University of Southern California Dissertations and Theses

Signaling identity: how race and gender shape what representatives say online

(USC Thesis Other)

Signaling identity: how race and gender shape what representatives say online

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content SIGNALING IDENTITY:
HOW RACE AND GENDER SHAPE WHAT REPRESENTATIVES SAY ONLINE
by
Whitney Hua
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POLITICAL SCIENCE AND INTERNATIONAL RELATIONS)
December 2023
2023 Whitney Hua

ii
Table of Contents
Table of Contents ....................................................................................................................................... ii
List of Tables............................................................................................................................................... v
List of Figures ..........................................................................................................................................viii
Abstract ........................................................................................................................................................x
Chapter 1......................................................................................................................................................1
Introduction.................................................................................................................................................1
Congressional Outreach and the Wild, Wild Web ............................................................................4
New Media and the Rhetorical Race to Power..........................................................................5
Speaking and Leading with Negativity......................................................................................8
Same Emotions, Differing Risks: Emotional Expression as a Relational Social Process............11
Research Motivation ............................................................................................................................17
Purpose of Study ..................................................................................................................................19
Design of Study ....................................................................................................................................21
Outline of Contents ..............................................................................................................................23
Chapter 2....................................................................................................................................................26
Costly Signals: Higher Risks, Uncertain Rewards ...............................................................................26
Identities Presented, Identities Perceived.........................................................................................27
Selective Signaling: Identities Picked and Primed for the Norm .........................................29
The Strategic Use of Emotional Appeals...........................................................................................29
Negative Discrete Emotions.......................................................................................................31
Positive Emotions ........................................................................................................................34
Positioning, Privilege, & Strategic Communication........................................................................36
Underrepresented Social Identities....................................................................................................38
Gender Effects ..............................................................................................................................40
Race Effects ...................................................................................................................................41
Moderating Influence of Partisan Norms .........................................................................................44
A Dynamic Model of Congressional Communication....................................................................47
Chapter 3....................................................................................................................................................54
Data and Methods ....................................................................................................................................54
Congressional Tweets as Data............................................................................................................55
The Curious Case of Twitter......................................................................................................56

iii
Data Collection.............................................................................................................................62
Operationalizing Key Terms and Concepts .....................................................................................63
Threat Perception in Competitive Electoral Contexts ............................................................63
Explanatory Variables.................................................................................................................64
Analytic Methods .................................................................................................................................68
Measuring Emotion-Associated Rhetoric ................................................................................68
Multivariate Regression Analysis of Rhetorical Strategies....................................................73
Chapter 4....................................................................................................................................................74
Racing on a Staggered Track: Race-Based Constraints on Negative Emotive Appeals .................74
Navigating Risk: Strategic Communication in an Evolving Congress.........................................75
Who Gets to Go Negative?..................................................................................................................78
Negativity’s Thorns..............................................................................................................................80
The Road Less Privileged....................................................................................................................83
Examining the Interaction of Race & Gender..........................................................................89
Race & Gender Constraints through Partisan Lens................................................................91
Party Moderating Effect – Separate RGIDs Subset by Party.................................................98
Visualizing the Effects of MCs’ Intersecting Race & Gender Identities by Party...............99
House of Anger..........................................................................................................................110
Chapter 4 Summary ...........................................................................................................................113
Summary: Representative Negativity? ...........................................................................................115
Chapter 5..................................................................................................................................................118
Feeling the Heat: Electoral Threat, Institutional Positioning, and Challenger Negativity ..........118
2020 Incumbent and Challenger Campaign Tweets .....................................................................121
Challenging Expectations in Challenger Rhetoric.........................................................................124
Challenging With and Without Privilege .......................................................................................126
Challengers Analysis - Democrat vs. Republican Models with RGID and Party
Interaction Term.........................................................................................................................128
Incumbents Analyses ................................................................................................................132
Who Runs Lowest? Challengers, Incumbents, and Social Identity ............................................142
Claiming a Seat, Finding a Voice: Rhetorical Shifts After Electoral Victory .............................145
Her Disgust, His Disgust: Negativity’s Qualitative Gap..............................................................147
Chapter 5 Summary ...........................................................................................................................150
Chapter 6..................................................................................................................................................154
On a Positive Note ..................................................................................................................................154
Positively Constrained: The Challenge of Challengers ................................................................159
A Positive Spin: Incumbent Messaging ..........................................................................................163
113th to 116th Women Stay Positive ...............................................................................................165
Anticipation Proclamation—BIPOC Women’s Keys to the House.............................................182
Chapter 6 Summary ...........................................................................................................................186

iv
Constrained to Taking the High Road....................................................................................186
Identities of the Constrained....................................................................................................187
Discerning Positive Valenced Emotions.................................................................................188
Chapter 7..................................................................................................................................................190
Conclusion ...............................................................................................................................................190
Summary of Findings ........................................................................................................................192
Significance of Study.................................................................................................................197
The Social Underpinnings of Congressional Communication ....................................................198
The Negativity Divide and the Uneven Platform of Digital Media...................................200
Tweeted Boundaries: Constrained through the Prism of Race and Gender.....................202
Signaling Identity: Advancing a Dynamic Framework of Strategic Communication .............205
Limitations and Future Directions...................................................................................................207
Concluding Thoughts ........................................................................................................................211
References ................................................................................................................................................213
Appendix .................................................................................................................................................241

v
List of Tables
Table 1. Number of tweets and MCs by race and gender (113th to 116th Congress).....................60
Table 2. Time frame and average number of tweets per MC by Congress and period..................60
Table 3. Number of candidates, accounts, and tweets in 2020 congressional campaign tweet
dataset by race and gender (RGID)........................................................................................................62
Table 4. Number of k features by NRC sentiment or emotion category...........................................72
Table 5. OLS regressions of all 113th to 116th MCs’ monthly average use of negative valenced
emotions on Twitter (RGID as factor levels).........................................................................................85
Table 6. Examples of fear-associated rhetoric in 113th to 116th MC office account tweets. .........88
Table 7. OLS regressions of all 113th to 116th MCs’ monthly average use of negative valenced
emotions on Twitter (woman*white).....................................................................................................90
Table 8. Model estimates of RGID*Republican effects from separate regressions of 113th to
116th MC monthly average use of negative valenced rhetoric..........................................................93
Table 9. Examples of disgust-associated rhetoric in 113th to 116th MC office account tweets.....96
Table 10. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of negative sentiment...........................................102
Table 11. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of angry appeals. ..................................................103
Table 12. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of disgust appeals.................................................104
Table 13. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of fear appeals.......................................................105
Table 14. Examples of anger-associated rhetoric in 113th to 116th MC office account tweets....113
Table 15. OLS regressions of all 2020 House candidates’ monthly average use of negative
valenced emotions on Twitter (RGID as factor levels)......................................................................123
Table 16. Total number of unique tweets, month observations, accounts, and candidates by
incumbency status, race/gender, and party.......................................................................................126

vi
Table 17. OLS regressions of 2020 House challenger candidates’ monthly average use of
negative valenced emotions on Twitter (RGID as factor levels)......................................................128
Table 18. Model estimates of RGID*Republican effects from separate regressions of 2020
challenger candidates’ monthly average use of negative valenced rhetoric. ................................131
Table 19. OLS regressions of 2020 House incumbent candidates’ monthly average use of
negative valenced emotions on Twitter (RGID as factor levels)......................................................133
Table 20. Model estimates of RGID*Republican effects from OLS regressions of 2020
incumbents’ monthly logged average use of negative valence rhetoric in office tweets.............135
Table 21. Effect of incumbency status on 2020 House candidates’ monthly average use of
negative sentiment, anger, disgust, and fear appeals by RGID.......................................................143
Table 22. Examples of disgust-associated rhetoric in 2020 House candidate tweets....................149
Table 23. OLS regressions of all 2020 House candidates’ monthly average use of positive
valenced emotions on Twitter (RGID as factor levels)......................................................................157
Table 24. OLS regressions of 2020 House challenger candidates’ monthly average use of
positive valenced emotions on Twitter (RGID as factor levels).......................................................160
Table 25. Model estimates of RGID*Republican effects from separate regressions of 2020
challenger candidates’ monthly average use of positive valenced rhetoric...................................162
Table 26. Model estimates of RGID*Republican effects from separate regressions of 2020
incumbent candidates’ monthly average use of positive valenced rhetoric..................................164
Table 27. OLS regressions of all 113th to 116th MCs’ monthly average use of positive valenced
emotions on Twitter (woman*white)...................................................................................................166
Table 28. OLS regressions of all 113th to 116th MCs’ monthly average use of positive valenced
emotions on Twitter (RGID as factor levels).......................................................................................168
Table 29. Model estimates of RGID effects (subset by party) from separate regressions of
113th to 116th Dem. & Rep. MCs’ monthly average use of positive sentiment. ...........................171
Table 30. Model estimates of RGID effects (subset by party) from separate regressions of
113th to 116th Dem. & Rep. MCs’ monthly average use of anticipation appeals. ........................174
Table 31. Model estimates of RGID effects (subset by party) from separate regressions of
113th to 116th Dem. & Rep. MCs’ monthly average use of joy appeals. ........................................175
Table 32. Model estimates of RGID effects (subset by party) from separate regressions of
113th to 116th Dem. & Rep. MCs’ monthly average use of trust appeals. .....................................176
Table 33. Subgroup main effects from separate OLS regressions of all 113th to 116th MCs’
monthly logged average use of positive valenced emotions on Twitter........................................179

