Emotional Appeals in Political Debates: How Language Emotions Shape Voter Behavior

Do emotional appeals in political debates sway the electorate? This article examines the influence of candidates’ emotional appeals during U.S. Presidential Debates on polling outcomes between 2004 and 2020. Despite the significance of debates in shaping voter perceptions, the specific impact of emotional appeals, defined in terms of negativity and positivity in candidates’ language, remains underexplored. Utilizing a mixed-methods approach, we analyze debate transcripts with the Linguistic Inquiry and Word Count (LIWC-22) tool to quantify emotional appeals and correlated these findings with polling data from RealClearPolitics. Our study diverges from traditional analyses by concentrating on the emotional appeals present in debates and their subsequent effects on voter preferences. Our findings indicate that neither negativity nor positivity in debates significantly alter voter behavior, suggesting the need for a broader understanding of factors influencing voter decisions. This research contributes to the field of political communication by highlighting the limited impact of debate aggressiveness on polling results and by encouraging future studies to explore the relationship between political discourse and electoral outcomes. 

1. Introduction 

Elections in a democratic society embody the collective voice of the electorate, with political debates providing a powerful forum for candidates to connect with and influence voters. Beyond the articulation of policies, debates showcase candidates’ emotional appeals, both positive and negative, which can shape public perception. This article investigates how these emotional tones in language, specifically positivity and negativity, correlate with changes in voter preferences during U.S. presidential debates from 2004 to 2020. 

While existing research has examined the general influence of debates on voter perception, the impact of candidates’ emotional expressions remains less explored. In particular, our study examines whether positive or negative language in debates significantly affects voter polling, aiming to clarify if emotional appeal sways electoral behavior. By focusing on emotional tones beyond aggressiveness alone, this article contributes nuanced insights into political communication, offering practical implications for campaign strategies.

Our analysis draws on a literature review of past studies on debate influence, followed by a context section that clarifies key components of our research question. Using the Linguistic Inquiry and Word Count (LIWC-22) tool, we qualitatively assess emotional tones in debate language, allowing for a precise measurement of positive and negative expressions. Our mixed-methods approach integrates quantitative and qualitative methods to analyze debate transcripts and RealClearPolitics polling data.

Our findings suggest that while positivity in debates may show a slight association with changes in voter preferences, the relationship is not statistically significant. This result challenges the assumption that emotional tone strongly impacts voter behavior, indicating that other, unexplored factors might play a more significant role in shaping electoral decisions. Through our study, we contribute to a deeper understanding of the nuanced relationship between candidates’ emotional language and voter behavior.

2. Literature Review

Impact of Debate Format on Voter Perception and Candidate Evaluation

Mitchell S. Mckinney and Benjamin R. Warner (2013) discussed a similar topic, where they studied how presidential debates function in different campaign contexts, and the election years they looked at are 2000, 2004, 2008, and 2012. In total, their analysis included 6,775 debate viewers, of which 4,308 were general election presidential viewers, 880 vice presidential debate viewers, and 1,587 primary debate viewers. 58% of respondents were female, 41% were male, and 1% did not disclose their gender. The participants consisted of undergraduate students from schools all across the United States, and they were recruited by faculty members. The participants would view the presidential debates in a group setting and would fill out pre and post-debate surveys which asked about their attitudes towards politics.

Mckinney and Warner broke the results down based on the measures that were used to evaluate. For candidate preference/voter choice, they saw that the general election debate had little effect on voter choice with 86% of participants still choosing their original candidate. However, the primary debates had a greater influence on voter choice, with 35% of participants switching their candidate choice. For candidate evaluation, post-debate, there was a positive increase in score for Democratic candidates in all three types of debates, but only a significant increase in score for Republican candidates in the primary and vice presidential debates. Once exposed to all three debate types, there was a significant increase in the participants’ political information efficacy, or PIE: voters’ perceived ability to understand and influence political processes. Finally, the study showed that after viewing the debates, there was an overall decrease in participants’ cynicism (McKinney and Warner 2013).

