From “A Safer World and a More Hopeful America” to “Restore The Soul of The Nation”: The Use of Emotions in U.S. Presidential Campaign Ads
Master’s Thesis Authored by Anny Xu Student ID: 13346156 Supervised by Dr. Linda Bos
Graduate School of Communication
M.Sc. Communication Science – Political Communication June 25, 2021
Word Count: 7.305
Emotions have been a staple of the American political campaigning landscape to this day.
There are several contemporary trends that could have influenced the way emotions are being used nowadays in comparison to earlier decades, with one of these developments being the increasing shift from traditional media to social media platforms for distributing campaign ads in an efficient and less costly manner. With the phenomenon of mediatization and the
internalization of (mass) media logic by politics, this shift raises the question whether politicians and campaign staff have adapted their strategies to meet social media logic. In order to enhance social transmission and engagement by users with campaign ads, and increase the likelihood of ads going viral, it was expected that more positive and activating emotions are employed. A longitudinal content analysis of U.S. presidential campaign ads from 2004, 2012 and 2020 has shown that while positive emotions have not been increasingly used over time, there is some confirmation for a greater use of activating emotions.
Furthermore, it was postulated that enthusiasm as both a positive and activating emotion would become the most commonly used emotion in campaign ads. This was however not the case. Lastly, the role of humor, which is distinct to internet culture with humorous content achieving disproportionate attention across social media platforms, was investigated for campaign ads. The results emphasized a minor relevance of humor for campaign ads. Overall, the study was not able to provide hard evidence for a change in the usage of emotions in presidential campaign ads due to a shift from traditional mass media to social media. Instead, the complexity of campaign strategy and impact of other factors (e.g., issues, candidates) were highlighted. Nevertheless, an alignment with social media logic regarding activating emotions was found, in addition to differences in the use of emotions between those selected elections.
Emotions have always been a staple of American political campaigns (e.g., Ridout, Franklin Fowler, & Branstetter, 2010), and looking at very successful and old campaign ads like
“Daisy Girl” but also popular contemporary ads like “Yes We Can” might prove the
importance of emotional appeals. The use of emotions as a campaign strategy is thereby based on a variety of effects: Emotions can increase interest in politics, attention (see Brader, 2005) and participation (see Valentino et al., 2011; Weber, 2012). However, it can also lead to less rational decisions, manipulation and enhance affective polarization (see McLaughlin et al., 2020), which is problematic on a normative level but might be desirable for political advertisers and candidates. While emotions have been utilized in political advertising for a long time, there are several trends that could have influenced the way political messages are being transported now in comparison to earlier decades. Developments like mediatization (Livingston, 2009), declining policy substance in political communication, or the rise of populism with its rhetoric that heavily relies on emotional cues and attacks, also enhancing the use of negativity in recent years (see Nai, Martiñez I Coma & Maier, 2019), could have influenced the strategies political ads are based on. Furthermore, there has been a shift from traditional channels to the Internet when it comes to political campaigning (Owen, 2018).
While it might be more obvious how the first mentioned trends and developments could influence the content of political advertising, it is perhaps more abstract to understand how media platforms themselves could have an impact on the design of political ads including the level of emotionality and type of emotion.
The underlying argument is based on the network media logic of social media platforms that differs from the media logic of traditional mass media (Klinger & Svensson, 2014). Distribution on social media platforms is built on the logic of virality (Klinger &
Svensson, 2014) and the associated ease of sharing content is probably one of the reasons why
social media are viewed as cheaper and more effective than traditional media by marketeers (Berger & Milkman, 2012). This also offers the opportunity for politicians to greatly increase their reach without spending more when distributing campaign ads on social media. As the cost of federal elections in the U.S. are immensely high, creating political ads that are more likely to be shared or even go viral would be very much in the interest of politicians and their campaign strategists. While it has been shown that emotional aspects of content affect
whether content is shared or not (see Heath, Bell & Sternberg, 2001; Berger & Milkman, 2012), research on the role of emotions for social transmission of political content has been conducted in the context of moral rhetoric (see Brady et al., 2017) but is still rather limited within the scope of political campaigning and social media logic (e.g., Borah, 2016). Thus, it is still necessary to verify if politicians have indeed adapted their strategies to meet social media logic. Therefore, this study aims at answering the following research question: How has the increasing shift from traditional media to social media platforms influenced the use of emotions in U.S. presidential campaign ads?
Political communication has been impacted by a number of developments over time, such as individualization, rationalization or economization to name a few (Blumler & Kavanagh, 1999), but the arguably most dominant aspect has been mediatization. Mediatization has not only shifted sources of information and allowed the media to be independent from political institutions, but also resulted in the internalization of media logic by political actors in their processes (Strömbäck, 2008). Media logic refers to the assumptions and processes for
constructing messages within a particular medium (Altheide, 2004), which could be seen as a playbook that politics follow to ensure presence in a medium. It goes without saying that this development of political communication evolved over multiple decennia, where democracies
are finding themselves in the Age of Professionalization and Commercialization of political communication (see Blumler & Kavanagh, 1999). Media logic has been criticized with regard to its scarce empirical evidence as well as lack of differentiation between different media (Brants & van Praag, 2017). However, when it comes to the evolution of campaign communication in the new media era specifically, the research interest of this study, three distinct phases on the use of new media in election campaigns can be construed (Owen, 2018).
In the first phase Old Media, New Politics (1992-1994), candidates used established nonpolitical and entertainment media to circumvent mainstream press gatekeepers (Owen, 2018). This era is still living on as political candidates continuously seek favorable and widespread coverage in popular magazines and appearances in talk shows (Baum, 2005).
With the introduction of technological innovations, i.e., the advent of the internet, the second phase New Media, New Politics 1.0 (1996-2006) led to candidates implementing text-based websites by 2000, and basic interactive elements on campaign websites becoming the standard in the 2004 U.S. election. The final phase New Media, New Politics 2.0 (2008- present) can be differentiated from the previous period by innovations in digital election communication that revolve around networking, collaboration and community building as well as active engagement. Notable developments include the use of social networking sites such as Facebook but also video sharing sites like YouTube in 2008, and the feature of Twitter and other microblogging sites more prominently in the 2010 midterm elections (Owen, 2018).
Although other authors have also postulated a new era of campaigning which is defined by the introduction of information and communication technology, especially the internet (see Karlsen, 2009; Norris, 2000), and the internet is claimed to have revolutionized political campaign communications (Panagopoulos, 2009), scholars haven’t reached a
consensus about the actual impact of the internet on political practice. On the one hand, the normalization hypothesis suggests that offline patterns are being replicated in the online sphere (Foot & Schneider, 2006) and that the internet solely serves as an extended tool to distribute the same information used in offline campaigning (Gibson et al., 2005; Schweitzer, 2008). On the other hand, the innovation hypothesis posits that online campaigning is
disengaged from and substitutes typical offline patterns leading to differences between on- and offline campaigning (Schweitzer, 2008). Nevertheless, scholars found that integrating these new platforms posed implications for electioneering and affected the strategy behind campaigns (e.g., Franklin Fowler et al., 2020).