vii
Table 34. Subgroup main effects from separate OLS regressions of 2020 challenger
candidates’ monthly logged average use of positive valenced emotions on Twitter...................181
Table 35. Examples of anticipation-associated rhetoric in 113th to 116th MC office account
tweets........................................................................................................................................................184

viii
List of Figures
Figure 1. Simplified diagram of main processes & concepts in theoretical framework.................15
Figure 2. Detailed theoretical framework of strategic congressional communication. ..................52
Figure 3. Monthly average number of tweets per MC (113th to 116th Congresses).......................59
Figure 4. Diagram of Plutchik’s (2001) proposed set of eight basic emotions and
corresponding emotional states..............................................................................................................71
Figure 5. Diagram of individual perception and appraisal process of risk environment for
minority RGIDs from framework...........................................................................................................77
Figure 6. Plot of minority RGID (factor levels) effects compared to baseline level (white male)
from regressions of all 113th to 116th MCs’ monthly average use of negative valenced
emotions. ....................................................................................................................................................86
Figure 7. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of negative sentiment per tweet. ..........................................................102
Figure 8. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of anger-associated rhetoric per tweet. ...............................................103
Figure 9. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of disgust rhetoric. ...............................................104
Figure 10. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of fear rhetoric. .....................................................105
Figure 11. Plot of correlation matrix coefficients between emotion categories. ............................111
Figure 12. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of negative appeals. ....................................................................138
Figure 13. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of angry appeals. .........................................................................139
Figure 14. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of disgust appeals........................................................................140
Figure 15. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of fear appeals..............................................................................141

ix
Figure 16. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of positive sentiment per tweet. ...........................................................171
Figure 17. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of anticipation -associated rhetoric per tweet. ...................................175
Figure 18. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of joy-associated rhetoric per tweet. ....................................................176
Figure 19. Model estimates (w/ 95% CIs) of separate RGID effects of 113th to 116th MCs’
monthly average use of Trust Appeals in office account tweets. ....................................................177
Figure 20. Diagram of Plutchik’s (2001) wheel of emotions (as shown in Figure 4, p. 70) ..........186

x
Abstract
The rise of social media in the realm of political communication has profoundly
transformed both the media landscape and the strategies employed by elected officials to
cultivate their public image and connect with constituents. Platforms like Twitter have expanded
the reach of political messaging, providing opportunities for emerging voices to bring about
change. With the potential to go viral with a single tweet or post, Twitter incentivizes elected
officials to craft engaging rhetoric, often leaning towards negativity. This is particularly evident
during political campaigns and increasingly during periods of governance.
Despite our understanding of how politicians use digital platforms strategically, an
important question remains: who has the freedom to employ negative rhetoric, and who is
limited? While all candidates share the desire for reelection, current research tends to overlook
the diverse experiences shaped by candidates' social identities. Instead, explanations tend to
attribute rhetorical strategies to institutional dynamics, assuming a universal experience among
political figures.
Challenging this prevailing notion, I argue that candidates' strategies and rhetorical
decisions, both during campaigns and while in office, are shaped by a complex process that can
be best understood through an intersectional lens. Previous studies on gender or race in political
communication have yielded inconclusive results. To address this gap, I propose a theoretical
framework that incorporates insights from race, ethnic politics, and sociology. This framework
recognizes the unique experiences tied to a candidate's racial and gender identities and examines
how their interplay reflects their relative position within the U.S. hierarchy of privilege. Drawing
on social dominance theory, my model proposes a hierarchy of privilege that influences the

xi
rhetorical latitude of political actors based on audience expectations and biases rooted in their
race/gender intersectional identity (RGID). A lower position in this hierarchy corresponds to
heightened rhetorical constraints, limiting what a political figure can safely articulate without
facing backlash.
Using a comprehensive database of tweets from House legislators spanning the 113th to
116th Congresses, as well as tweets from 2020 election candidates, I conduct sentiment analysis
through multivariate regression models. This analysis aims to identify patterns of sentiment, both
negative and positive, in relation to the intersecting racial and gender identities of political actors.
This research addresses questions about the prevalence of negative rhetoric across demographics,
the nuances of emotional communication across different race and gender intersections, and the
role of party affiliation in shaping these dynamics.
From the fourth to the sixth chapters of this dissertation, I provide empirical evidence
supporting the significant impact of a political actor’s intersecting racial and gender identities
moderate emotive rhetoric in public discourse, effectively answering the question that yes,
political actors from minoritized RGIDs communicate in ways that are distinct from white male
counterparts who represent the dominant group in the United States. Importantly, these social
identity driven effects persisted when political party was held constant, across varying electoral
contexts, and amongst incumbent candidates and challenger candidates.
This dissertation uncovers the complex layers of rhetorical constraints tied to the
intersectional backgrounds of U.S. political figures, both within and outside of Congress. The
insights from this study emphasize the significant influence of intersectional identities on
congressional communication and the differential freedom afforded to political elites in
expressing their emotions. This research serves as a foundation for further exploration of how
intersectional identities shape the behavior of political elites in the U.S. political landscape.

1
Chapter 1
Introduction
In 2018, U.S. Congresswoman Lauren Underwood (D-IL 14th District) was making her
inaugural bid for Congress in one of the most competitive House races Democrats were hoping
to flip that year. On top of the known disadvantages associated with running against an
established four-term incumbent during a midterm year, Underwood and her campaign
described having to face specific obstacles throughout the electoral process due to her intersecting
identities as a Black woman1. From the lack of initial support to get on the ballot, to the lack of
trust in their electability once on it (Herndon 2018), Underwood’s campaign points to the varied
ways in which group-based social hierarchies, and the systems of privileges they inform, can
translate into markedly tangible consequences for members of nonprivileged groups. For women
candidates from racially minoritized2 backgrounds (i.e., Black, Indigenous, Persons of Color or
1 In recognition of the social and cultural dimensions of gender identity, I use the terms "women" and
"women candidates" in this project to describe the specific group of congressional elites who identify as
women.
2 Adapted from the American Psychological Association’s (2021) guidelines for inclusive language,
‘minoritized’ is used in this dissertation as an alternative to ‘minority’ to underscore the role of societal

2
BIPOC3), this means having to negotiate both gendered and racialized social processes across
campaign and governing contexts in how they approach their public-facing communications and
present themselves to voters.
As sociologist Erving Goffman (1959) suggests, all social interaction is performative where
we wear masks to put our audience at ease—presenting ourselves to fit the parts that we are
playing and meet our audience’s expectations. This feels particularly apt for public-facing
legislators who must curate their image to best signal their alignment with voters. The struggles
Underwood faced in her inaugural congressional bid speaks to the experiences of
underrepresented candidates who must navigate complex yet distinct, multifaceted social
hierarchies with their campaign's communication strategy. While political in nature, these
interactions and the amount of agency individual actors have in their self-presentation when
forming communicative relationships do not take place outside the social structures that maintain
imbalanced group-based perceptions and expectations. This necessitates further consideration of
how political actors, in spite of their shared electoral goals, should not be assumed to perceive
and assess the electoral environment in the same way; but rather, members of underrepresented
backgrounds experience varying constraints in what they can say and do.
structures in systemically subjugating specific groups of individuals. The term ‘minoritized’ helps clarify
and acknowledge those whose social identity and position have been constructed through discriminatory
social processes and power dynamics.
3 The term "BIPOC" refers to whether the political actor is from a nonwhite historically excluded and
marginalized racial group. While all nonwhite racial and ethnic identities are considered racially
minoritized given their shared positioning in U.S. systems of power relative to the privileged (White), it is
important to recognize the distinct histories and experiences rooted in specific institutionalized forms of
racialized subjugation. In this dissertation, I conceptualize "BIPOC" as a term of solidarity in
acknowledgement of the historical context and initial intent behind the term "women of color", which
originated at the 1977 National Women’s Conference to signify the shared commitment for collaboration
alliance amongst an alliance of minoritized groups of nonwhite women to collaboration (Grady 2020).