Furthermore, Diana B. Carlin, Eric Morris, and Shawna Smith (2001) studied how the debate format influences clash between candidates. They specifically studied the 2000 election and used transcripts from the Bush-Gore debates to analyze this topic. They divided the transcripts up into units, where each segment where the candidate spoke uninterrupted was considered a unit. After analyzing these units using nine different category schemes, they found that presidential debates do produce clash, and that each debate format produced different forms of clash (Carlin et al. 2001). Moreover, not only does the debate format influence how the debate goes, general election debates versus primary debates also each have different effects on viewers. Benoit, McKinney, and Stephenson (2002) found that primary debates have a much greater effect than general election debates on change in voter preference (Benoit et al. 2002).

Effects of Emotional Tone and Aggressiveness on Candidate Evaluation and Voter Attitudes

Research has also been done focusing on the aggressiveness of candidates during the election debates. Daniel John Montez and Pamela Jo Brubaker (2019) explored which Republican and Democratic candidates were more punished for showing aggression in primary debates. The authors utilized content analysis and divided the study into two phases. In the first phase, they analyzed and compared two 2016 presidential primary debates from each party, and in the second, they chose two of the three 2016 general election debates. Overall, their results showed that front-runners in the race were the greatest victims of aggression, and that aggression increased as each debate progressed (Montez and Brubaker 2019).

Shelly S. Hinck, Robert S. Hinck, and Edward A. Hinck (2013) coded debate scripts from nine 2012 Republican primary debates according to level of face threat, target of message, and subject of disagreement. It was found that the 2012 Republican primary debates were overall less threatening than general campaign debates, however, when Republican candidates did utilize face threats against each other in the primary debates, their attacks were more intense. Furthermore, it was found that a candidate’s standings in the polls also determines how likely they are to be victims of attacks from other candidates (Hinck et al. 2013). Similarly, McKinney, Kaid, and Robertson (2001) also found that front-runners in the race usually end up receiving the most attacks during the debates (McKinney et al. 2001).

Influence of Debate Exposure on Knowledge, Perception of Candidate Character, and Voting Preference

Kenneth Winneg and Kathleen Hall Jamieson (2017) utilized panel survey data from viewers of the first and third debates to measure their changes in knowledge on policy issues, their belief about a candidate’s character, and how much of a potential threat to the nation each one would be if elected. The researchers ran paired sample t-tests for the tests of the knowledge hypothesis and research questions, as well as to measure any changes in the evaluation of presidential qualification and potential threat to the nation’s well-being. Furthermore, they ran a logistic regression to control for education to further answer the questions about the effect of viewing post-debate coverage.

As for the results, the researchers recorded a positive knowledge gain among viewers regarding issue stances after the first debate. There was little movement of viewers’ assessments of whether each candidate was qualified to serve as president and if elected, would pose a threat to the well-being of the nation (Jamieson and Winneg 2017). On a broader scale, extensive debate research has found that exposure to debates does affect viewer perceptions of candidates. Katz and Feldman (1962), explored the 1960 Kennedy-Nixon debates, where they found that the audience was significantly more focused on analyzing the character of the candidates and their presentations, as compared to their actual stances on policies (Katz and Feldman 1962). Furthermore, Mike Yawn, Kevin Ellsworth, Bob Beatty, and Kim Kahn (1998) focused on a 1996 Arizona Republican primary debate and how it affected voters’ attitudes. They conducted a pretest-posttest quasi-experimental design and they found that the debate led respondents to change their electability and viability assessments of the candidates, which resulted in significant changes in the respondents’ vote preferences (Beatty et al. 1998).

Moderation of Partisan Attitudes Through Debate Exposure

Sarah Brierly, Eric Kramon, and George Kwaku Ofosu (2020) explored the effects of the parliamentary debates on the 2016 Ghana election. Their main research question focused on why debates influence voters’ attitudes. For their sample, they selected 1,991 participants from three different constituencies in Ghana. They randomly assigned participants to different segments of the debate (their treatment) which corresponded with different potential causal channels. The five treatments that each participant could be assigned to were a placebo video, a personal background segment, a policy segment, a full debate video, and a full debate audio. Their outcomes were measured through surveys given to each person immediately after they received their treatment. The most important result from this study concluded that debates moderated the political attitudes of partisans, which made them more favorable towards members of another party as opposed to wanting to vote for co-partisan candidates from their own party (Brierley et al. 2020).