Strategic campaigning in the internet era
Each U.S. campaign cycle since 1994 has generated its own internet story (Raisinghani &
Weiss, 2011) and achieving tremendous success building buzz (Panagopoulos, 2009) –
ranging from the creation of web sites, usage of email or net-organization of house parties and adoption of social networking sites (Gueorguieva, 2009). By the 2006 U.S. midterm election many of the larger campaigns already had internet strategists on staff or as consultants (Greenfield, 2007), which is not surprising given that the technological changes of the twentieth century are likely to shift the strategic landscape of campaign communication and alter the content of campaign messaging (Franklin Fowler et al., 2020). Although there are few systemic analyses of how the web is used (Druckman, Kifer, & Parkin, 2010), there is evidence that the internet and especially social media are used in a different manner compared to offline campaigning and that other considerations are made when setting up online
messages and online presences.
When it comes to functionality, the use of social media for political campaigns is more active and goal-directed (Owen, 2018) with candidates using ads more for mobilization of supporters than for persuasion (Franklin Fowler et al., 2020; Williams & Gulati, 2009) as well
as promotions of fund-raising and volunteering efforts being in the foreground of strategic online campaigns (Bor, 2014; Cogburn & Espinoza-Vasquez, 2011). Moreover, strategies concerning content and style differ from traditional media, which might be due to the different objectives of online campaigns. With social media, candidate-centered campaigning is being supplemented by personality-centered campaigning that goes beyond conveying political messages (Vergeer, Hermans, & Sams, 2011) and is accompanied by a presentation of the candidate in a more informal manner (Bor, 2014). This also facilitates creating a sense of informality and intimacy (Panagopoulos, 2009) that could foster parasocial interaction and relationship (see Horton & Wohl, 1956) which could be beneficial for both offline and online mobilization. Campaign strategists are mindful about conveying a positive online presence (Bor, 2014) which is also visible in ads designed for the internet being overwhelmingly positive and engaging in less attacking of the opponent as well as employing humor at a greater likelihood compared to traditional political ads such as on TV (Ridout et al., 2010;
Franklin Fowler et al., 2020).
Despite these empirical findings, some studies suggest that political use of new media rather confirms the normalization hypothesis because it is predominantly used for one-way, top-down mass messaging instead of interactive communication (McCullagh, 2000;
Druckman, Kifer, & Parkin, 2010; Pew, 2012; Bor, 2014). This especially holds true for candidates higher in popularity on social media, who become less social as a result of logistical difficulties to maintain individual, two-way communication (Vergeer et al., 2011).
One measure that was adopted early on was the usage of YouTube as an alternative communication channel to distribute political video ads (see Gueorguieva, 2009; Ridout, Fowler Franklin, & Branstetter, 2010; English, Sweetser, & Ancu, 2011). In the pursuit of combining both offline and online strategies in a seamless and efficient manner, distributing the classic 30 second ad on social media might serve as the ideal intersection by using a traditional one-way format with its many benefits (e.g., Ridout et al., 2010), but also taking
advantage of social media’s features. As previously laid out, changes and adaptations in content and style arise with the implementation of social media platforms and adhering to the logic of social media, at least to a certain extent, might be crucial for the success of campaign strategies.
Social media logic
Social media as a communication channel offers not only an alternative way to reach younger demographics but is far more efficient and less costly than traditional tools (Raisinghani &
Weiss, 2011; Gueorguieva, 2009). It allows more candidates to engage in campaigning and advertise specifically (Franklin Fowler et al., 2020). Additionally, the ability of campaigns to achieve broad dissemination and access to voters is potentially unlimited (Franklin Fowler et al., 2020; Gueorguieva, 2009). Indeed, the potential of messages in form of videos becoming
“viral”, and therefore giving an item the maximum exposure possible (Nahon et al., 2011), is being considered one of the important areas that are revolutionizing campaigns
(Panagopoulos, 2009). This opportunity emerges from social media’s logic of distribution that is based on networks and popularity.
A key feature of the internet and social media is that they are networked and put power at the edges, i.e., users, and in the connections between them (Palfrey, 2004). This fosters fast dissemination of content through a network effect that predicts that networks tend to grow in an exponential manner allowing for messages to reach a wide audience even though
candidates’ first-degree networks might be rather small (Vergeer et al., 2011). In fact, the Obama campaign in 2008 made strategic use of this feature by explicitly asking supporters to send reminders to their networks about attending the Iowa caucus (Williams & Gulati, 2009).
Moreover, social media’s distribution of content is asymmetric as platforms enhance the dominance of popular content (Klinger & Svensson, 2014). Popularity on social media is two- way traffic with algorithms automatically assessing differentiated value to and users’
engagement with the item at hand resulting in increased visibility (van Dijck & Poell, 2013).
Here, platforms have their own thermometers for measuring aggregated popularity (van Dijck
& Poell, 2013), e.g., “most viewed” videos on YouTube or “trending topics” on Twitter, but also engagement in the form of likes, comments and shares can be seen as indicators.
Furthermore, likes and shares can be considered as social endorsement and higher social endorsement in turn also means momentum for a candidate (Borah, 2016) resulting in a loop because people engage with content more once they observe it is popular (Salganik, Dodds, &
Watts, 2006). Thus, the challenge of candidates is to create content that voters will forward within their networks, endorse and engage with, and that potentially becomes viral because otherwise, the message will not reach beyond a limited circle of people (Klinger & Svensson, 2014).
Power of emotions in (online) political campaigning
Looking into reasons for social transmission on social media, Baek et al. (2011) have found that information sharing, and entertainment accounted for the most important reasons for why people share content. In fact, humorous content, which is one gateway to entertainment, is deeply engrained in internet culture, where it is both expected and rewarded (Phillips &
Milner, 2017), and achieves disproportionate attention across social media platforms (Davis, Love, & Killen, 2018). However, if we look at virality of content, which is on a larger scale than individual social transmission and includes factors like engagement and algorithm, empirical research on reasons why content goes viral across the internet is still limited (Botha
& Reyneke, 2013). Nevertheless, previous scholars have allocated emotions a role in why messages are shared or even go viral (see Heath, Bell, & Sternber, 2001; Phelps, 2004;
Dobele et al., 2007). In the context of moral and political discourse, social media communications containing moral and emotional cues are consistently associated with
increased virality (Brady et al., 2017; Valenzuela, Piña, & Ramirez, 2017). This can partly be
explained through the prioritization of emotional (and moral) over neutral content in visual attention that results in an attentional capture that such content produce, and which embodies a prerequisite to determine which content drives online engagement (Brady, Gantman, & van Bavel, 2020). After attention is secured, emotions facilitate sharing of content and increase its likelihood to become viral. Once the content of a video is relevant to individuals, their
emotional reaction to it causes them to share it with their network, which in turn has
emotional reactions to the video as well and indicates that emotional contagion possibly is a key determinant of viral marketing (Botha & Reyneke, 2013). However, not all emotions lead to equal chances of social transmission. Results of Berger and Milkman (2012) suggest that content is more likely to become viral the more positive it is, but arousal also promotes activation and sharing regardless of their positive or negative valence (Berger, 2011). While these findings do indicate which aspects might boost social transmission and virality, some of the older studies might not reflect the status quo anymore, particularly because it is a recent phenomenon and there might be new developments.