3
Facing voter expectations that often reflect prevailing identity-based stereotypes4,
underrepresented candidates and their campaigns must operate with further restrictions on the
strategies they adopt to present themselves in a way that simultaneously conforms to voters’
stereotypical beliefs (Kahn 1996) and meets established norms or expected standards held for
public officeholders (Dittmar 2015), while also being mindful to avoid appeals in their public
communications that can trigger voters’ negative identity-based perceptions (Cryer 2019).
In this dissertation, I develop a more nuanced understanding of congressional
communication through the conceptual lens of intersectionality theory (Crenshaw 1989) that
better addresses the complexities of an increasingly diverse media landscape and electorate.
Bringing together perspectives in legislative behavior, race and ethnic politics, and social
psychology, I explain how representatives’ race/gender intersectional identity (RGID)5 shape
what and how representatives say in their public rhetoric in today’s new legislative arena—social
media. I posit that what representatives say can be understood as a relational dynamic between
the electoral environment, and the mobility and access granted to a particular political actor in
the broader social context to levy influence on the scope of debate. With a focus on both structural
and individual level influences, I specify a dynamic theoretical framework that develops a
nuanced understanding of how racialized and gendered social processes importantly
contextualize and influence individuals’ threat perception in the electoral environment, creating
distinct incentive structures that inform their rhetorical approaches.
Drawing upon an extensive collection of congressional Twitter communications, this
study examines the nuanced ways in which race and gender shape what representatives choose
4 See: Sanbonmatsu (2002); Huddy and Terkildsen (1993); Dolan (2014); Hatemi et al. (2012); Bauer (2013);
McElroy and Marsh (2010). 5 I refer to ‘race/gender intersectional identity’ as ‘RGID’ interchangeably in this dissertation.

4
to publicly express—and what they choose to withhold—to claim a seat at the proverbial table
and secure it. I offer an alternative explanation of why representatives make the rhetorical choices
they do that centers race and gender as foundational forces in electoral politics and
communication. Given the relative absence of intersectionality in the theoretical development of
congressional and campaign communication research, this dissertation sheds new light on the
significance of distinct race/gender identities (RGIDs)6 in shaping legislators’ rhetorical
strategies, providing a foundational perspective that better reflects the complexities of
understanding legislative behavior given a steadily diversifying Congress.
Congressional Outreach and the Wild, Wild Web
In a representative democracy, politicians are expected to appeal to voter preferences
given the electoral incentives to do so. A wide range of studies point to the persisting influence
of political elites and their discourse in shaping public opinion (Zaller 1992; Santoro et al. 2021;
Druckman et al. 2013), candidate evaluations (Rogowski and Stone 2020; Woessner 2005), and
policy preferences (Martin et al. 2020; Flores 2018; Hopkins 2018; Broockman and Butler 2017;
Gabel and Scheve 2007). Elite cues that are attributable to a particular party are particularly
impactful on the formation of issue-related positions (Guisinger and Saunders 2017; Bullock
2011), with those who can control the argument are more likely to successfully translate belief
into policy (Waldman and Jamieson 2003). As ideologies are expressed and importantly,
“confirmed, changed and perpetuated through discourse” (Van Dijk 2006), this reinforces the
6 Throughout this dissertation I use the acronym ‘RGID’ interchangeably for ‘race and gender intersectional
identity’ with the aim clarifying writing and improving overall readability.

5
importance of continued research on the communication strategies of elected representatives
given the capacity of elite rhetoric to drive what people know and think about politics.
As electorally motivated political actors, Members of Congress (MCs) engage in the
'permanent campaign' (Blumenthal 1982) to continually solicit and win public support. While in
office, they develop a representational 'homestyle' both in their district activities and public
rhetoric (Fenno 1978), using their congressional outreach strategically to influence voters’
evaluations of them (Grimmer 2013). MCs increasingly rely on social media platforms to express
and develop their representational styles (e.g., Russell 2021), making strategic decisions about
their constituent engagement and self-presentation online in pursuit of maintaining the electoral
connection (Mayhew 1974).
New Media and the Rhetorical Race to Power
The race for the public’s attention is more competitive than ever with increasing diversity
of media avenues for sharing and receiving political information. Elected officials must compete
online to compete electorally— with greater exposure and news coverage, the public may
evaluate the party and its policy proposals more favorably (Iyengar 1990; Iyengar and Kinder
2010). The media-centric nature of modern political campaigns lends greater power to headlines
and the actors who can attract them. In the hybrid media system where social media and
traditional news media continuously interact and shift in power (Chadwick 2017), political elites
can shape the news coverage around them with a single tweet. As a particularly memorable
example of this, prior research examining news coverage of 2016 presidential candidate
communication on Twitter found that Donald Trump’s controversial tweets received
significantly more coverage from traditional news media than Hillary Clinton’s more

6
conventional tweets (Crigler et al. 2017), amplifying Trump’s candidacy and perhaps
unintentionally setting the news agenda with his sporadic tweets.
The dynamic, reactive environment specific to Twitter’s platform7 adds an important
dimension to modern campaign strategy that makes it distinct from how candidates and their
campaigns make use of other social media platforms that vary in perceived audiences,
functionality, and norms (Kreiss, Lawrence, and McGregor 2018). More competitive information
environments such as Twitter—where both the technical functionalities and substantive aims of
the platform are designed to facilitate user engagement and interactivity—produce a highly
reactive news environment that captures greater temporal variation than static platforms such as
candidate websites, which have significantly less message turnover. As the first form of social
media widely adopted by members of Congress, Twitter remains one of the primary platforms
along with YouTube and Facebook that are used for congressional communications (Strauss
2018).
As social media platforms play an increasingly central role in electoral politics, the noise
and interference of campaign "crosstalk" (Just et al. 1996) is now a structural element of the
electoral landscape that changes what it means to control the scope of debate. While it still holds
that power is held by those who are able to control the scope of debate (Schattschneider 1975),
the amount of opportunity and points of access for political actors to levy influence on how
political information is distributed and ultimately, understood by the public, has altered
dramatically. In today’s hybrid nature of political communication, how public-facing officials
7 In the context of this dissertation project, I examine Twitter prior to Elon Musk’s tumultuous acquisition
of the social media company that began April 14, 2022 and concluded on October 27, 2022. It is worth noting
that Musk has since made various policy changes to the platform, which likely has implications for how
the platform is used that research in political communication must grapple with moving forward.

7
connect with their constituencies has shifted the production of political information away from
traditional organizational settings that are more hierarchical in terms of distribution of power
and rather, towards a more cyclical and interactive process that lends greater emphasis on
individual actors (Chadwick 2017).
The rapid expansion of available communication tools, particularly those with an
emphasis on interactive functionality like most social media platforms, how legislators
communicate can often matter more than what they say. The use of rhetorical strategies in the
right information environment, for example, can help elected officials facilitate more favorable
constituent evaluations and increase their perceived credibility (Hwang 2013).
Conveniently accessible digital platforms like Twitter, for example, have provided more
opportunities for new faces and voices to arise (Bennett and Segerberg 2013), offering potentially
game-changing exposure and reach with a single tweet. There is, however, good reason to doubt
that all political actors can navigate these new terrains the same way if not on equal footing.
While the incentive of staying in office remains constant for all public-facing politicians,
the strategic considerations that shape the communication approach and rhetorical choices of
individual candidate-actors vary by their relational positioning in both the electoral environment
and the broader social context. In other words, how electorally motivated actors adapt their
strategic messaging and rhetorical appeals is best understood as a dynamic and relational process
that engages with the structural influences of the broader social context. The amount of agency a
candidate-actor has within a given electoral environment is structured by the extent to which they
are constrained by their relative social positioning, thereby shaping how expressive they are in
their strategic rhetoric while in office and on the campaign trail. Certain attention-grabbing tactics
such as going negative or engaging in negative campaigning, for example, come with inherent
risks that may not necessarily affect all candidates equally. By virtue of individual differences, a