Context

All of these studies explore important issues surrounding presidential debates and have done a tremendous job in providing more insight as to the effects these debates truly have. Much of the research mentioned above focuses on the broader aspect of how presidential debates (general or primary) swing voter opinions, while some of the research hones in on specific characteristics and mannerisms of presidential candidates participating in the debates.

Contrary to a lot of these studies, Vincent Pons and Caroline Le Pennec (2019) offer a different perspective on the influence of debates, arguing that debates may have limited impact on voters’ final decisions. Their study contends that while debates are high-profile events, they often do not change voter preferences significantly due to the existing familiarity voters have with candidates and their positions. They suggest that dramatic moments in debates are unlikely to sway undecided voters significantly, and other, non-debate factors such as campaign messaging and broader political events likely play a more substantial role (Pons and Le Pennec 2019). This contradiction indicates more research is needed. 

In our research, we are eager to explore how aggressive mannerisms displayed by a candidate impact the support they garner from voters that view the debates. Although prior research has explored the aggressiveness of candidates on the debate floor, it has mostly focused on why aggressiveness occurs during debates and who ends up falling victim to it the most. There has yet to be substantial research on the overarching issue of how aggressiveness affects polling numbers post-debate, something that we believe is an important and pressing topic that needs to be explored.

Our research is situated within the broader context of political communication, an area that has garnered significant interest for its profound impact on voter behavior and decision-making. In recent years, the nature of political discourse, especially in presidential debates, has evolved, prompting a re-evaluation of its effects on the electorate. Presidential debates, serving as pivotal platforms for candidates to present their policies and personalities, have a profound influence on shaping public opinion and voter preferences.

The concept of positive and negative emotions in political debates is not new, yet it remains a relatively underexplored dimension in political science research. Historically, debates have been analyzed for their content, rhetorical strategies, and overall impact on election outcomes. However, the specific role of emotional expression in debates—defined in terms of the tone, language, and assertiveness of candidates—has not been thoroughly examined, particularly in relation to its direct impact on voter preferences.

This gap in the literature becomes even more pronounced when considering the changing landscape of political communication. The rise of social media, the 24-hour news cycle, and the increasingly polarized political environment have transformed how candidates communicate and how their messages are received by the public. These changes have made it imperative to re-examine traditional assumptions about political discourse, including the effectiveness of aggressive tactics in debates.

Our study aims to bridge this gap by focusing on recent U.S. Presidential Debates, specifically the debates chosen for their relevance in today’s political climate and the availability of extensive data, including debate transcripts and voter preference surveys (Martherus 2020). By analyzing these debates, we seek to understand how the aggressive rhetoric used by candidates influences voter opinions in the context of contemporary political dynamics.

3. Design

This study is structured around a pivotal aspect of political communication: the interplay between a candidate’s emotional expression in presidential debates and the electorate’s subsequent voting behavior. Central to our investigation is the hypothesis: “Does the emotional tone conveyed in presidential debates correlate with shifts in voter preferences?” Our objective is to illuminate the potential causal relationships between the emotional tenor of candidates during debates—specifically, their use of negative versus positive language—and changes in voter intention.

We looked at general debates from 2004 – 2020 because that is when Real Clear Politics (Real Clear Politics 2019) started their polling on general elections. We also used some, but not all, primary debates from 2008 – 2020 because the transcripts dataset (Martherus 2020) did not cover them all, and Real Clear Politics (RCP) did not start their polling on primaries until 2008. The selection of debates within our dataset was ultimately subject to the constraints of our RCP polling and debate transcripts. The total entries analyzed were 183 (where 153 were primary and 28 were general) from 39 debates.

