twitter Style: An Analysis of How House Candidates Used Twitter in Their 2012 Campaigns. Political Science & Politics, 47(2), 454-462.
Geer, J. (2009). Fanning the flames: The news media’s role in the rise of negativity in
presidential campaigns (Discussion paper). Cambridge, MA: Joan Shorenstein Center on
the Press, Politics, and Public Policy, Harvard University.
Political communication (pp. 136–145). Baton Rouge, LA: LSU Press.
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, 1612–1621.
Groshek, J. & Clough Groshek, M. K. (2013). Agenda Trending: Reciprocity and the Predictive Capacity of Social Networking Sites in Intermedia Agenda Setting across Topics over Time. Media and Communication, 1(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, 288–300.
Hansen, K. M., & Pedersen, R. T. (2008). Negative campaigning in a multiparty system.
Scandinavian Political Studies, 31, 408–427.
Hayes, A. F. 2013. Introduction to Mediation, Moderation and Conditional Process
Analysis. New York, NY: Guilford Press.
Hemphill, L., Otterbacher, J., & Shapiro, M. (2013). What’s congress doing on twitter?
Proceedings of the 2013 conference on computer supported cooperative work (CSCW 2013), San Antonio, TX.
Jungherr, A. (2016). Twitter use in election campaigns: A systematic literature review.
Journal of Information Technology & Politics, 13(1), 72-91.
Kim, S. T., & Lee, Y. H. (2006). New functions of Internet mediated agenda-setting: Agenda rippling and reversed agenda-setting. Korean Journal of Journalism and Communication
Studies, 50(3), 175–205.
Lassen, D. S., & Brown, A. R. (2011). Twitter: The electoral connection? Computer Review, 29, 419–436.
In R. S. Fortner & P. M. Fackler (Eds.), The handbook of media and mass communication
theory (pp. 249–268). West Sussex, UK: John Wiley & Sons.
McCombs, M., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public
Opinion Quarterly, 36, 176–187.
Office of the Press Secretary (2011). White House hosts Twitter @townhall. The White House. Retrieved from: http://www.whitehouse.gov/the-press-o ce/2011/06/30/white- house-host-twitter-townhall.
Ornstein, N., & Mann, T. (Eds.). (2000). The permanent campaign and its future. Washington, DC: American Enterprise Institute & The Brookings Institution. Ott, B. L. (2017). The age of Twitter: Donald J. Trump and the politics of debasement.
Critical Studies in Media Communication, 34:1, 59-68.
Perlmutter, D. D. (2008). Political blogging and campaign 2008: A roundtable. The
International Journal of Press/Politics 13, 160–170.
Trump, D. [realDonaldTrump]. (2015, March 13). The failing @nytimes is truly one of the worst newspapers. They knowingly write lies and never call to fact check. Really bad people!. Retrieved from:
https://twitter.com/realdonaldtrump/status/709089928324489216?lang=en
Reinemann, C., Stanyer, J., Scherr, S., & Legnante, G. (2012). Hard and soft news: A review of concepts, operationalizations and key findings. Journalism, 13(2), 221-239.
Russell Neuman, W., Guggenheim, L., Mo Jang, S., & Bae, S. Y. (2014). The Dynamics of Public Attention: Agenda‐Setting Theory Meets Big Data. Journal of Communication, 64(2), 193-214.
Sayre, B., Bode, L., Shah, D., Wilcox, D., & Shah, C. (2010). Agenda setting in a digital age: Tracking attention to California Proposition 8 in social media, online news, and
conventional news. Policy & Internet, 2, 7–32.
Semetko, H. A. (1996). The media. In L. LeDuc, R. G. Niemi, and P. Norris (Eds.),
Comparing democracies. Elections and voting in global perspective (pp. 254-279).
London: Sage.
Skovsgaard, M. (2014). A tabloid mind? Professional values and organizational pressures as explanations of tabloid journalism.
Sparks, C. (2000). Introduction: the panic over tabloid news. In: C. Sparks & J. Tulloch (Eds.)
Tabloid Tales: Global Debates over Media Standards (pp 1-40). Lanham, MD: Rowman
and Littlefield.