Political campaigns are not only designed for social transmission, but also take other strategic considerations into account when it comes to using emotions, such as the effect on voters. Among others, these include inducing information processing, increasing willingness to vote and reducing the role of partisanship in vote decision (Brader, 2005; Ridout & Searles, 2011). The most dominant model for applying emotions to politics is the Theory of Affective Intelligence (Marcus, Neuman, & MacKuen, 2000). Although emphasis has been put on enthusiasm and anxiety as broader dimensions, it should be mentioned that the influence of distinct emotions such as anger, fear, hope and sadness on political participation have also been examined (see Bodenhausen, Kramer, & Susser, 1994; Valentino, Gregorowicz &
Groenendyk, 2007). The task for campaign strategists is consequently to identify and use emotions in political ads that would enhance desired effects on voters and tap into social media’s logic of distribution simultaneously.
Hypotheses and research questions
When it comes to expectations regarding the use of emotional appeals in campaign ads over time, with an increasing shift towards social media as communication channels as well as continuous professionalization and familiarization with new media environments, the
previously mentioned observations already give some indication about developments. Given that content inducing positive emotions are more likely to be viral and shared (see Berger &
Milkman, 2012; Botha & Reyneke, 2013) and therefore, show longer lifespans on platforms than negative content (Wu et al., 2011), it would be plausible that positive emotions are increasingly used over time in campaign ads (H1). Furthermore, political social media use is more action-oriented and intends to mobilize supporters, which should thus lead to an
increased application of high-arousal, activating emotions that would not only be
advantageous for evoking offline behaviors in voters but also enhance engagement and social transmission on social media. Indeed, the most common emotional appeals in the 2008 and 2012 Facebook campaign strategies were enthusiasm, fear and anger (Borah, 2016). The expectations are consequently that activating emotions are increasingly used in campaign ads (H2). Combining the relevance of positive valence with the activation potential of emotions, it would be reasonable, if enthusiasm has become the most commonly used emotion in campaign ads (H3).
Lastly, humor as an emotional appeal could serve as one indicator for an adaptation to social media logic. Humorous content is distinctly of the internet (Davis et al., 2018) and does not only drive social media usage by inducing feelings of amusement and entertainment but is popular and has been shown to capture attention and encourage sharing (Panagopoulos, 2009;
Botha & Reyneke, 2013). Therefore, it is not surprising that a clear intersection between humor and politics exists online (Wells et al., 2016). In the realm of political campaigning, humor was a commonly used emotion in the Facebook campaign strategy of Barack Obama in
both 2008 and 2012, with this type of posts receiving the highest number of likes and shares (Borah, 2016). However, humor hasn’t really been employed in offline political ads, and these findings from past campaign strategies are not enough to base expectations upon, which is why the following question is raised: How has the use of humor in campaign ads evolved over time? (RQ)
In order to analyze whether and how the use of emotions has changed and adapted to social media logic, a longitudinal content analysis was conducted. A quantitative over qualitative approach was chosen to gain a systematic overview on trends and developments regarding the use of emotional appeals in campaign ads. Furthermore, the content analysis was conducted manually and not automatically as emotions are a human response and complex to
operationalize for automation. Lastly, this study analyzed videos which is mostly not suited for automation.
Selection of research units
The research units at hand were U.S. presidential campaign video ads, a staple of American campaigning (e.g., Ridout et al., 2010). Ads from the 2004, 2012 and 2020 presidential elections were collected to integrate ads from an era that was mainly aligned with television and traditional mass media channels, a time when classic social media platforms like
Facebook, YouTube and Twitter have been established and adopted by a broad audience, and the most recent election with increasing technological sophistication of social media and new platforms (e.g., TikTok, Snapchat, etc.). Furthermore, ads from both primaries and general elections were included with the aim to diversify the variety of candidates and potentially their personal style of campaigning.
The population from which the sample was drawn was partly gathered from Stanford University’s “Political Communication Lab” (PCL), which archived all U.S. presidential and mid-term election ads from 1994 to 2016. The 2004 primary election ads as well as both primary and general election ads from 2012 were taken from there. Due to technical issues, the 2004 general election ads could not be accessed and therefore, ads were manually drawn from YouTube and “The Living Room Candidate” based on the complete list of ads provided on the PCL website. However, it was not possible to retrieve all of them, and 45% were added to the population. As for the 2020 election, there was no publicly available set of all ads due to its recency. Hence, primary and general election ads were manually drawn from YouTube by going through each candidate’s channel and retrieving ads from their playlists. While some channels have labeled their videos in an explicit manner that made them identifiable as ads, others did not engage in such categorization. In these cases, length and style of title often served as cues, such as 30 or 60 second videos with one-word titles followed by name of candidate for President 2020, e.g., “Indivisible | Joe Biden For President 2020”. In doubt, videos were watched to evaluate whether they were ads or not. Features like sponsorship statement and message approval statements by the candidate served as additional cues to determine ads. The distinction between ads designed for the primary race or general election was made based on the publishing date of the video. The collected ads included both televised ads and those that were solely distributed in the online sphere. In total, the population
consisted of 1701 ads, with 135 from 2004, 751 from 2012, and 815 from 2020. Ads in Spanish or from independent candidates were not included in the population.
The sampling was randomized and contained 150 cases with equal distribution of 50 cases per election year to prevent distortions and biases. In order to reflect the higher political
importance of the general election while also ensuring variance and diversity in different
campaigning styles by incorporation more political actors into the analysis, the sample consisted of 40% primary election ads and 60% general election ads. The sample contained ads from 21 candidates with most ads coming from or featuring nominees of the general election. The 2004 and 2012 subsamples contained an (almost) equal balance of both
candidates: George Bush and John Kerry were represented with 16 ads each, Barack Obama with 22 and Mitt Romney with 20 ads. The 2020 subsample included 23 ads from Joe Biden and 15 ads from Donald Trump. This imbalance stems from the larger number of ads put out by the Joe Biden campaign. Furthermore, both parties were somewhat evenly distributed with 54% of ads associated with the Democratic party and 46% of ads with the Republican party.
With regards to sponsors of ads, the vast majority of ads were sponsored by the candidate (85.2%) and only a small number by either party (6.7%) or an interest group/PAC (8%). The length of ads ranged from 15 to 169 seconds, although most ads were created in the common format of 30 seconds (59.3%) or 60 seconds (15.3%).
The main focus of the analysis were emotional appeals (see codebook in Appendix A for specific wording of the variables). On a general level, the presence of emotional appeals was coded in addition to whether emotional appeals were the dominant appeals in ads and the overall emotionality. Furthermore, for a number of specific emotions, for example
enthusiasm, humor and anger, the presence as well as the strength of appeal (no appeal, some appeal and strong appeal) were coded but later transformed into a binary variable due to low intercoder reliability. The majority of these variables were adopted from Brader (2006).
Specific variables were later merged into several indices. Enthusiasm and humor were used to create the variable positive emotions (M = .15, SD = .25, 0 = no, 1 = yes), while anger, fear and sadness formed negative emotions (M = .13, SD = .22, 0 = no, 1 = yes). Moreover, activating emotions (M = .18, SD = .24, 0 = no, 1 = yes) consisted of enthusiasm, anger and
fear, while non-activating emotions (M = .11, SD = .32, 0 = no, 1 = yes) only included sadness appeals due to insufficient intercoder reliability of the other variables pride and compassion.