Table 1: Data Description

Year	Type	Coverage in Dataset	RCP Polling Availability	Notes
2004	General	Yes	Yes	First year of RCP polling on general elections
2008	General	Yes	Yes	Full RCP polling and transcripts available
2008	Primary	Partial	Yes	Limited transcripts coverage for primaries
2012	General	Yes	Yes	Complete coverage
2012	Primary	Partial	Yes	Limited transcripts coverage for primaries
2016	General	Yes	Yes	Complete coverage
2016	Primary	Partial	Yes	Limited transcripts coverage for primaries
2020	General	Yes	Yes	Complete coverage
2020	Primary	Partial	Yes	Limited transcripts coverage for primaries

The debate selection is therefore focused on periods when the proliferation of the internet and social media was expanding, though at different rates over the years. Consequently, the impact of debates on voting intentions may have evolved across this time. As a result, the conclusions of this article, like those of many studies in political communication, are subject to potential bias stemming from the increasing role of social media.

We operationalize emotional expression using the Linguistic Inquiry and Word Count (LIWC-22) tool, which serves as a robust instrument for parsing psychological constructs within spoken or written text. The LIWC-22 (Linguistic Inquiry and Word Count) tool is a text analysis software designed to quantify emotional, cognitive, and structural components of language. It works by comparing input text to an internal dictionary with thousands of words categorized by psychological and linguistic attributes, allowing it to identify and measure tone, emotions, and topics within text. Commonly used in psychology, social sciences, and marketing, LIWC-22 analyzes tone in political speech by identifying word categories like positive (e.g., “success,” “achieve”) and negative (e.g., “failure,” “risk”). By applying LIWC-22 to the transcripts of the U.S. Presidential Debates, we quantify the levels of negativity and positivity, thereby offering a measurable proxy for the candidates’ aggressiveness and positivity in their debate performances. We define aggressiveness as the prevalence of negative emotional words relative to the total word count, while positivity is measured similarly through the incidence of positive emotional expressions. In other words, aggressiveness is equated to the negativity score.

In our analysis of the dataset (Martherus 2020), we considered the dynamics between debate sentiments and voting behavior changes. Our analysis began with a focus on the role of negativity in political debates. Within the context of this study, our use of the word ‘negativity’ refers to the use of critical or negative words, critical language, or sentiments by participants in a political debate. Using language that criticizes opponents, draws attention to shortcomings, or highlights the negative elements of ideas or initiatives is considered negative language.

The dataset constructed for this research, PS169 Dataset: Real Clear Politics, encompasses variables such as the date of the debate, type (primary or general), the order of the debate, candidate names, polling figures before and after the debates, and the calculated negativity, positivity, and combined emotional tone scores. Polling data, reflecting voter sentiment 24 hours prior and 48 hours subsequent to each debate, were sourced from RealClearPolitics daily averages. This time frame was chosen to ensure a sufficiently responsive window for capturing the influence of debates on the electorate’s preferences.

Our analysis employs a mixed-method approach. The core of our quantitative strategy involves a series of regression analyses—specifically focusing on negativity, positivity, and their aggregated impact (negativity + positivity)—to investigate their correlation with the change in votes. We used a linear regression model to find general trends in correlation. We did not investigate other models given the small magnitude of data points.

Qualitatively, we supplement our analysis by examining the emotional outliers found in political debates. As we analyze and interpret the quantitative data to see whether or not changes in voter preferences are correlated with the emotional tone of discussions, we evaluate the tone, vocabulary, and general emotional content of the comments made by the candidates during debates. The qualitative method involves identifying and analyzing situations in which the emotional tone of debates deviates from the typical range. Finding outliers encourages more qualitative research into certain situations affecting candidates’ speech: emphasizing associations found in candidates’ language use and emotional expression categories such as “power,” and “family”. These approaches allow readers to juxtapose the quantitative LIWC-22 findings with the qualitative narratives emerging from the coverage. The integration of these perspectives will enable us to present a comprehensive view of the extent to which debate sentiments sway voter decisions.

Our research is backed by a qualitative evaluation of the findings of scholars who have studied the function of language in political discourse and its possible influence on voter behavior. As we discussed earlier, the experts that we have chosen as an instrument to strengthen our findings are Pons and Le Pennec (Pons and Le Pennec 2019). The experts were chosen on the basis of their political science specialization, academic connections, and particular contributions to research on the influence of political disputes. The purpose of the qualitative assessment is to determine whether the conclusions of the chosen experts corroborate our theory.