8
political actor’s electoral risk environment is shaped by contextual threats to their political
survival, which in turn, informs a similarly distinct incentive structure.
Drawing from evolutionary psychology research on the role of individual threat
perception in strategic decision-making, Druckman et al. (2009) contend that heightened
perception of threat acts as a stimulus for political actors to adopt risky behaviors and actions
when they perceive threat to their political survival (McDermott, Fowler, and Smirnov 2008).
Some campaign strategies are more effective in achieving this goal, but perhaps the most studied
(and most controversial) is the use of negative campaigning or "going negative", which has long
been recognized for its value in attracting media coverage (Maier and Nai 2020) and prompting
voter attention in campaigns (Kern 1989).
Speaking and Leading with Negativity
Appeals to emotions are fundamental to modern congressional public discourse and have
been particularly salient with the ubiquity of appeals to negativity on Twitter. Scholars have
linked the use of this technique to those candidates in less advantaged positions, such as
challengers (Druckman, Kifer, and Parkin 2020), political outsiders or anti-establishment figures
(Hansen and Treul 2021), populists (Gerstlé and Nai 2019), and underdogs (Auter and Fine 2016),
who utilize such tactics to garner greater attention. Seeking to engage voters and develop the
electoral connection (Mayhew 1974), elected officials may weave varying emotional appeals into
their communications for different objectives. For instance, a candidate seeking to galvanize their
constituents for the purpose of securing votes may deploy appeals to anger.
While negative rhetoric on Twitter has become a lasting hallmark of congressional
communication in both governing and campaign contexts (Russell 2018; Gervais, Evans, and

9
Russell 2020), it remains unclear whether we can assume that all candidates can uniformly benefit
from employing rhetoric that is generally deplored. This raises questions about how negative
discrete emotions are used by underrepresented candidates across diverse contexts, and the ways
in which it may differ from the conventional white male candidate.
Considering that audiences often form perceptions of messages based not only on the
content but also the sender, one might infer that messages delivered by individuals from
historically underrepresented racial or gender groups could be received differently from those
conveyed by their white male counterparts. By attending to the potential variance brought on by
the interplay of race and gender, this research sheds light on how disparate identities influence
the emotional appeals adopted in representatives’ discursive strategy.
In this dissertation, I examine the relationship between social identity and congressional
communication, looking at how RGID shape distinct patterns of rhetorical appeals in line with
the constraints corresponding with their relative social positioning. In what ways does race and
gender shape the strategic appeals representatives make in their congressional outreach? And to
what extent is congressional public outreach on social media influenced—or rather,
constrained—by the intersection of their race and gender identities?
This suggests congressional strategic communication reflects a dynamic, relational model
in terms of interactions between individuals and groups within the broader social context.
Indeed, prior research on campaigns and messaging strategy has shown great variation in
effectiveness of strategies and tactics across different electoral contexts and time periods,
providing an apt reflection of “the complex web of electoral dynamics” (Dittmar 2020, 312). The
effectiveness of strategic rhetoric relies on the intersubjective acknowledgement of the
candidate’s presented identities, highlighting the core tenet of strategic communication being an
inherently relational process and should be modeled as such in theories of elite rhetoric.

10
Understanding strategic messaging requires further consideration of the relational
dynamics between social groups that allows for prioritization of individual social identities
where relevant, but also accounts for a more inclusive intersection of social identities, ideologies,
and partisanship. In this context, congressional communication resembles a multifaceted
relationship that establishes incentive structures for elites to follow when communicating with
their voter base. To mobilize votes, political actors try to align themselves with groups they hope
to gain support from by forging a shared sense of identity (Edelman 1964; Kreiss, Lawrence, and
McGregor 2020; Leach 2011). As with any form of communication, the sender must establish the
channel with the receiver. In political communication, this critical step is done through partisan
signals and speaking to partisan ideology. Signaling partisan identity is the most direct way to
calibrating the audience of potential voters to the speaker. From here, an elite can foster stronger
connections with constituents by communicating more relatable qualities through the conduit of
aligned partisan ideologies.
As Jost (2021) posits, ideological frameworks are socially constructed and subjectively
validated, asserting that "interpersonal and intergroup factors determine whether or not the
discursive superstructure developed by political elites becomes a shared social representation
that successfully penetrates public consciousness" (77). This emphasizes that effective political
communication hinges on the elite’s ability in signaling their shared group membership with
their target audience or voter base. Thus, to form a comprehensive understanding of the strategic
appeals used in congressional outreach, consideration of how multiple identities inclusive of
associations with both political and social groups is increasingly relevant. The ability of political
elites to effectively signal their membership within specific social groups is instrumental. Prior
research has proposed ways partisan can be signaled such as through credit-claiming messages
for popular programs (Grimmer, Westwood, and Messing 2014). However, in the contemporary

11
polarized landscape, political identity has increasingly converged with social identity (Greene
2004; Mason 2015).
This evolving landscape necessitates a recalibration of elites’ outreach strategies,
requiring they tailor their communication to resonate with specific audiences to then leverage
these communications to signal their shared group membership. By framing their policy choices
in carefully curated terms, they can reaffirm their alignment with constituent preferences and
bridge gaps where incongruencies exist (Grose, Malhotra, and Parks Van Houweling 2015). Such
strategic messaging is crucial in fostering connections and sustaining political credibility when
political and social identities are so intertwined.
Built upon the basis of social identity theory, group norms play a vital role in capturing
the distinctive properties of groups (Hogg and Reid 2006; Tajfel 1986). These norms are contextdependent prototypes that shape the attitudes and behaviors of group members, with certain
influential individuals, especially leaders, exerting greater normative influence (Hogg and Reid
2006). Recent research has highlighted the impact of environmental factors on group behaviors,
emphasizing the need for a more comprehensive understanding of how group norms operate
within varying contexts to better understand group behaviors (Brown 2020; Hogg and Reid 2006;
Smith 2019).
Same Emotions, Differing Risks: Emotional Expression as a Relational
Social Process
Emotions, frequently conceptualized as individual experiences, can also function as a
collective social phenomenon, reflecting shared sentiments that are sensitive to the broader socio-

12
political landscape (Phoenix 2020). This recognition underscores the intersubjective foundations
that shape the conception and experience of emotions as dual processes—both internal and
external. In treating emotions as inherently relational and socially constructed, dynamic models
offer a schema for comprehending how individual emotional experiences are generated and
collective emotional expressions are ascribed significance.
In navigating the polarized climate of contemporary political discourse and positioning
themselves within today's hybrid media system, underrepresented candidates from marginalized
communities confront unique racialized and gendered prejudices, shaped by their relative
positioning within these social dynamics play a pivotal role in structuring their psychological
perceptions (Zou and Cheryan 2017). Considering that voters do not perceive nor evaluate all
candidates with the same set of expectations, it is reasonable to suspect that candidates differ in
both how they perceive their risk environments as well as how they form their rhetorical
strategies in response.
These prejudices shape particular perceptions of electoral threats and, consequently,
dictate the strategic behaviors and rhetorical choices these candidates employ in response. Such
candidates may incur higher electoral costs for employing conventional tactics like "going
negative," which may backfire. These racialized and gendered social processes thus constrain the
messaging and rhetorical branding available to underrepresented candidates.
In this dissertation, I explore the relationship between social identity and congressional
communication, addressing the extent to which representatives’ RGID inform distinct
approaches to their use of strategic appeals. How does race and gender shape the rhetorical
strategies representatives make in their congressional outreach? And to what extent do these
rhetorical strategies reflect the privileges and constraints associated with representatives’
strategies and perspectives? Understanding the ways in which representatives' RGID shape their

13
rhetorical strategies in making the electoral connection is an important and increasingly relevant
area of inquiry for scholars interested in political communication.
I apply an intersectional approach to best capture the effects of disparate and intersecting
social identities and anchor them to a single framework. Introduced by Kimberlé Crenshaw
(1989), the concept of intersectionality provides a fundamental understanding that a person's
identities are not merely layered but intertwined, shaping unique experiences and perceptions.
When applied to research in political communication, it becomes apparent that the experience
and strategic behavior of a congressional candidate cannot be fully understood without
considering how one’s distinct social identities from birth bestows an implicit value in systems
of privilege.
I propose a dynamic model that explains how and why the intersectionality of race and
gender influence candidate appraisals of electoral threat as a function of the advantages or
constraints imposed by their relative social positioning in group-based hierarchies. This in turn
prompts candidates to assess their available resources (i.e., the positive or negative social value
attributed to groups with either high perceived status or inferior status, respectively) and
determine their ‘coping’ potential, or capacity to take on risks in forming their strategy
responding to said threat. Integrating theoretical perspectives from race and ethnic politics and
social psychology, this framework explicates their interconnections and implications for political
discourse in electoral environments.
An overview of the theoretical framework is presented in Error! Reference source not
found., which details the underlying concepts of the process in which political actors perceive
electoral threats through the lens of their social positioning informed by their RGID. Candidateactors then individually appraise the risk in terms of the privileges, or lack thereof, their social
positioning may afford in order to assess if they have sufficient social resources or capital that

14
allows them to accommodate the potential ramifications of employing a more expressive
rhetorical approach that draws more attention to themselves. Taken together, the model
conceptualizes how political elites appreciate electoral threats and respond accordingly, while
involving the domains of societal structure, internal processing, and externalized response.