In videos ads, other factors also contribute to their perceived emotionality. To account for this aspect, audio and visual characteristics were also operationalized into the type of music and sound effects being used, and the dominant color scheme (Brader, 2006). Lastly, the dominant expression of the featured candidate (Kaid & Johnston, 2001) or main
protagonist of the ad was coded.
Additionally, formal variables like corresponding election year, length of the ad, party, type of sponsor and type of race (contested vs. uncontested) were included. In order to
account for other factors that exert an influence on the use of emotions in campaign ads, a number of control variables were included. These are for one political dynamics such as election round (primary vs. general election) and challenger or incumbent status of a candidate, and also the individual candidate as individual traits and preferences have an influence on campaign style (see Nai, Martinez i Coma & Maier, 2018). Furthermore, certain political issues are more prone to be framed emotionally than others (see Skonieczny, 2018;
Sulkin, Moriarty & Hefner, 2007). Therefore, the dominant issues (see Franz et al., 2020;
Kaid & Johnston, 2001) of the ad were coded as well. The original codes for dominant issues were later merged from 21 into seven categories due to intercoder reliability. Economic (23.5%) and social issues (20.8%) were the most predominant issues while a significant set of ads didn’t have an emphasis on any issue (16.1%) and, for example, highlighted the candidate instead.
Coding decisions were mostly based on the content of the ads except for few formal
characteristics. The codebook included more detailed instructions on emotional appeals due to
them being at the core of this study and their level of complexity and subjectivity. For all emotion-related variables, coders were instructed to pay attention to both explicit emotional appeals, such as rhetoric, and implicit contributing factors like music and imagery. Moreover, for each emotion-specific appeal an example ad was listed to make them less abstract and more tangible. Coding decisions for different levels of emotional intensity and strength of appeal were to be derived from the presence of said appeal on multiple dimensions. If an appeal was only translated to one dimension, for example only rhetoric, visual or audio, it would be considered a moderate appeal and the ad somewhat emotional. In contrast, if the appeal was manifested in various dimensions through rhetoric, visual and/or auditive cues, the appeal was strong and the ad very emotional overall. The initial coder training consisted of a thorough explanation and discussion of the codebook and instruction to watch the above- mentioned sample ads.
Two coders were involved in data collection with the main coder coding the whole sample. To assess the reliability of the coding, a randomly selected subsample of initially 10%
(n = 15) was double-coded. Three indices were used to calculate reliability scores: per cent agreement, Kappa and Krippendorff’s Alpha. The test showed insufficient reliability for almost all variables, and consequently, additional coding and training was necessary. For the second training session both coders coded a number of ads together and established common ground through discussion. To increase the statistical power of reliability tests, the second subsample for double-coding comprised 30 randomly selected ads. Intercoder reliability was then calculated using the complete subsample of 30% (n = 45). This time five emotion- specific variables showed sufficient reliability surpassing the threshold of .67. The variables emotional appeal (Krippendorff’s alpha = .64), overall emotionality (Kalpha = .57) and enthusiasm (Kalpha = .58) were included in data analysis despite their insufficient intercoder reliability. This decision was made due to their importance for this study’s objective, yet the results need to be interpreted with caution. Due to low intercoder reliability scores, other
emotion-specific variables like pride and compassion had to be neglected for data analysis.
The exact reliability scores can be found in table B1 of Appendix B.
Before conducting the needed analyses to test the postulated hypotheses and answer the research question, exploratory analyses were executed to gain a better understanding on the use of emotional appeals before diving into the specificities. In general, emotional appeals were present in 53.3% of the sample. An one way analysis of variance showed that the effect of year was significant (F(2, 147) = 8.85, p .001). Post hoc analyses using the Duncan criterion indicated that emotions were significantly more used in the recent 2020 election (M
= 0.76, subset 1) than in 2004 (M = 0.46, subset 2) and 2012 (M = 0.38, subset 2). In fact, there is a positive relationship between the use of emotional appeals and the time dimension with the election year being a significant predictor (β = 0.15, t(148) = 3.08, p = .002; R2 = 0.06, F(1, 148) = 9.49, p = .002). By controlling for other factors, namely dominant issue of the ad, fixed effects for candidates, election round (primary or general election), incumbent or challenger, as well as length of ad, the effect of election year diminishes and instead,
individual candidates emerge as the most relevant predictors (R2 = 0.15, F(2, 129) = 10.97, p
< .001), in this case Joe Biden (β = 0.46, t(129) = 4.24, p < .001) and Donald Trump (β = 0.34, t(129) = 2.67, p = .009). Furthermore, looking into the extent of emotionality and differentiating between somewhat emotional and very emotional ads excluding controls, a linear regression has provided support that the overall emotionality of ads has increased over time (β = 0.26, t(148) = 3.80, p <. 001, R2 = 0.09, F(1,148) = 15.03, p < .001).
The use of positive emotions
First of all, it was expected that positive emotions are increasingly used in campaign ads (H1).
To determine the change over time, a linear regression was calculated with positive emotions
as the criterion and the election year as the predictor. Control variables were left out at this stage. The analysis indicated no increase in the use of positive emotions (β = 0.02, t(148) = 0.02, p > .05; R2 = 0.003, F(1, 148) = 0.37, p > .05). An additional one-way analysis of variance demonstrated that positive emotions were significantly more used in campaign ads during both 2004 (M = 0.18) and 2020 elections (M = 0.21) than in 2012 (M = 0.07; F(2, 147)
= 4.74, p = .01). However, positive emotions are overall not more frequently employed than negative emotions as a paired sample t-test showed (t(149) = 0.772, p > .05). Furthermore, negative emotions were found to have been increasingly used over time (β = 0.07, t(148) = 3.37, p = .001; R2 = 0.07, F(1, 148) = 11.33, p = .001) suggesting the opposite development than expected. Based on these results, the first hypothesis must be rejected.
To investigate the role of other factors, the control variables dominant issue, candidate, election round, incumbent or challenger and length of ad were added to a hierarchical multiple regression model. This has led to the result that certain issues and individual candidates have a negative impact on the use of positive emotions. The results are depicted in table 2 and the complete results from the regression analyses including all covariates can be found in table B2 of Appendix B.
Table 2. Regression analysis summary for positive emotions.
Predictor β t(126) p
Control variables Dominant issue
Social issues -0.20*** -3.81 .000
Military, national security, terrorism -0.24** -3.30 .001
Economic issues -0.15** -2.89 .005
Mitt Romney -0.20** -3.01 .003
Table 2 (continued).
Newt Gingrich -0.30* -2.20 .03
Election year -0.02 -0.26 .79
Notes. R2 = 0.21, F(5, 126) = 6.55, p < .001, ***p < .001, **p < .01, *p < .05
The use of activating emotions
The second main hypothesis postulated that activating emotions are increasingly used in campaign ads (H2). First of all, a paired sample t-test was calculated and indicated that
activating emotions (M = 0.18) are overall more often used than non-activating emotions (M = 0.11, t(149) = 2.11, p = .036). However, due to the removal of other emotion-specific
variables as mentioned previously, sadness was the only non-activating emotion. With regards to the use of activating emotions over time, a linear regression has found a significant effect ( = 0.05, t(148) = 2.14, p < .05; R2 = 0.03, F(1, 148) = 4.56, p < .05). However, adding the same control variables as previously to a hierarchical multiple regression model diminished the effect of the election year on the use of activating emotions. Instead, specific issues, several candidates and the election round emerged as relevant predictors as table 3 illustrates.