 

4. Results

Contrary to the initial hypothesis, our findings revealed that negativity in debates did not significantly influence voting changes, as evidenced by a p-value of 0.731. Using a linear regression model to find a basic correlation trend, the estimated coefficient of -0.02734 hinted at a slight negative impact on vote changes with increased negativity, but the statistical insignificance of this result, coupled with a low R-squared value (0.001364), suggested limited explanatory power of negativity alone. This was visually corroborated in Figure 1, which depicted no discernible trend.

 

Shifting our focus to positivity, we investigated its potential effects on voting behavior changes using that model. Intriguingly, the positivity coefficient of 0.08297 indicated a mild positive correlation with changes in votes. However, the non-significant p-value of 0.290 and a negligible adjusted R-squared value (0.001511) emphasized the lack of robust statistical backing for positivity’s influence on voting changes— despite a minor correlation as seen in Figure 2.

 

Our exploration extended to understanding the combined effect of negativity and positivity on voting behaviors. Recognizing that emotional tones in debates often interplay, impacting audience perceptions in a multifaceted manner, we summed these two variables to capture the overall sentiment expressed. This approach was underpinned by the rationale that the cumulative emotional tone, encompassing both positive and negative elements, might offer a more holistic view of the debate’s emotional landscape and its potential influence on voter decisions. The sum of these sentiments, represented by a coefficient of 0.03420, however, did not significantly affect voting changes (p-value: 0.577). The adjusted R-squared value remained low (-0.00786), indicating that even when considering the combined sentiment factors, the model could not adequately explain the variance in voting changes. This was visually corroborated in Figure 3, which did not display any clear trend, reinforcing the notion that other factors, possibly beyond the scope of debate sentiments, play a more crucial role in influencing voting behavior. 

Lastly, we investigated whether the effects of negativity and positivity varied between primary and general debates. Through models incorporating interaction terms between debate types and the sentiment variables, we found no significant differences in effects across debate types. The p-values for the interaction terms (0.230, 0.730, 0.557) indicated that the relationship between debate sentiments and voting changes was consistent, irrespective of the debate being primary or general. This was visually echoed in Figures 4 – 6, where no distinct patterns emerged. 

5. Discussion

We find compelling evidence that neither negativity nor positivity nor their combination significantly influences changes in voting patterns. This consistency was observed across both primary and general debates. The findings underscore the likelihood of other, unexplored factors playing a more decisive role in swaying voting behaviors. These results encourage a broader perspective in political science research, acknowledging the complex and multifaceted nature of political discourse and its impact on voter decisions. Future research may benefit from incorporating a wider array of variables and embracing diverse analytical approaches to unravel the nuanced layers of political influence on voter behavior. 

The results of our analysis have provided valuable insights into the complex relationship between emotional sentiments expressed in political debates and changes in voting behavior. While our findings did not yield significant correlations between negativity, positivity, or their combination and shifts in voter preferences, they raise important questions and offer opportunities for a nuanced understanding of political communication and its impact on electoral outcomes. 

As for the studies that came before ours, none of them truly explored how much of a factor the overall aggressiveness of a candidate plays in their polling numbers. Although we found that aggressiveness plays no significant influence on voting patterns, it helps further the studies conducted by previous researchers on this topic, particularly by helping researchers such as McKinney and Warner realize that the shift in voting patterns they saw in their studies is likely not correlated to the negativity or positivity of candidates. Our findings illustrate that although some candidates might be more prone to direct attacks, or may even be the ones doing the attacking, it provides them with no significant boost in the polling numbers. This information can inform research on the mannerisms and characteristics of candidates, such as the study conducted by Carlin, Morris, and Smith.  

Moreover, it is crucial to acknowledge the multifaceted nature of political discourse. The absence of a strong connection between negativity and changes in voting patterns challenges the prevailing notion that a candidate’s use of negative language in debates directly sways the electorate. This suggests that voters may be more discerning and resilient in the face of adversarial rhetoric, with their choices driven by a broader array of considerations. It prompts us to question whether the electorate is becoming desensitized to negativity or if other factors, such as policy positions, candidate credibility, or external events, play more significant roles in influencing their decisions. 