Swanson, D., & Mancini, P. (1996). Politics, media and modern democracy: An
international study of innovations in electoral campaigning and their consequences.
Westport, CT: Praeger.
Thelwall, M., Buckley, K., & Paltoglou, G. (2012), Sentiment strength detection for the social web, Journal of the American Society for Information Science and Technology, 63(1), 163–173.
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.
Tumasjan, A., Sprenger, T., Sandner, P., & Welpe, I. (2010). Predicting elections with Twitter: What 140 characters reveal about political sentiment. Proceedings of the 4th
International Conference of Weblogs and Social Media, Washington, D.C.
Trilling, D. (2015). last3200tweets.py. Retrieved from:
https://github.com/uvacw/tcst/blob/master/last3200tweets.py
Vaccari, C., & Valeriani, A. (2013) Follow the leader! Direct and indirect flows of political communication during the 2013 general election campaign, New Media & Society, 17(7), 1025-1042.
Van Aelst, P., & De Swert, K. (2009). Politics in the news: Do campaigns matter? A comparison of political news during election periods and routine periods in Flanders (Belgium). Communications, 34, 149-168.
Appendix A Hard news words Table A1
112 hard news words and counts in complete dataset.
Word Count Word Count Word Count
bill 8388 enforcement 1120 investment 488
congress 6976 police 1087 isis 482
office 4978 democracy 1024 workforce 471
law 3760 constitution 935 reforms 471
obamacare 3671 medicare 934 roundtable 458
trumpcare 3616 industry 905 legislative 454
jobs 3450 policies 895 delegation 452
tax 3402 conference 876 taxpayers 450
security 3167 gov 864 agriculture 442
healthcare 2741 schools 845 transportation 440
repeal 2697 political 844 subcommittee 436
funding 2632 regulations 835 terrorist 405
economy 2422 financial 802 regulatory 396
legislation 2349 amendment 783 technology 390
job 2045 employees 747 terrorists 390
congressional 1952 fund 692 cybersecurity 360
reform 1940 premiums 689 environmental 355
education 1930 chairman 674 constitutional 348
business 1900 terrorism 660 commerce 344
military 1886 climatechange 650 refugee 342
program 1869 manufacturing 647 ambassador 333
government 1792 congressman 639 unconstitutional 321
republicans 1783 mayor 639 immigrant 315
investigation 1723 environment 635 society 315
workers 1703 consumers 625 investments 313
policy 1588 gunviolence 617 holocaust 268
committee 1570 taxes 616 advocacy 264
secretary 1543 commission 603 university 263
administration 1536 housing 598 regulation 256
businesses 1445 citizens 579 payments 250
economic 1374 refugees 564 credits 248
safety 1373 income 551 entrepreneurs 239
epa 1362 taxpayer 541 socialsecurity 223
research 1290 sanctions 536 scientists 220
infrastructure 1262 discrimination 533 lawmakers 218
medicaid 1202 labor 531 educators 213
Appendix B
End of the campaign variables Table B1
Min, max, mean and standard deviation of end of the campaign variables.
Variable min max M SD
Positive sentiment last week .76 4.00 1.92 .55 Positive sentiment last month .21 4.00 1.88 .41
Positive sentiment all but last week
.18 2.88 1.81 .31
Positive sentiment all but last month
.23 2.88 1.81 .30 Tweet frequency week .00 24.29 1.28 1.92 Tweet frequency month .00 16.87 1.38 1.59
Tweet frequency all but last week
.15 10.19 1.63 1.55
Tweet frequency all but last month
Appendix C
Regression summary tables Table C1
Summary of hierarchical regression analysis for variables predicting hard news mentions over number of tweets (N = 864).
Model 1 Model 2 Variable B SE B B SE B (Constant) .420 .039 .429 .042 Gender .006 .015 .013 .005 .015 .012 Age .003 .001 .169** .003** .001 .166 Years in office (ln) -.004 .008 -.020 -.003 .008 -.014 Period -.104 .011 -.298** -.104** .030 -.298 Chamber .030 .014 .070** .051** .024 .118 Republican .013 .012 .036 .014 .012 .041 Independent .030 .084 .012 .022 .085 .009 Campaigning -.030 .032 -.060 PeriodxCampaigning .000 .032 .000 R2 .12 .12 F for change in R2 15.92** .61
Note: Years in office had its logarithm taken due to skew.