The complete results from the regression analyses including all covariates can be found in table B3 of Appendix B.
Table 3. Regression analysis summary for activating emotions.
Predictor β t(125) p
Control variables Dominant issue
Crime, gun control 0.32** 3.06 .003
Environmental concerns 0.58* 2.58 .01
Table 3 (continued).
Dennis Kucinich 0.58* 2.58 .01
Donald Trump 0.10 1.53 .128
Joe Biden 0.13* 2.44 .02
General Election Independent variable
0.09* 2.05 .04
Election year -0.07 -0.51 .61
Notes. R2 = 0.21, F(6, 125) = 5.48, p < .001, ***p < .001, **p < .01, *p < .05
Although the variable election year doesn’t have a significant effect on the criterion in the multiple regression model per se, both nominees for the presidential election of 2020 emerged as predictors in the regression model. To examine whether the effects of candidates caused the change, another regression was executed with all previous control variables excluding candidates. Again, issues on crime, gun control ( = 0.32, t(145) = 3.29, p = .001) and environmental concerns ( = 0.56, t(145) = 2.47, p = .015) as well as general election (
= 0.1, t(145) = 2.50, p = .02) were significant predictors, and no effect for election year was detected (R2 = 0.12, F(3, 145) = 6.53, p < .001). Although the effect of election year on the use of activating emotions weakened substantially after adding control variables, the result of the linear regression provides some confirmation for the second hypothesis, albeit limited.
The use of enthusiasm
Due to enthusiasm being a positive and activating emotion, it was hypothesized that enthusiasm has become the most commonly used emotion in campaign ads (H3). On a descriptive level, enthusiasm was the most employed emotional appeal (M = 0.25) followed by fear (M = 0.15) and anger (M = 0.13). In order to investigate whether the use of enthusiasm
has increased over time, a linear regression was employed. The analysis showed no support for an enhanced utilization of enthusiasm appeals for later campaign ads compared to earlier ones (β = 0.02, t(148) = 0.46, p > .05; R2 = 0.001, F(1, 148) = 0.21, p .05). To gain a better understanding of the development, an additional one-way analysis of variance was carried out. This showed that ads from 2012 (M = 0.06) used significantly less enthusiasm appeals than in 2004 (M = 0.32) and 2020 (M = 0.36; F(2, 147) = 7.73, p = .001). Although
enthusiasm is overall the most used emotion in campaign ads, there hasn’t been a linear increase over time, hence the hypothesis has to be rejected. Regressing the other emotional appeals by year has shown that anger has seen the most increase over time ( = 0.1, t(148) = 3.01, p = .003, R2 = 0.06, F(1, 148) = 9.06, p = .003) followed by sadness ( = 0.08, t(148) = 2.56, p = .011, R2 = 0.04, F(1, 148) = 6.56, p = .011).
The use of humor
Based on the observation that humorous and amusement-invoking content is often being engaged with and socially transmitted on social media, it was questioned how the use of humor in campaign ads has evolved over time (RQ). The initial linear regression showed no significant effect of election year on the use of humor (β = 0.01, t(148) = 0.42, p .05) and a supplementary one-way analysis of variance did not make any significant differences between the elections apparent (F(2, 147) = 0.35, p .05). Furthermore, data displayed that humor appeals were applied very little in campaign ads with only nine out of 150 ads containing this emotional appeal.
The main objective of this study was to investigate how the increasing shift from traditional media to social media platforms has influenced the use of emotions in U.S. presidential campaign ads. For that, a longitudinal quantitative content analysis was conducted, which
would not allow for statements of causation, but, as a first step, would give some indication whether politicians have adapted their strategies to meet social media logic.
Looking at campaign ads from the presidential elections in 2004, 2012 and 2020, the analyses provided mixed results. In case of an alignment with social media logic, it was expected that more positive and activating emotional appeals would be integrated because content with positive appeals is more likely to be shared and to go viral (see Berger &
Milkman, 2012; Botha & Reyneke, 2013). Regarding the use of positive emotions, they were not increasingly used in ads and generally not more often used than negative emotions. In fact, more and more campaign ads contained negative emotional appeals, which could be a result of the increase in activating emotions that includes anger and fear. Activating emotions are not online beneficial for inducing offline political behavior (see Brader, 2005) but also foster engagement and social transmission on social media, which is important to spread the message beyond one’s own, limited circle (Klinger & Svensson, 2014) and profit from social media platforms’ algorithm (see Klinger & Svensson, 2014; van Dijck & Poell, 2013).
Consistent with the formed expectations, a positive trend was found for activating emotions which include enthusiasm next to anger and fear. This trend is likely driven by the more frequent use of anger. However, the development over time was explained by other factors such as certain issues, individual candidates and characteristics of the race that impact the implementation of activating emotions in campaign ads. Furthermore, an enhanced use of enthusiasm appeals would have insinuated that ads are created with (social) media logic in mind. While enthusiasm was the single most applied emotional appeal, there was no positive linear trend while there was for anger and sadness. Lastly, the development of humor as an emotional appeal in campaign ads was examined and would have provided support for a convergence of campaign strategy and social media logic as humorous content is dominant on social media and likely to be socially transmitted and engaged with. The study was, however, not able to reveal a significant trend, but rather displayed the minor role of humor in
campaign ads. This fortified previous doubts about the use of humor in ads in contrast to the use of humor in social media posts, which was commonly used in past election campaigns (Borah, 2016), and point towards different strategies depending on the communication medium.
Comparing the use of emotional appeals between the elections has also exhibited significant differences. Positive emotions and especially enthusiasm were less prominent in 2012 than in 2004 and 2020. While there are many factors that could have stunted the use of those emotional appeals, it is likely that the issues on 2012 candidates’ agenda promoted different frames. In 2012, emotional appeals were used the least and the dominant issue overall was of economic nature, which is negatively correlated with the use of positive emotions. It is plausible that these dips sprung up from the Great Recession from previous years, with consequences like high unemployment rates affecting the following years and making economic issues highly important for the 2012 election. Therefore, this context might have forced a different, maybe more rational approach in place of eliciting emotions.
Interestingly, Joe Biden and Donald Trump were found to use both emotional appeals in general and activating emotional appeals more often. Additionally, the emotionality of campaign ads has increased over time. On the one hand, this can nod towards an adaptation of campaigns to social media logic, but on the other hand might also accredit that the candidate and his/her personality and style is an important factor in campaign design. In this case, Donald Trump’s populist rhetoric, which is associated with emotional appeals like fear and anger (see Gerstle & Nai, 2019; Salmela & von Scheve, 2018), could have driven Joe Biden to respond with equal if not heightened emotionality in his ads.