15
Figure 1. Simplified diagram of main processes & concepts in theoretical framework. ELECTORAL THREAT ENVIRONMENT | STRESSOR AVOID RISK EXPRESSIVE TAKE RISK CONSTRAINED RESOURCES AVAILABLE INSUFFICIENT RESOURCES RESOURCES SUFFICIENT APPROACH STRATEGIC + SOCIAL CONSTRAINTS IDENTITY GROUP INTERSECTIONAL GENDER RACE & ETHNICITY POSITIONING PRIVILEGED NONPRIVILEGED POSITIONING SOCIAL GROUP POSITIONING

16
Starting from the left of Error! Reference source not found., social dominance theory
posits that a socially constructed group hierarchy shapes a structure of privilege. In this context,
individuals are categorized based on the visible, straight forward attributes "assigned at birth,"
namely, their skin color and biological sex. This assignment places them either in the privileged
majority group or the subordinate minority group, bestowing advantages, or imposing
constraints accordingly (Gordon 2004; Scheurich 1993; Nieto et al. 2010). Distinct party norms and
ideologies further form unique group expectations, which in turn direct the constraints applied
to minority groups.
Understanding their social positioning within systems of structured privilege lays the
foundation for candidates to assess their risk environment and gauge threats to their political
survival through each of their individual appraisal processes. Finally, as rational actors,
candidates balance potential benefits, risks, and constraints to make the strategic move to best
serve their self-interests and political survival. Figure 1 visualizes the processes of the proposed
framework.
Underrepresented candidates estimate the risk in their environment in relation to their
offsetting resources—high social value accessible only to the privileged group. Being denied these
advantages results in a harsher risk environment for political actors at subordinate social ranks.
The privileged group in the social hierarchy, in this case white men, benefit from their established
“high social value” and benefit from a baseline of positive stereotyping from outgroups. These
privileges situate them in a distinctly advantaged position in their ability to effectively convey
their messages with most audiences with comparatively lower risk associated with negativity.
Congressional elites motivated by their desire to secure reelection, focus their campaign
and in-office rhetorical strategies to develop a presentational style that primes certain image

17
perceptions to best appeal to their perceived audience/constituency. I argue that the social
positioning of representatives at the intersection of race and gender identities shapes distinct
perceptions of their risk environment. The racialized and gendered social processes underlying
the electoral and political landscape in which the communicative relationship is developed must
be considered to develop a more nuanced and accurate picture of how communication strategies
are formed between the sender and intended audience/target group. All communication,
particularly in the form of political persuasion, is fundamentally relational, developed with
consideration of how the sender of the message can best relate and connect with the target
audience. Recognizing that baseline differences exist in terms of how an individual sender’s social
status and value is perceived by others within the broader context is essential to understanding
the incentive structure candidates are operating from—i.e., the constraints and amount of agency
afforded by their social positioning—when forming their rhetorical strategies.
Research Motivation
With the prospect of going viral with a single game-changing tweet, Twitter’s platform
offers explicit incentives for elected representatives to make rhetorical choices that best enable
them to attract higher engagement—implicitly structuring the content and even the timing of the
messages they choose to share. This relates to growing normative concerns regarding the extent
to which Twitter’s unique communication platform not only exacerbates negativity in politics but
may also reinforce the barriers that continue to inhibit underrepresented candidates from
achieving equitable electoral consequences.
Moreover, I argue that an enhanced focus on individual variation within theoretical
frameworks is imperative for a nuanced understanding of congressional outreach. While the

18
context in which messages are disseminated shapes their reception and impact, the identity of the
source or sender is equally crucial and deserves theoretical accommodation. By incorporating this
individual-level variable into theoretical considerations, one can more effectively elucidate the
conditions that contribute to the unique incentive structures under which individual
representatives and candidates operate when making strategic assessments and decisions.
A primary motivation of this research is to address the relative lack of theoretical
development in studies of Congress and political communication more broadly that situates race
as foundational rather than auxiliary to frameworks of legislative behavior and strategic
messaging. While work in race and ethnic politics has historically been isolated from other fields
of research, it has notably remained so in studies of political communication despite identity
politics, particularly with regards to gender, being well attended to. Relatively less research, to
my knowledge, has systematically analyzed how legislators and candidates from historically
minoritized backgrounds make use of new media technologies across varying electoral contexts
with a similar level of attention afforded to addressing the foundational role of race and gender
in theories of political communication.
I highlight an oftentimes implicit assumption in prior research that takes homogeneity of
political experience across political actors as a given and argue that doing so provides an
incomplete understanding of congressional communication. This assumption of homogeneity
perpetuates an exclusionary approach that systematically overlooks variance in how
underrepresented political actors perceive their environments and therefore, how they respond
to them as well.
More specifically, I theorize that actors’ perception and strategic response to electoral
threat is uniquely shaped by the privileges and constraints they experience—and expect to
experience—relative to those in power not only in the political and institutional sense, but within

19
the social hierarchical structure as well. I develop a theoretical framework that addresses the
systematic influence of race and gender in shaping the incentive structure or conditions under
which candidates as individual actors strategically assess their environment offline or online,
informing what they say and do in that space accordingly. Rather than assuming that candidates
share comparable political experiences while on the campaign trail or in office, I argue that what
representatives say is best understood as a dynamic process in which individual perceptions of
the electoral environment are shaped by the level of access, mobility, and agency a particular
political actor is granted from broader social structures to levy influence on the scope of debate.
Purpose of Study
This dissertation takes an intersectional approach to develop a more nuanced and
comprehensive understanding of congressional strategic behavior and communication that
acknowledges the structural influence of group-based hierarchies in the broader social context.
Specifically, I theorize that the relative social positioning formed at the intersection of race and
gender distinctly shape the constraints that political actors employ in their strategic appeals when
seeking to connect and develop constituent relationships. Essentially, this RGID influences the
strategic behavior and rhetoric legislators use in varying electoral contexts. Through the
conceptual lens of intersectionality theory, an individual’s set of identities are assumed to not be
merely layered but intertwined (Crenshaw 1989), and together are instrumental in shaping
unique experiences and perceptions of the political world.
When applied to research in political communication, it is clear that the experience and
strategic behavior of a congressional candidate cannot be fully understood without considering
how one’s distinct social identities from birth bestows an implicit value in systems of privilege.

20
It acknowledges that these intertwined identities are pivotal determinants in molding
individuals' experiences and actions, most certainly in political contexts. This multifaceted
intersectionality shapes the unique experiences of marginalized racial and gender groups,
thereby influencing the rhetorical appeals in their public communications. Ultimately, the aim is
to examine how race, gender, and their intersectional overlap shape the political narrative
expressed by these congressional elites in their public outreach.
This dissertation helps illuminate a theoretical basis for unifying studies of congressional
communication that typically consider candidate campaign messaging and legislative outreach
in separate contexts. Indeed, work in race and ethnic politics is notably isolated from other fields
of research but is particularly so within political communication scholarship. To my knowledge,
there has been relatively limited systematic research on how historically marginalized
representatives and candidates make use of new media technologies in different electoral and
media contexts. Similarly, insufficient attention has been afforded to addressing the foundational
role of race and gender in theories of political communication.
While adopting an intersectional analytic approach inevitably introduces additional
complexities and ‘messiness’, by doing so scholars are better able to capture the political realities
that marginalized groups experience (Smooth 2011). Intersectionality offers a more holistic
perspective on communicative relationships that may be better suited to studying the nuanced
and complex dynamics that arise in legislative outreach from increased diversity among
representatives and the electorate at large. This dissertation focuses on the intersecting influence
of race and gender in shaping the strategic appeals of congressional rhetoric. In recognizing the
multidimensionality of social identities and the compounding effects of these communication
strategies and political processes, this dissertation contributes a more comprehensive