Table C2
Summary of hierarchical regression analysis for variables predicting domestic over foreign mentions (N = 864).
Model 1 Model 2
Variable B SE B IRR B SE B IRR
(Constant) -.030 .179 0.970 -.066 .193 Gender .080 .069 .041 1.083 .078 .069 .040 1.081 Age .003 .003 .046 1.003 .003 .003 .045 1.003 Years in office (ln) -.011 .038 -.012 0.989 -.008 .038 -.009 0.992 Period -.152 .052 -.099** 0.859 -.035 .138 -.023 0.966 Chamber .159 .066 .084** 1.172 .212 .109 .112* 1.236 Republican .049 .056 .032 1.050 .053 .056 .035 1.054 Independent .678 .388 .060* 1.970 .657 .390 .058* 1.929 Campaigning -.009 .147 -.004 0.991 PeriodxCampaigning -.136 .149 -.088 0.873 R2 .02 .02 F for change in R2 2.73** .60
Note: Years in office had its logarithm taken due to skew.
*p < .05. **p < .01. Table C3
Summary of hierarchical regression analysis for variables predicting domestic over foreign states mentions (N = 864).
Model 1 Model 2
Variable B SE B IRR B SE B IRR
(Constant) .953 .244 .927 .264 Gender .312 .094 .115** 1.366 .312 .094 .116** 1.366 Age -.002 .004 -.017 0.998 -.002 .004 -.017 0.998 Years in office (ln) .046 .052 .036 1.047 .046 .052 .036 1.047 Period -.019 .071 -.009 0.981 .022 .188 .010 1.022 Chamber -.610 .090 -.230** 0.543 -.622 .148 -.234** 0.537 Republican .223 .076 .104** 1.250 .222 .077 .104** 1.249 Independent .173 .530 .011 1.189 .178 .532 .011 1.195 Campaigning .042 .201 .014 1.043 PeriodxCampaigning -.049 .203 -.023 0.952 R2 .07 .07 F for change in R2 8.94** .03
Note: Years in office had its logarithm taken due to skew.
Table C4
Summary of hierarchical regression analysis for variables predicting positive sentiment (N = 864). Model 1 Model 2 Variable B SE B B SE B (Constant) 1.637 .066 1.647 .071 Gender .047 .025 .063* .046 .025 .063* Age .000 .001 .006 .000 .001 .005 Years in office (ln) -.018 .014 -.052 -.017 .014 -.050 Period .060 .019 .102** .053 .051 .090 Chamber .064 .024 .089** .078 .040 .108* Republican .142 .021 .244** .143 .021 .245** Independent -.009 .143 -.002 -.014 .144 -.003 Campaigning -.024 .054 -.028 PeriodxCampaigning .008 .055 .014 R2 .08 .08 F for change in R2 10.76** .01
Note: Years in office had its logarithm taken due to skew.
Table C5
Summary of hierarchical regression analysis for variables predicting negative sentiment (N = 864). Model 1 Model 2 Variable B SE B B SE B (Constant) 1.442 .052 1.453 .057 Gender .048 .020 .078** .049 .020 .080** Age .001 .001 .046 .001 .001 .049 Years in office (ln) .029 .011 .099** .027 .011 .092** Period .032 .015 .066** -.023 .040 -.046 Chamber .026 .019 .043 -.014 .032 -.023 Republican -.167 .016 -.339** -.170 .016 -.345** Independent .019 .114 .005 .035 .114 .010 Campaigning .027 .043 .038 PeriodxCampaigning .064 .043 .129 R2 .18 .18 F for change in R2 26.52** 2.33*
Note: Years in office had its logarithm taken due to skew.
Table C6
Summary of hierarchical regression analysis for variables predicting average sentiment (N = 864). Model 1 Model 2 Variable B SE B B SE B (Constant) .195 .084 .194 .090 Gender -.001 .032 -.001 -.003 .032 -.003 Age -.001 .001 -.024 -.001 .001 -.027 Years in office (ln) -.047 .018 -.100** -.044 .018 -.094** Period .027 .024 .035 .075 .064 .096 Chamber .038 .031 .039 .092 .051 .094* Republican .309 .026 .392** .313 .026 .397** Independent -.028 .181 -.005 -.049 .182 -.008 Campaigning -.050 .069 -.045 PeriodxCampaigning -.056 .069 -.070 R2 .19 .19 F for change in R2 28.65** 1.19
Note: Years in office had its logarithm taken due to skew.