The lack of a statistically significant impact of positivity on voting changes also invites further scrutiny. While our analysis hinted at a mild positive correlation, it was not strong enough to draw concrete conclusions. This result may suggest that voters prioritize other aspects of a candidate’s performance or message over the level of positivity displayed during debates. It prompts us to explore whether voters are more responsive to substantive policy proposals, relatability, or authenticity in candidates, rather than their overall tone. Additionally, the absence of a clear distinction between primary and general debates in our findings raises questions about the extent to which different debate contexts influence voter responses to positivity. 

Our investigation into the combined effect of negativity and positivity highlights the complexity of emotional expression in debates. The lack of a significant impact suggests that the interplay between these emotional dimensions does not strongly shape voter preferences. However, this outcome encourages us to delve deeper into the nuances of emotional discourse. Are certain combinations of negativity and positivity more impactful than others? Do voters respond differently when candidates balance negative and positive messages? These questions prompt us to consider the subtleties of emotional expression and how they may affect voter decision-making. 

We were able to support our claim using the findings of Vincent Pons and Caroline Le Pennec. Before an election, candidates in the United States campaign for several months, providing voters enough time to make up their minds (Pons and Le Pennec 2019). Any dramatic moments during a discussion usually have a limited impact and disappear before election time. We also discovered that debates tend to be brief affairs with little bearing on people’s decisions. Voters take into consideration a wider range of information, of which the material from debates is only one part. Since voters are already familiar with the candidates, the final presidential debate will have less significance. In summary, our findings juxtaposed with those of these experts, indicating that although a candidate may exhibit aggressive language during a debate, the overall impact of debates on election results is very small, with other, non-debate factors exerting a more significant influence. 

Notably, the visual data exhibited a greater spread of outliers in terms of emotional expression levels during primary debates as compared to general elections. This could be indicative of a more polarized or energized discussion within primary debates, possibly reflecting the strategic behavior of candidates aiming to appeal to more partisan subsections of the electorate, in line with the Median Voter Theory. The Median Voter Theory explains that candidates tend to moderate their positions in general elections to appeal to a broader voter base, whereas in primaries, they may adopt more extreme positions to galvanize the base within their own party (Cukierman and Spiegel 2019).

The observed outliers in our analysis potentially signal instances where the emotional intensity of debates diverges from the expected norm, offering fertile ground for further investigation. These atypical cases might shed light on the circumstances under which candidates depart from the median voter theorem’s predictions, perhaps due to the perceived benefits of energizing their core supporters or the influence of contemporary political dynamics. Further insights from the Linguistic Inquiry and Word Count (LIWC) program should be included into research studies in order to get a more sophisticated comprehension of qualitative data. By offering a methodical and quantitative examination of language, LIWC enables scholars to explore the complexities of communication in more detail. Beyond conventional qualitative techniques, LIWC provides a methodical way to find patterns in texts about emotional tone, social behavior, and cognitive processes. We are able to improve the integrity of our findings by adding objective language metrics to the study by incorporating LIWC insights into the research publications. As we examine intricate subjects like social dynamics, sentiment analysis, or the psychological components of language, this tool becomes quite helpful. 

Throughout this study, we were able to achieve insights regarding our candidates using the Linguistic Inquiry and Word Count (LIWC) program which provided a sophisticated analysis of qualitative data. By offering a methodical and quantitative examination of language, LIWC enables scholars to explore the complexities of communication in more detail. Beyond conventional qualitative techniques, LIWC provides a methodical way to find patterns in texts about emotional tone, social behavior, and cognitive processes. As we examine intricate subjects like social dynamics, sentiment analysis, or the psychological components of language, this tool becomes quite helpful. 

The LIWC data showcased various attributes of debates that are worth mentioning. Within our data from the LIWC program, we noticed that some political candidates stood out in various categories compared to others. For example, John Kasich, a Republican candidate in the 2016 primaries, had a notably larger score in the ‘family’ category (score of .48) compared to his opponents. Language related to familial responsibilities, relationships, feelings, and activities tend to fall under this category. A political candidate’s deliberate attempt to establish a personal and emotional connection with people is usually seen when they bring up family issues repeatedly in their speeches or other communications. By showcasing relevant parts of their life and ideals, political candidates may humanize themselves through family-related discussions.