Limitations and future research
There are limitations to this study, with the first one already mentioned: The chosen design was a longitudinal quantitative content analysis from which causal relationships cannot be
derived, but rather assesses trends and developments over time. Moreover, there are gaps and disparities in the research unit population because not all 2004 campaign ads could be
retrieved, and the collection of 2020 campaign ads was different to other election years. Since all 2020 campaign ads were drawn from YouTube and it is not comprehensible whether Stanford University’s Political Communication Lab did archive all offline and online ads, a bias in the sample can’t be fully ruled out. Furthermore, it is unclear, whether the 2012 election with its decreased use of emotional appeals and positive emotions was an exception from the general trend that consequently could have distorted results from data analyses. To address both issues, future studies with the same research design should include all
presidential elections from 2000 onwards, when the internet became widely adopted, and complement or match existing archives of ads with ads uploaded on the candidates’ social media platforms.
Additionally, the posed research question could also be approached through a qualitative content analysis that would allow for a more in-depth examination of how emotional appeals are utilized, and how campaign ads are set up to evoke an emotional reaction. This could primarily be through music or images for example, or in an explicit or implicit manner. Alternatively, conducting interviews with campaign managers and staffers is presumably the best and most direct method in determining the role of social media (logic) in campaign strategy (e.g., Kreiss, 2016; Plasser & Plasser, 2002) and ad production. However, the feasibility of such is arguably low, especially within the scope of this study, and it is uncertain how much information campaign managers can disclose on strategies, which is an essential part of their work.
Besides the above-mentioned limitations, there was a general challenge of coding emotions which also translated to insufficient intercoder reliability for some variables. Here, the difficulty lies in the subjectivity of detecting and assessing emotions, and there are
multiple layers to it. Not only might certain appeals evoke emotional responses in some but not in others, which results in different coding decisions regarding the presence of an emotional appeal, but there are also emotions that are harder to grasp than others. This was the case for enthusiasm, pride and compassion in contrast to anger, fear, sadness and humor/amusement when looking at the intercoder reliability. Thus, it is difficult to
operationalize emotions, especially for audiovisual formats where emotions can be induced on various dimensions. Even though this subjectivity complicates research, it reflects the reality where not every ad will cause the same emotional response in viewers. Perhaps this aspect could be considered in content analyses by adding more coders and running comprehensive training to establish common and essential elements (e.g., explicit statement of emotion) or combination of elements (e.g., explicit statement of emotion paired with emotional music) that make content emotional.
On the whole, this study was not able to provide hard evidence for a change in the usage of emotional appeals due to a shift from traditional mass media to social media. Although there is a positive trend towards more emotional campaign ads, the only alignment with social media logic would be that activating emotions are being used more frequently – and these are also highly relevant emotions to elicit political actions among voters and not only among internet users. Thus, the integration of those emotional appeals can’t be solely attributed to the distribution of ads on social media. Nevertheless, differences between the selected elections were made apparent and emphasized how complex campaign strategy is and the many factors that are taken into account and shape the setup of campaigns.
Altheide, D. (2004). Media Logic and Political Communication. Political Communication, 21, 293-296.
Baek, K., Holton, A., Harp, D., & Yaschur, C. (2011). The Links That Bind: Uncovering Novel Motivations for Linking on Facebook. Computers in Human Behavior, 27, 2243-2248.
Baum, M. (2005). Talking the Vote: Why Presidential Candidates Hit the Talk Show Circuit.
American Journal of Political Science, 49(2), 213-234.
Berger, J. (2011). Arousal Increases Social Transmission of Information. Psychological Science, 22(7), 891-893.
Berger, J., & Milkman, K. (2012). What Makes Online Content Viral? Journal of Marketing Research, 49, 192-205.
Blumler, J., & Kavanagh, D. (1999). The Third Age of Political Communication: Influences and Features. Political Communication, 16, 209-230.
Bodenhausen, G., Kramer, G., & Susser, K. (1994). Negative Affect and Social Judgement:
The Differential Impact of Anger and Sadness. European Journal of Social Psychology, 24, 45-62.
Bor, S. (2014). Using Social Network Sites to Improve Communication Between
Political Campaigns and Citizens in the 2012 Election. American Behavioral Scientist, 58(9), 1195-1213.
Borah, P. (2016). Political Facebook Use: Campaign Strategies Used in 2008 and 2012 Presidential Elections. Journal of Information Technology & Politics, 13(4), 326-338.
Botha, E., & Reyneke, M. (2013). To Share or not to Share: The Role of Content and Emotion in Viral Marketing. Journal of Public Affairs, 13(2), 160-171.
Brader, T. (2005). Striking a Responsive Chord: How Political Ads Motivate and Persuade Voters by Appealing to Emotions. American Journal of Political Science, 49(2), 388- 405.
Brader, T. (2006). Campaigning for Hearts and Minds: How Emotional Appeals in Political Ads Work. Chicago, IL: University of Chicago Press.
Brady, J., Gantman, A., Van Bavel, J. (2020). Attential Capture Helps Explain Why Moral and Emotional Content Go Viral. Journal of Experimental Psychology: General, 149(4), 746-756.
Brady, W., Wills, J., Jost, J., Tucker, J., & van Bavel, J. (2017). Emotion Shapes the Diffusion of Moralized Content in Social Networks. Proceedings of the National Academy of Sciences, 114(28), 7313-7318.
Brants, K., & Van Praag, P., (2006). Signs of Media Logic Half a Century of Political Communication in the Netherlands. Javnost – The Public, 13(1), 25-40.
Brants, K. & Van Praag, P. (2017). Beyond Media Logic. Journalism Studies, 18(4), 395-408.
Cogburn, D., & Espinoza-Vasquez, F. (2011). From Networked Nominee to Networked Nation: Examining the Impact of Web 2.0 and Social Media on Political Participation
and Civic Engagement in the 2008 Obama Campaign. Journal of Political Marketing, 10, 189-213.
Davis, J., Love, T., & Killen, G. (2018). Seriously Funny: The Political Work of Humor on Social Media. New Media & Society, 20(10), 3898-3916.
Dobele, A., Lindgreen, A., Beverland, M., Vanhamme, J., & Van Wijk, R. (2007). Why Pass on Viral Messages? Because They Connect Emotionally. Business Horizons, 50, 291- 304.
Druckman, J., Kifer, M., & Parkin, M. (2009). Campaign Communications in U.S.
Congressional Elections. American Political Science Review, 103(3), 343-366.
Druckman, J., Kifer, M., & Parkin, M. (2010). Timeless Strategy Meets New Medium: Going Negative on Congressional Campaign Web Sites, 2002-2006. Political
Communication, 27(1), 88-103.
English, K., Sweetser, K., & Ancu, M. (2011). YouTube-ification of Political Talk: An Examination of Persuasion Appeals in Viral Video. American Behavioral Scientist, 55(6), 733-748.
Foot, K., & Schneider, S. (2006). Web campaigning. Cambridge, MA: MIT Press.
Franklin Fowler, E., Franz, M., Martin, G., Peskowitz, Z., & Ridout, T. (2020). Political Advertising Online and Offline. American Political Science Review, 115, 130-149.