21
understanding of congressional communication that better reflects today’s sociopolitical
landscape.
Design of Study
To investigate the impact of race and gender on the rhetorical strategies of elected officials
in the age of social media, this study builds upon existing research that explores how lawmakers
utilize Twitter beyond election campaigns, specifically in the context of governance and
representation (Russell 2021; Golbeck et al. 2010; Herring 2009; Shapiro and Hemphill 2015).
While the proposed framework in this dissertation is applicable to all public-facing officials, the
focus of the study is on the U.S. House of Representatives. This choice is motivated by the
frequency of reelections and the potential turnover that members must regularly face, both in
terms of their individual races and majority party control.
From two robust and comprehensive datasets of public tweets from political actors, the
present study aims to elucidate the role of the disparate RGIDs of elites on the public discursive
strategies political actors subsequently adopt. The first cohort consists of incumbent
representatives in the 113th to 116th Congresses on their official House Twitter accounts, this
study analyzes the congressional outreach messages shared online from 2013 to 2020 and will
encompass alternating periods of governance and election in addition to a time-period with
presidential administrations equally split between Democrat and Republican.
The second dataset of analysis is a robust collection of public tweets made during the 2020
campaign year. This dataset carries over the 2020 incumbent data from the first cohort and
combines it with the 2020 challenger candidate tweets. By utilizing social media as the primary
source of data, this study provides a detailed view of the dynamic nature of congressional

22
communication. Twitter, as the chosen platform, not only reflects the wide range of strategies and
messaging employed by individual candidates and representatives, but also facilitates regular
and frequent interactions, resulting in a rich and extensive dataset.
In addition to its representational relevance, the accessibility of data collection through
Twitter API further supports its selection as the main data source for this study. This choice
enables the creation of a comprehensive dataset that spans multiple congressional sessions and
presidencies. Through this expansive dataset, this study offers valuable insights into the diverse
strategies and appeals that shape legislators' rhetorical choices, issue emphasis, framing
techniques, and other fundamental elements of political communication.
By integrating intersectionality as a conceptual lens, I extend prior analytic approaches
focused on the extent to which candidates decide to take on more or less risk in their strategic
messaging, advancing scholarship in congressional communication and campaigns. The
framework developed in this study provides a more comprehensive understanding of how social
structures and processes importantly contextualize the intergroup dynamics that underlie the
way relational communications are generally perceived and individually formed. This research
contributes to the growing body of work examining the influence of identity-based group
dynamics in the electoral process that typically focus on the ‘demand’ side of communication,
such as candidate evaluations or other constituent outcomes. By engaging with emerging
theoretical dialogue and inquiry on, expands our knowledge of the complexities of political
communication.
My study navigates the confluence of racial identity, gender, and political
communication, with specific focus on congressional Twitter campaign strategy. The significance
of this subject matter is accentuated in the light of its resonance with the ongoing socio-political
dynamism. This dissertation is a contribution to the incipient application of intersectionality to

23
political communication theory. By incorporating multiple axes of identity into campaign
research, including the concept of "intersectional stereotyping" proposed by Doan and HaiderMarkel (2010), my dissertation addresses this research gap and enriches our understanding of
how intersectionality shapes the political communications landscape.
Furthermore, underpinning this investigation is a monumental assembly of a
congressional Twitter database. This repository stands unique in its completeness of
demographic coding and the timeframe of congresses it envelops. Not only does this database
serve as an enriched resource for the present study, but it also introduces a platform for
prospective explorations in political communication and intersectionality.
Outline of Contents
The first chapter of this dissertation introduces the intersection of racial identity, gender,
and political communication, with a specific focus on the digital communication strategies of
congressional candidates on Twitter. The importance of the topic is highlighted in terms of its
relevance to current political and social dynamics.
The second chapter presents a comprehensive literature review, to support and develop
the basis of the theoretical framework of the dissertation that is grounded in social identity theory
and outgroup framing. This entails more in-depth discussions of emotive rhetoric and how
certain patterns of sentiment appeals may correlate to the electoral threat political elites from
different racial and gender backgrounds may experience more or less of. Furthermore, by
examining the existing scholarship on congressional communication, race and ethnic politics, and
political psychology, I will highlight key gaps in the literature (particularly the paucity in

24
intersectional approaches in studying political communication strategy). From there I will present
my relevant hypotheses that I will test in the chapters that follow.
Chapter three discusses the methodological approach of the study, elaborating on the
rationale for the chosen methods and data selection. It also provides a general overview of the
data collection process in the study prior to discussing the methods used to conduct the text
analysis and empirically test the expectations specified by my theoretical framework. Supporting
materials are included in the Appendix.
In chapter four, I examine the negative emotive rhetoric used in representatives’ public
outreach messages shared on their verified official House Twitter accounts and present the
findings derived from my analysis. Leveraging text data from over 2 million tweets shared by
members of the 113th to 116th Congresses from January 3rd, 2013, to November 3rd, 2020, I
demonstrate how electorally-motivated representatives adopt distinct strategies in their in-office
public outreach that reflect the social constraints—or lack thereof—associated with their
intersectional positioning relative to the group-based hierarchies embedded in the broader social
context.
Chapter five is a continued exploration of the use of negative emotive rhetoric by
candidates during the 2020 election. Of particular interest is how the intensified scrutiny and
electoral threat may further influence how various candidates perceive their own risk
environments and how that may impact how they may strategically communicate online.
Chapter six shifts focus from negative to positive emotive rhetoric that MCs in the 113th
to 116thg congresses may adopt as well as the candidates of the 2020 election. Given the limited
literature published to explore the salience of positive emotions in political communication, this
will help provide insights into how positive sentiment and appeals to discrete positive emotions
may vary based on disparate social identities and electoral environments.

25
Finally, the seventh chapter concludes the dissertation by summarizing the study, its key
findings, and contributions. It also discusses the limitations of the study, including suggestions
for possible new directions that future research can take to further explore the implications of
intersectionality in political communication and campaigning. The results of this study reveal the
underlying complexities of the social dimensions implicit to the strategic considerations political
actors must navigate in their communication strategies.
This dissertation aims to contribute to the understanding of how racial identity, gender,
and politics intersect and impact political communication strategies. It bridges a paucity in the
literature by employing an intersectional approach to analyze the Twitter strategies of
congressional candidates, providing new insights into political communication in a digital age.

26
Chapter 2
Costly Signals: Higher Risks, Uncertain Rewards
In this chapter, I review existing literature on congressional strategic behavior and
outreach given the proliferation of new media technologies and an increasingly overcrowded
information environment (Bawden and Robinson 2020), forever altering the strategies and
rhetorical choices legislators use in their public outreach to maintain the electoral connection. The
now necessary adoption of social media for political communication grants legislators more
opportunity and discretion in signaling the identities they want to present, and priming specific
image perceptions that can best combat negative ones (Druckman et a. 2004). Despite core social
identities like race and gender being widely shown to matter when it comes to shaping how
candidates are perceived as well as how they perceive themselves, theoretical development in
existing research surprisingly remains inattentive to investigating the ways in which social
identities may intersect and motivate distinct strategies for candidates operating from varying
social positions within group-based hierarchies.
As public-facing elected officials, representatives are likely to experience greater external
influences in both magnitude and variety of source (Russell 2021); but how these influences are
individually experienced and responded to across varying intersecting identities requires further

27
theoretical and empirical investigation. Racial identities and the corresponding racialized
attitudes and prejudices, for example, influence the way people encounter, process, and engage
with information (Cryer 2020), driving key differences in how candidates perceive and evaluate
the electoral environment. This in turn shapes different incentive structures for candidates’
strategic behavior and rhetorical choices in their efforts to navigate and define themselves amidst
the shifting interactions between traditional and new media technologies that take place within
the hybrid media system (Chadwick 2014). As ideologies are expressed and more importantly,
are “confirmed, changed and perpetuated through discourse” (Van Dijk 2006), this highlights the
influential role elite rhetoric plays in shaping the public’s understanding of political issues.
The potential electoral advantages borne out of attracting larger audiences and greater
virality incentivize politicians to strategically frame their public outreach in ways that are most
appealing and persuasive. This raises the question: what kind of strategic language is used by
congressional members, and to what extent are these rhetorical strategies affected by their
intersecting identities? While not a perfect process, research shows that politicians do engage
with citizens on social media by raising and discussing topics that citizens bring to their attention
(Barbera et al. 2014), making social media communication by political elites more and more
consequential for the study of American politics.
Identities Presented, Identities Perceived
While the incentive of staying in office remains constant for all public-facing politicians,
the strategic considerations that shape the communication approach and rhetorical choices of
individual candidate-actors vary by their relational positioning in both the electoral environment
and the broader social context. In other words, how electorally motivated actors adapt their