Former President Barack Obama was shown to have the highest average in the ‘pro-social’ section compared to the other candidates within the same category. In LIWC, language that exhibits empathetic characteristics is classified as ‘pro-social’. Obama may decide to take a pro-social position during debates for a variety of calculated reasons, frequently as an effort to win over supporters. This ties in with his overall stances as he frequently discusses topics pertaining to economic and social inequality. His policies supported social programs that help disadvantaged populations, fought for a fair tax system, and attempted to reduce economic disparity.

Another candidate score that we thought was worth mentioning was John Kerry, a Democratic candidate for the 2004 general election. He received a high average in the ‘power’ category compared to other candidates, and according to a journal posted by Yla Tausczik and James Pennebaker on the meaning of words in LIWC, the power category is intended to record expressions and vocabulary related to authority, control, and influence (Tausczik et al. 2010). The 2004 presidential election was a close race between Kerry and former President George W. Bush, and many citizens were in search of a powerful leader as the United States was at the head front of the Iraq War. 

Applying language that is associated with power can convey a sense of assertiveness and leadership. It’s possible that Kerry had to project a strong, assured image in order to reassure people about his capacity for effective management. Although our results did not find a significant connection between the use of various types of emotional affect and shifts in voter preferences, it was interesting for us to discuss the various correlations we found along the way, inviting future researchers to investigate them. 

It is worth noting that our study focused on a specific time frame, primarily examining debates between 2004 and 2020. Future research should expand this temporal scope to capture evolving trends in political communication. The rise of social media and the changing media landscape have transformed the ways in which candidates communicate with the public. Exploring the impact of emotional sentiments in televised debates prior to this evolving context could yield different results. 

Furthermore, we recommend that future researchers explore the other variables that the Linguistic Inquiry and Word Count (LIWC-22) tool can produce, building on the aggregate dataset we have constructed. LIWC-22 offers a wide range of linguistic and psychological constructs beyond negativity and positivity. Investigating variables such as cognitive processes, social dynamics, or even more granular emotional categories could provide deeper insights into the relationship between language, emotions, and voting behavior. Building on the foundation of our dataset, researchers can uncover new dimensions of political communication that may influence voter decisions. 

6. Conclusion

In conclusion, our research on the impact of a candidate’s emotional expressions in debates on polling outcomes has produced findings that advance our knowledge of the dynamics of political communication. Our findings disproved conventional thinking by showing that changes in voting patterns were not significantly affected by the negativity or positivity stated by candidates. Although there was very little positive association between positivity and changes in votes, the absence of statistical significance calls into question the assumed impact of emotional tone on voter choices.

Within the broader context of political communication research, our work filled a crucial gap by concentrating on the role that emotional expressions have in debates. We used regression analysis concentrating on the years 2004 to 2020 and using the Linguistic Inquiry and Word Count (LIWC-22) tool. A thorough investigation of the emotional terrain of discussions and its possible influence on voter intentions was made possible by this research approach. The lack of noticeable effects of negativity, positivity, or their combination of changes in voting patterns raises the possibility that other, unidentified factors may be more important in determining voter behavior.

We backed this finding with our qualitative analysis drawn from the literature of Pons and Le Pennec (Pons and Le Pennec 2019). Our findings cast doubt on beliefs regarding the direct influence of emotional language on voters and motivate future investigators to explore the diverse aspects of political conversation. This emphasizes the necessity of more research into the developing patterns in political communication, particularly in light of the shifting media environment and the emergence of social media, as well as the applicability of sentiment analysis tools in social science: a contribution this study advances. Our research opens the door for more investigation and improvement of our knowledge of political communication in democracies by offering insightful contributions to the current conversation on how presidential debates affect voters’ judgments.

References

Benoit, W., McKinney, M., and Stephenson, M. (2002). Effects of Watching Primary Debates in the 2000 U.S. Presidential Campaign. Journal of Communication, 52(2), 316–331.