Franz, M., Franklin Fowler, E., Ridout, T., & Wang, M. (2020). The Issue Focus of Online and Television Advertising in the 2016 Presidential Campaign. American Politics
Research, 48(1), 175-196.
Gibson, R., Wainer, L., & Ward, S. (2005). Online Participation in the UK: Testing a
‘Contextualised’ Model of Internet Effects. British Journal of Politics and
International Relations, 7, 561-583.
Greenfield, H. (2007, January). YouTube Boom May Mean New Jobs on Campaigns.
National Journal Magazine.
Gueorguieva, V. (2009). The Impact of Alternative Communication Channels. In
Gueorguieva, V. (ed.) Politicking Online: The Transformation of Election Campaign Communications (Chapter 13). New Brunswick, NJ: Rutgers University Press.
Heath, C., Bell, C., & Sternberg, E. (2001). Emotional Selection in Memes: The Case of Urban Legends. Journal of Personality and Social Psychology, 81(6), 1028-1041.
Horton, D., & Wohl, R. (1956). Mass Communication and Para-Social Interaction.
Observation on Intimacy at a Distance. Interpersonal and Biological Processes, 19(3), 215-229.
Kaid, L., & Johnston, A. (2001). Videostyle in Presidential Campaigns: Style and Content of Televised Political Advertising. Westport, CT: Praeger.
Karlsen, R. (2009). Campaign Communication and the Internet: Party Strategy in the 2005 Norwegian Election Campaign. Journal of Elections, Public Opinion and Parties, 19, 183-202.
Klinger, U., & Svensson, J. (2014). The emergence of network media logic in political
communication: A theoretical approach. New Media & Society, 17(8), 1-17.
Kreiss, D. (2016). Seizing the Moment: The Presidential Campaign’s Use of Twitter during the 2012 Electoral Cycle. New Media & Society, 18(8), 1473-1490.
Livingston, S. (2009). Foreword: Coming to Terms with ‘Mediatization’. In: Lundby, K.(ed.).
Mediatization: Concept, Changes, Consequences. New York, NY: Peter Lang, ix-xi.
Marcus, G., Neuman, W., & MacKuen, M. (2000). Affective Intelligence and Political Judgement. Chicago, IL: University Chicago Press.
McCullagh, D. (2000, January 29). Web Changing Politics? Not. Wired. Retrieved from http://www.wired.com/politics/law/news/2000/01/33960
McLaughlin, B., Holland, D., Thompson, B., & Koenig, A. (2020). Emotions and Affective Polarization: How Enthusiasm and Anxiety About Presidential Candidates Affect Interparty Attitudes. American Politics Research, 48(2), 308-316.
Nahon, K, Hemsley, J., Walker, S., et al., (2011). Fifteen Minutes of Fame: The Power of Blogs in the Lifecycle of Viral Political Information. Policy & Internet, 3(1), 6-33.
Nai, A., Martiñez I Coma, F., & Maier, J. (2019). Donald Trump, Populism, and the Age of Extremes: Comparing the Personality Traits and Campaigning Styles of Trump and Other Leaders Worldwide. Presidential Studies Quarterly, 49(3), 609-643.
Norris, P. (2000). A Virtuous Circle? Political Communications in Post-Industrial Democracies. Cambridge, UK: Cambridge University Press.
Owen, D. (2018). New Media and Political Campaigns. In K. Kenksi, & K. Hall Jamieson
(Eds.), The Oxford Handbook of Political Communication. New York, NY: Oxford University Press.
Palfrey, J. (2004, December 7). Internet Politics 2004: The Good, the Bad and the Unknown.
Personal Democracy Media. Retrieved from
https://personaldemocracy.com/content/internet-politics-2004-good-bad-and-unknown Panagopoulos, C. (2009). The Digital Pulse of the 2008 Campaign. In Gueorguieva, V. (ed.)
Politicking Online: The Transformation of Election Campaign Communications.
New Brunswick, NJ: Rutgers University Press.
Pew Research Center’s Project for Excellence in Journalism. (2012, August 15). How the
presidential candidates use the Web and social media. Retrieved from
Phelps, P., Lewis, R., Mobilio, L., Perry, D., & Raman, N. (2004). Viral Marketing or Electronic Word-of-Mouth Advertising: Examining Consumer Responses and Motivation to Pass Along Email. Journal of Advertising Research, 44(4), 333-348.
Phillips, W., & Milner, R. (2017). The Ambivalent Internet: Mischief, Oddity, and Antagonism Online. Hoboken, NJ: Wiley.
Plasser, F., & Plasser, P. (2002). Global Political Campaigning: A Worldwide Analysis of Campaign Professionals and Their Practices. Westport, CT: Praeger Publishers.
Raisinghani, M., & Weiss, R. (2011). The Impact of the Internet on Politics: The “Net Effect”
on Political Campaigns and Elections. International Journal of E-Politics, 2(4), 29-40.
Ridout, T., Franklin Fowler, E., Branstetter, J. (2010). Political Advertising in the 21st Century: The Rise of the YouTube Ad. Paper prepared for the annual meeting of the American Political Science Association, September 2-5, 2010, Washington, D.C.
Ridout, T., & Searles, K. (2011). It’s my Campaign I’ll Cry if I want to: How and When
Campaigns Use Emotional Appeals. Political Psychology, 32(3), 439-458.
Salganik, M., Dodds, P., & Watts, D. (2006). Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market. Science, 311, 854-856.
Salmela, M., & von Scheve, C. (2018). Emotional Dynamics of Right- and Left-Wing Political Populism. Humanity & Society, 42(4), 434-454.
Schweitzer, E. (2008). Innovation or Normalization in E-Campaigning? A Longitudinal Content and Structural Analysis of German Party Websites in the 2002 and 2005 National Elections. European Journal of Communication, 23, 449-470.
Skonieczny, A. (2018). Emotions and Political Narratives: Populism, Trump and Trade.
Politics and Governance, 6(4), 62-72.
Strömbäck, J. (2008). Four Phases of Mediatization: An Analysis of Mediatization of Politics.
The International Journal of Press/Politics, 13(3), 228-246.
Sulkin, T., Moriarty, C., & Hefner, V. (2007). Congressional Candidates’ Issue Agendas On- and Off-line. Harvard International Journal of Press/Politics, 12(2), 63-79.
Valentino, N., Gregorowicz, K., & Groenendyk, E. (2007). Emotions, Efficacy, and Political
Participation. Paper presented at the Annual Meeting of the International Society of Political Psychology, Portland, OR.
Valentino, N., Brader, T., Groenendyk, E., Gregorowicz, K., & Hutchings, V. (2011).
Election Night’s Alright for Fighting: The Role of Emotions in Political Participation.
The Journal of Politics, 73(1), 156-170.
Valenzuela, S., Piña, M., & Ramirez, J. (2017). Behavioral Effects of Framing on Social Media Users: How Conflict, Economic, Human Interest, and Morality Frames Drive News Sharing. Journal of Communication, 67, 803-826.
Van Dijck, J., & Poell, T. (2013). Understanding Social Media Logic. Media and Communication, 1(1), 2-14.
Vergeer, M., Hermans, L., & Sams, Steven (2011). Online Social Networks and Micro- Blogging in Political Campaigning: The Exploration of a New Campaign Tool and a New Campaign Style. Party Politics, 19(3), 477-501.