28
strategic messaging and rhetorical appeals is best understood as a dynamic and relational process
that engages with the structural influences of the broader social context. The amount of agency a
candidate-actor has within a given electoral environment correlates with their perceived ability
to take risks in the strategies and rhetorical choices they make in office and on the campaign trail
in pursuit of the ever-present goal to sustain relevancy and control the scope of debate.
The concept of social identity, which refers to an individual's sense of belonging and
identification with a particular group or community (i.e., race, gender, political ideology), can
serve as a critical determinant in how individuals perceive and assess risks (Tajfel and Turner
1986). Understanding how the interplay between social group identities and their dynamics
influence individual perception and assessment of risk enriches our understanding of the
strategic decisions and rhetoric representatives adopt. This is particularly relevant in today’s
political landscape given the unprecedented increase in candidates and elected officials from
diverse backgrounds8, which calls for theories that consider the advantages and constraints
unique to varying identities. The behavioral and communicative strategies employed by political
actors today are therefore likely to diverge from the assumptions previously posited in academic
literature.
This discernable shift perhaps warrants revisiting prevailing explanations of
congressional communication that often prioritize party politics and partisan group identity as
the focal point shaping candidate-actors’ strategic considerations and behavior. Beyond the
monopoly of scholarship centered around party-driven explanations, studies that have
investigated the role of social identity are primarily focused on a single identity such as race or
8 There has been a continued increase in both racial and gender diversity in each Congress since the 112th
Congress in 2011, with the current Congress having the most diverse congressional body to date (Pew
Research Center 2023).

29
gender identity separately. Situated within this changing environment, the foundational concepts
of social identity and intergroup dynamics take on heightened relevance. Linear models that
attempt to outline the effects of social identity variations are no longer sufficient, given the
intricate web of individual and collective emotions and attitudes coexisting in real-time.
Selective Signaling: Identities Picked and Primed for the Norm
The Strategic Use of Emotional Appeals
In an era of affective polarization where partisans are biased to view co-partisans
positively and opposing partisans negatively (Iyengar and Westwood 2015; Iyengar et al. 2019;
Webster and Abramowitz 2021), political actors are often criticized for wielding language as a
weapon to attack their opponents both offline and in online contexts (Yarchi et al. 2021). As
congressional races have become less competitive with just 72 districts classified as swing seats
after the 2016 election by Cook PVI (compared to 164 districts in 1997), elected officials have little
incentive to compromise in safer and more partisan concentrated districts (Huder 2013). Citizens
and political elites alike regularly take to social media platforms to voice their thoughts, with elite
communication playing an influential role in reinforcing partisan affinities and group identities
(Bäck et al. 2023). While contentiousness and partisan driven conflict may be the bedrock of
democracy (Schattschneider 1975), contemporary party politics is perceived by some scholars to
resemble a bitter and unrelenting sports rivalry, marked by an unprecedented alignment of
ideological and social divides (Abramowitz 2018; Finkel et al. 2020; Iyengar, Sood, and Lelkes
2012; Mason 2015; Mason and Wronski 2018).

30
Within the discursive arena of politics, the strategic usage of emotions has come to occupy
a central role, shaping the tonality of political communication, and determining voter behavior
(Lodge and Taber 2005; Brader 2005). Their prevalence within electoral systems and their
potential sway on public consensus underscore the need to examine the varying effects of discrete
emotions more closely (e.g., Marcus and MacKuen 1993; Huddy et al. 2005). A wide breadth of
extant scholarship focuses on the distinct role of specific emotional appeals used in political
discourse, going beyond the valence model that bifurcates emotions as either positive or negative
sentiment. This broader scope of analysis underscores the capacity of individual emotions to elicit
unique behavioral responses.
In the realm of politics, informed elites understand the impact of emotions and tactically
curate their messaging to strike chords with voters on an emotional level. In this schema, the
usage of emotional appeals in the rhetoric of congressional elites provides critical insights into
the risk-bearing strategies employed by candidates. My conceptual framework suggests that
these appeals serve to establish hypotheses ripe for empirical exploration.
The political implications of emotion are pervasive, having been empirically linked to
perceptions of candidates and vote choice (Lodge and Taber 2005; Brader 2005), media
consumption and learning (Marcus and MacKuen 1993; Huddy et al. 2005), policy attitudes
(Huddy, Feldman, and Cassese 2007), and political participation (Valentino et al. 2011; Rudolph,
Gangl, and Stevens 2000). Furthermore, both negative and positive emotions can be regarded as
"bold or controversial" based on their application and the partisan context in which they are
deployed.
Despite the risks, the strategic use of emotions enables political elites to connect with
voters, shaping the norms of political communication and helping to maintain their support base
(Gadarian and van der Vort 2017). By appealing to voters emotionally, candidates can shield

31
themselves from the advances of rival parties and influence public opinion, participation, and
voter attitudes (Gross 2008; Brader 2006; Lodge and Taber 2005; Marcus 2003; Valentino et al.
2011; Weber 2013). The strategic modulation of these emotional appeals often occurs in response
to electoral vulnerability or proximity (Brader 2006; Gross and Johnson 2016).
Negative Discrete Emotions
Negativity, and the discrete negative emotions of anger, disgust, and fear, has become a
cornerstone of political communication that is rooted in its attention grabbing nature (Soroka et
al. 2019), with implications for generating online constituent engagement (Diez, Gulías, and
Quesada 2021), challenging voter assessments of candidate positions (Geer and Vavreck 2014),
incivility in political dialogue (Mutz 2007), the tactics of 'going negative' or disseminating attack
messages (Lau et al. 2007), and the deployment of negative partisan rhetoric (Iyengar et al. 2019).
Over the past decade, a surge of research has probed the role of specific negative emotions
in politics, with a significant emphasis on the emotion of anger (Valentino et al. 2011; Weber 2013).
These studies highlight the utility of anger in affecting a range of political outcomes, such as
increasing voter mobilization (Mason 2018; DeSteno et al. 2004; Valentino et al. 2011; Valentino et
al. 2008), strengthening partisan loyalty (Webster 2020), and activating group attitudes thereby
prompting less systematic processing of (and greater receptivity to) persuasive messages (Rydell
et al. 2008). Appeals to negative emotions have consistently demonstrated their political utility to
trigger an array of behaviors that collectively benefit a campaign (Brader 2005). Despite these
political advantages, anger may be considered unprofessional or inappropriate, particularly for
politicians who are expected to uphold a certain level of decorum (Averill 1982; Marcus et al.
2000).

32
Fear appeals, often employed to galvanize a base or cast opponents negatively, can create
anxiety or panic and may even fuel false beliefs or conspiracy theories (Marcus, Neuman, and
MacKuen 2000; Merolla and Zechmeister 2009). As an inhibiting emotion, it is situationally
associated with greater uncertainty and less control, which can diminish an individual’s reliance
on political party ideology and be open to new government measures to maximize their
protection (Marcus and MacKuen 1993; Brader 2005; 2006; Valentino et al. 2008; Wagner and
Morisi 2019; Lupia and Menning 2009; Clifford and Jerit 2018; Nadeau et al. 1995).
The use of disgust as an emotional appeal has received relatively less attention than the
use of anger or fear, with much of prior research focused on the rhetorical influence of disgust in
issue-policy domains linked to moral values and judgments such as LGBTQ+ rights (e.g., Smith
et al. 2011; Cunningham, Forestell, and Dickter 2013; Inbar, Pizaaro, and Bloom 2012) and
abortion (Kam and Estes 2016). Disgust is the emotional root of moral judgment (Hatemi and
McDermott 2012) and can facilitate harsher moral judgment when primed as a moral emotion
(Ben-Nun Bloom 2014), appeals to disgust can substantially extend the reach of messaging
(Vosoughi et al. 2018; Brady et al. 2017; Clifford 2019; Clifford and Wendell 2016; Horberg et al.
2009). While not necessarily a potent factor for generating votes, disgust has been utilized as a
tool to characterize outgroups (Gadarian and van der Vort 2018; Jost 2021). Disgust sensitivity, or
the susceptibility of the audience to disgust appeals, is a predictor for negative attitudes towards
“lower groups” (Hodson and Costello 2007) and has been associated with those ascribing to
conservative ideology (Inbar, Pizarro, and Bloom 2009). Disgust appeals made by minoritized
elites, like other negative emotions, have been correlated with disproportionate electoral risk,
especially amongst women (Hua, Brisbane, Crigler, and Just 2019).
While negative emotional appeals have become an integral part of the political landscape,
their usage varies significantly depending on the candidates' social and electoral positioning.