Table C7
Summary of negative binomial regression for variables predicting tweet frequency (N = 864).
Model 1 Model 2
Variable B SE B CI (L, U) IRR B SE B CI L, U IRR
(Constant) 6.76** .2015 6.37, 6.16 863.7 6.740 .2161 6.31, 7.16 845.7 Gender .278** .0755 .13, .43 1.321 .280** .0757 .13, .43 1.323 Age -.011** .0033 -.02, -.01 .989 -.011** .0033 -.02, -.01 .989 Years in office (ln) .011 .0399 -.07, .09 1.011 .009 .0401 -.07, .09 1.009 Period -.183** .0580 -.30, -.07 .833 -.169 .1542 -.47, .13 .844 Chamber -.626** .0737 -.78, .48 .535 -.678** .1231 -.92, -.44 .507 Republican -.257** .0607 -.38, -.14 .773 -.262** .0614 -.38, -.14 .770 Independent .437 .4326 -.41, 1.29 1.548 .461 .4354 -.39, 1.31 1.586 Campaigning .085 .1653 -.24, .41 1.088 PeriodxCampaigning -.016 .1663 -.34, .31 .984 Chi-square 134.1** 134.0** AIC 11150 11154
Note: Years in office had its logarithm taken due to skew. AIC = Akaike’s Information
Criterion. CI L, U = Confidence Interval (Lower, Upper) *p < .05. **p < .01.
Table C8
Summary of simple linear regression for variables predicting tweet frequency (last week; N = 864). Model 1 Model 2 Variable B SE B B SE B (Constant) 4.426 .389 4.492 .420 Gender .417 .149 .094** .413 .149 .093** Age -.031 .007 -.187** -.031 .007 -.188** Years in office (ln) .031 .083 .015 .037 .083 .018 Last week -.348 .112 -.099** -.390 .300 -.111 Chamber -1.170 .143 -.269** -1.063 .237 -.245** Republican -.322 .121 -.092** -.314 .122 -.089** Independent .938 .844 .036 .896 .848 .035 Campaigning -.181 .320 -.036 PeriodxCampaigning .049 .323 .014 R2 .12 .117 F for change in R2 16.14** .17
Note: Years in office had its logarithm taken due to skew.
Table C9
Summary of simple linear regression for variables predicting tweet frequency (last month; N = 864). Model 1 Model 2 Variable B SE B B SE B (Constant) 4.183 .349 4.205 .378 Gender .377 .134 .094** .377 .134 .094** Age -.024 .006 -.165** -.025 .006 -.166** Years in office (ln) .005 .074 .003 .005 .075 .003 Last month -.294 .101 -.093** -.332 .269 -.105 Chamber -1.221 .128 -.310** -1.215 .213 -.308** Republican -.270 .109 -.085** -.269 .110 -.085** Independent .799 .758 .034 .796 .762 .034 Campaigning -.032 .288 -.007 PeriodxCampaigning .045 .290 .014 R2 .13 .133 F for change in R2 18.76** .01
Note: Years in office had its logarithm taken due to skew.
Table C10
Summary of simple linear regression for variables predicting positive sentiment (last week; N = 815). Model 1 Model 2 Variable B SE B B SE B (Constant) 1.698 .106 1.671 .114 Gender .090 .040 .082** .091 .040 .083** Age .001 .002 .016 .001 .002 .017 Years in office (ln) -.013 .022 -.025 -.014 .023 -.027 Last week .110 .031 .125** .152 .080 .172** Chamber .046 .038 .043 .029 .063 .027 Republican .075 .033 .085** .074 .033 .084** Independent .049 .222 .008 .055 .223 .009 Campaigning .049 .085 .039 PeriodxCampaigning -.049 .087 -.054 R2 .03 .03 F for change in R2 3.40** .22