Weber, C. (2012). Emotions, Campaigns, and Political Participation. Political Research Quarterly, 66(2), 414-428.
Wells, C., Van Thomme, J., Maurer, P., et al. (2016). Coproduction or Cooptation? Real-time Spin and Social Media Response during the 2012 French and US Presidential Debates.
French Politics, 14, 206–233.
Williams, C., & Gulati, G. (2009). The Political Impact of Facebook. Evidence from the 2006
Elections and the 2008 Nomination Contest. In Gueorguieva, V. (ed.) Politicking Online: The Transformation of Election Campaign Communications. Rutgers University Press.
Wu, S., Hofman, J., Mason, W., et al. (2011). Who Says What to Whom on Twitter.
Proceedings of the 20th International Conference on World Wide Web, Hyderabad, 28
March – 1 April, pp. 705-714. ACM.
Appendix A Codebook Formal characteristics
ID. List the corresponding ID number of the ad from the sample overview. To be created when drawing the sample
Year. Which election year is the ad from?
• 1 = 2004
• 2 = 2012
• 3 = 2020
Length. Please code the exact length of the ad in seconds, e.g., 65
Candidate. Which candidate is featured/endorsed in the ad? The candidate approving and broadcasting the ad is not necessarily the featured candidate of the ad, this is the case for attack ads. For attack ads code the candidate who produced the ad. Please code first and last name of the candidate.
Party. Which party does the endorsed candidate belong to? Which party is affiliated with the ad? The featured candidate is not necessarily a member of the party the ad is affiliated with, this could be the case for attack ads for example. For attack ads code the party which the candidate who produced the ad belongs to.
• 1 = Democrats
• 2 = Republicans
Election. Please code the election for which the ad was broadcasted.
• 1 = Primary/nomination
• 2 = General presidential election
• 0 = Uncontested race
• 1 = Contested race
Type of sponsor. Please code the sponsor of the ad. This information can usually be found at the end of each ad.
• 0 = Candidate
• 1 = Political party
• 2 = Interest group/PAC
Incumbent. Please code whether the featured/endorsed candidate in the ad is a challenger or incumbent (currently holds the position).
• 0 = Challenger
• 1 = Incumbent
Incumbent party. Please code “incumbent party” if the featured/endorsed candidate belongs to the same party as the current president.
• 0 = Challenger party
• 1 = incumbent party
Tone. Code “promotional” if the ad is predominantly endorsing the candidate him-/herself and highlighting his/her personality and actions. Code “attack” if the ad is predominantly
attacking one opponent and portraying him/her in a negative way. Code “comparative” for both promotional and attack elements.
• 1 = Promotional
• 2 = Comparative
• 3 = Attack
Music. Please code the kind of music that is used.
• 1 = Ad contains uplifting/sweet/sentimental/patriotic music or major chords
• 2 = Ad contains tense/somber music or minor chords
• 3 = Mix of uplifting and tense music or neutral music
• 999 = No music
Sound effects. Please code the type of sound effects that are used.
• 1 = Ad contains positive sound effects (e.g., laughter, cheers, applause, empowering chants)
• 2 = Ad contains negative sound effects (e.g., screams, sirens, crying, aggravating chants)
• 3 = Mix of positive and negative sound effects or neutral sound effects
• 999 = No sound effects
Dominant color. Please code the dominant color scheme of the ad.
• 1 = Black and white
• 2 = Dark or gray colors
• 3 = Ordinary or muted colors
• 4 = Bright colors (“colorful”)
• 5 = Mix of color schemes and none is dominant
Emotional appeal. Please code whether the ad utilizes emotional appeals.
• 0 = No appeal to viewer emotions
• 1 = Appeal to viewer emotions
Dominant appeal. Please code whether emotional or logical appeals are dominant. Logical appeals refer to rationality.
• 0 = Logical appeal dominant
• 1 = Emotional appeal dominant
• 2 = Neither or other appeal dominant
For the following emotion-specific appeals (from amusement to sadness), code “strong appeal” if it is the dominant appeal, or if the appeal is evoked on multiple dimensions through rhetoric, imagery and music, in contrast to only one dimension such as rhetoric for example.
Emotional appeals can both be explicit, e.g., statements like “You should be afraid of…” or “I am proud of this country”, and implicit, e.g., frightening music, use of symbols and imagery.
Amusement/humor appeal. Amusement/humor appeals can include jokes, witty comments, sarcasm, memes, etc. These clips include amusement/humor appeals:
• 0 = No attempt to elicit amusement (appeal to humor)
• 1 = Some appeal to amusement/humor
• 2 = Strong appeal to amusement/humor
Fear appeal. This ad includes fear appeals:
• 0 = No attempt to elicit fear/anxiety
• 1 = Some appeal to fear/anxiety
• 2 = Strong appeal to fear/anxiety
Enthusiasm appeal. Enthusiasm includes a strong excitement of feeling, a feeling of
inspiration, empowerment and hope that things are going well or that they can go well. This ad includes enthusiasm appeals: https://www.youtube.com/watch?v=iWtgWlHGpb8 (mainly words)
• 0 = No attempt to elicit enthusiasm/hope/joy
• 1 = Some appeal to enthusiasm/hope/joy
• 2 = Strong appeal to enthusiasm/hope/joy
Anger appeal. This ad includes anger appeals:
• 0 = No attempt to elicit anger/contempt/disgust
• 1 = Some appeal to anger/contempt/disgust
• 2 = Strong appeal to anger/contempt/disgust
Pride appeal. A pride appeal refers for example to satisfaction in what’s been accomplished or who they are. This ad includes pride appeals:
• 0 = No attempt to elicit pride
• 1 = Some appeal to pride
• 2 = Strong appeal to pride
Compassion appeal. Compassion refers to the sympathetic consciousness of others’ distress and a desire to alleviate it. It can also be understood as a sympathetic pity and concern for the sufferings or misfortunes of others. These ads include compassion appeals:
• 0 = No attempt to elicit compassion/empathy
• 1 = Some appeal to compassion/empathy
• 2 = Strong appeal to compassion/empathy
Sadness appeal. This clip includes sadness appeals:
• 0 = No attempt to elicit sadness/disappointment/regret
• 1 = Some appeal to sadness/disappointment/regret
• 2 = Strong appeal to sadness/disappointment/regret
Candidate expression. Code the dominant expression of the candidate. If the candidate is not present in the ad, code "88".
• 1 = Smiling/laughing
• 2 = Attentive/serious
• 3 = Frowning/glaring
• 4 = Sad/suffering
• 5 = Neutral or mix of expressions
• 88 = Not applicable/candidate not present
• 99 = Other (specify):
Protagonist expression. Determine the dominant protagonist of the ad (e.g., testimonials, people in the background) and then code his/her dominant expression.
• 1 = Smiling/laughing
• 2 = Attentive/serious
• 3 = Frowning/glaring
• 4 = Sad/suffering
• 5 = Neutral or mix of expressions
• 88 = Not applicable/no other person present
• 99 = Other (specify):
Overall emotionality. How emotional is the ad overall? Code very emotional in contrast to somewhat emotional, if emotional appeals are intensified by transporting them on multiple