Brierley, S., Kramon, E., and Ofosu, G. K. (2020). The Moderating Effect of Debates on Political Attitudes. American Journal of Political Science, 64(1), 19-37.

Carlin, D. B., Morris, E., and Smith, S. (2001). The Influence of Format and Questions on Candidates’ Strategic Argument Choices in the 2000 Presidential Debates. American Behavioral Scientist, 44(12), 2196–2218.

Cukierman, A., and Spiegel, Y. (2019). Majority Voting and Median Voter Theorem. https://www1.cmc.edu/pages/faculty/LGrant/ref_cards_full/Topic_referenceSheet_med_voter_thrm.pdf.

Montez, D. J., and Brubaker, P. J. (2019). Making Debating Great Again: U.S. Presidential Candidates’ Use of Aggressive Communication for Winning Presidential Debates. Argumentation and Advocacy, 55(4), 282–302.

Katz, E., and Feldman, J. J. (1962). The Debates in the Light of Research: A Survey of Surveys. In The Great Debates, edited by S. Kraus. Bloomington, IN: Indiana University Press.

Martherus, J. (2020). Introducing the Transcripts of U.S. Presidential Debates Data Set. https://ssrn.com/abstract=3611815.

McKinney, M. S., Kaid, L. L., and Robertson, T. A. (2001). The Front-Runner, Contenders, and Also-Rans: Effects of Watching a 2000 Republican Primary Debate. American Behavioral Scientist, 44(12), 2232–2251.

McKinney, M. S., and Warner, B. R. (2013). Do Presidential Debates Matter? Examining a Decade of Campaign Debate Effects. Argumentation and Advocacy, 49(4), 238–258.

Pons, V., and Le Pennec, C. (2019). How Do Campaigns Shape Vote Choice? https://www.nber.org/system/files/working_papers/w26572/revisions/w26572.rev2.pdf.

RealClearPolitics. (2012). Republican Presidential Nomination – Polls. https://www.realclearpolitics.com/epolls/2012/president/us/republican_presidential_nomination-1452.html.

RealClearPolitics. (2016). Republican Presidential Nomination – Polls. https://www.realclearpolitics.com/epolls/2016/president/us/2016_republican_presidential_nomination-3823.html.

RealClearPolitics. (2020). Democratic Presidential Nomination – Polls. https://www.realclearpolitics.com/epolls/2020/president/us/2020_democratic_presidential_nomination-6730.html.

RealClearPolitics. (2019). General Election: Trump vs. Biden. https://www.realclearpolitics.com/epolls/2024/president/us/general-election-trump-vs-biden-7383.html.

RealClearPolitics. (2016). RealClearPolitics – 2016 Election Maps – Battle for White House. https://www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton-5491.html.

RealClearPolitics. (2012). General Election: Romney vs. Obama. https://www.realclearpolitics.com/epolls/2012/president/us/general_election_romney_vs_obama-1171.html.

RealClearPolitics. (2008). Polling Data – General Election: McCain vs. Obama. https://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html.

RealClearPolitics. (2004). Presidential General Election: Bush vs. Kerry. https://www.realclearpolitics.com/epolls/2004/president/us/general_election_bush_vs_kerry-939.html.

Hinck, S. S., Hinck, R. S., Dailey, W. O., and Hinck, E. A. (2013). Thou Shalt Not Speak Ill of Any Fellow Republicans? Politeness Theory in the 2012 Republican Primary Debates. Argumentation and Advocacy, 49(4), 259–274.

Tausczik, Y. R., and Pennebaker, J. W. (2010). The Psychological Meaning of Words: LIWC and Computerized Text Analysis Methods. Journal of Language and Social Psychology, 29(1), 24–54.

Winneg, K., and Jamieson, K. H. (2017). Learning From the 2016 U.S. General Election Presidential Debates. American Behavioral Scientist, 61(4), 362–378.

Yawn, M., Ellsworth, K., Beatty, B., and Kahan, R. (1998). How a Presidential Primary Debate Changed Attitudes of Audience Members. Political Behavior, 20, 155–181.