33
Challengers, outsiders, and underdog candidates often resort to negativity as a 'disruptive
strategy,' seeking to gain traction in an otherwise asymmetrical playing field (Druckman, Kifer,
and Parkin 2010). These candidates face a high-risk, high-reward scenario where negativity could
either catapult them into the spotlight or trigger a backlash that further marginalizes them.
In a similar sense to politically disadvantaged candidates, unconventional candidates
from minoritized race and/or gender identities also bear disproportionate risks when appealing
to negative emotions relative to their white male counterparts (Averill 1982; Marcus et al. 2000;
Phoenix 2020; Bauer 2017; Krupnikov and Bauer 2014). In that regard, it is crucial to note that
congressional communication does not occur in a vacuum but is deeply embedded in a relational
social dynamic. A candidate's identity—be it racial, gender, or socioeconomic—interacts with
prevailing cultural narratives, shaping how their emotional appeals are received. For instance, a
woman expressing anger may be perceived differently from a man displaying the same emotion.
Likewise, a Black candidate expressing fear and appealing to uncertainty could be subject to
discriminatory biases that further disadvantage them (Piston et al 2018), thus altering the riskreward equation for them vis-à-vis other candidates.
In a diverse socio-political landscape, candidates are not equally at liberty to take the same
risks. The intersectionality of race, gender, and other social identities imposes a 'risk ceiling' on
underrepresented candidates, making certain rhetorical strategies riskier for them than for their
majority counterparts. Therefore, although negativity might offer an expedient path to visibility,
the potential for appeals made by minoritized actors to backfire, whether by aggravating negative
group-based prejudices and stereotypes, imposes social constraints on their strategic behavior
that members of the dominant group in privileged positioning do not encounter. This leads to the
expectation that minority RGID actors will be more constrained in their strategic approach
relative to their white male counterparts.

34
Positive Emotions
Positive emotions play a key role in shaping behavioral tendencies and is differentiated
from positive mood based on their association with initiating approach behavior (i.e., activity
engagement) (Cacioppo et al. 1993; Davidson 1993; Frijda 1994) or continued action (Carver and
Scheier 1990; Clore 1994). Through this lens, the emergence of positive valenced emotions such
as hope, pride, joy, and compassion tend to fortify daily habits and customary attitudes
(Cacioppo, Gardner, and Berntson 1999; Davidson 1993; Watson, Wiese, Vaidya, and Teilegen
1999; Carver and Scheier 1990; Clore 1994; Marcus and MacKuen 1993). A positive disposition
fosters heuristic processing, predominantly stemming from a sense of security. Positive emotions
are thought to have evolutionary roots, primarily functioning as a countermeasure to alleviate
the negative sensations that ensue from distressing events (Fredrickson and Levenson, 1998;
Fredrickson et al. 2000). Prior research on positivity has largely focused on enthusiasm as an
isolated emotion (Brader 2005), examining its relationship with voting patterns due to its inherent
activating characteristics and its association with generating “interest”.
Despite its ubiquitous and salient effects, positive emotions have not had the same
academic attention that negative sentiment has due to a variety of reasons that has stirred
confusion amongst scholars. Barbara Fredrickson (2001) argued that while negative emotions
arise from direct and sometimes obvious cause-and-effect relationships often triggered by stress,
it would be erroneous to apply the model of negative emotional dynamics onto positivity (Frijda
1986; Frijda et al. 1989; Lazarus 1991; Levenson 1999; Oatley and Jenkins 1996). Positive emotions,
rather, may activate through a “Broaden and Build” model, which in short, proposes a steady
feedback loop of serial feelings of satisfaction that fosters positive associations and builds a desire
to continue or return (Fredrickson 2001). Emotion scholars note that people have a baseline mild

35
level of positive affect that facilitates the execution of daily routines, which has also caused
challenges in detection of positive emotions in research (Diener and Diener 1996; Cacioppo et al.
1999).
There is a myriad of motivations for politicians to appeal to positivity in their campaigns,
among the most notable is its use by incumbent candidates who benefit from positivity’s role as
a deterrent to change (Druckman, Kifer, and Parkin 2009). The other notable and often desired
outcome is the development of “interest”. Establishing interest and engagement of constituents
is an ideal goal for any electorally minded actor, which may be preferred given the detrimental
experiences of negative emotions are subverted. Development of interest and engagement
through positive emotions is, however, not as linear as appeals to negative valenced emotions.
Mustering public interest through positive sentiment takes more time and energy and has been
shown to be an inferior instigator of online political engagement in comparison to negative
emotions, especially anger (Young 2023).
Despite the paucity in positivity scholarship relative to negativity, understanding the
patterns of positive emotional appeals that different RGID groups employ is an essential aspect
to capture. First, I posit that within the atmosphere of heightened electoral threat during an
election, the utility of positivity becomes more localized as a foil to the preferably employed
negative rhetoric. This becomes particularly noteworthy when considering the candidate's
specific social positioning, and it underscores the idea that candidates who might feel
marginalized or underrepresented might adopt a more restrained strategic approach in their
campaigns that could in turn be characterized by a stronger commitment to positive sentiment
messaging relative to an unconstrained white male candidate who may feel comfortable with an
unbridled approach to negative emotive appeals.

36
Secondly, and importantly, positive emotive appeals function to build interest and
sustained engagement of an individual in an activity suggesting at least one pathway through
appeals to positive sentiment that could help a candidate reach their electoral goals. This becomes
particularly salient for those averse to the risks of negativity appeals. Therefore, given its potential
to activate a voter base through appeals whose risk profile for a sender is substantially lower than
negativity, candidate-actors who are constrained in their use of negativity would be incentivized
to rely on a distinct pattern of positive valenced emotional appeals to reach their electoral goals.
Since the desired end point is activating their partisan base’s interest, these appeals are likely to
resemble previously described appeals to enthusiasm. Finally, due to their uniquely
disadvantaged identities in both racial and gender axes, BIPOC women actors are therefore the
most likely candidate-actors with subordinate social positioning to employ these appeals more
frequently than their white male counterparts.
Positioning, Privilege, & Strategic Communication
In the United States, the understanding of one's standing within the race-gender hierarchy
develops long before an individual decides to run for office. Driven by the social dominance
theory, this socially constructed model of identity starts at birth with matriculation into societal
groups based on external, straight-forward cues - often race and gender. These groups are
organized into a hierarchy with a single privileged/dominant group at the top and the remainder
are subordinate. This structure of privilege creates implicit biases and normative perceptions of
RGIDs (Sidanius and Pratto 1999), often forming denigrating stereotypes of subordinate groups
and beneficial stereotypes for the privileged group. These norms subsequently frame behavioral

37
expectations for RGIDs, placing inherent constraints on political candidates who hold
subordinate positions within this hierarchy (Bonilla-Silva 2006).
It remains clear that how race and gender may influence candidate strategy can be
understood through the lens of intersectionality, in which RGID are determinant of groups’
relative social positioning and public perceptions within the broader social context. Racialized
and gendered dynamics distinctly complicate the risk environment for racial minorities and
women who must navigate voters’ mixed expectations of different minoritized social identities.
Owing to their subordinate positioning in one or both social hierarchy axes, political actors
belonging to underrepresented intersectional identity groups are thus expected to assess and
respond to electoral threat in distinct ways from those assigned privileged positioning.
Importantly, the perception of strength shapes distinct perceptions of threat, and therefore
different risk assessments in the electoral environment. The privileged positioning within the
social hierarchy and its associated advantages significantly impacts the candidates' emotional
responses to perceived electoral threats. Research has found that those with higher status within
a group have a greater proclivity to respond to threats with anger than lower status members
(Tiedens, Elsworth, and Mesquita 2000). Thus, white male candidates, who enjoy privileged
positioning due to their intersecting identities, are more likely to respond to perceived electoral
threats with anger and corresponding appeals compared to candidates from marginalized and
underrepresented groups. This should result in a more pronounced use of anger in campaign
contexts and other influences that threaten their electoral safety, such as the competitiveness of
the district.
Underrepresented candidate-actors originating from historically marginalized
backgrounds confront a unique set of challenges, particularly when juxtaposed against their
white male counterparts in the privileged social tier. Status as a disadvantaged social identity in

38
either gender or racial identity has garnered scholarly attention interested in espousing and
explaining variations in behavior. The next section will discuss extant knowledge and
investigations in these identity realms.
Underrepresented Social Identities
Democrat challenger Colin Allred, who ran in 2018 for Texas’ 32nd congressional district,
describes that being a Black man means having to “cross more T’s and dot more I’s” in his
messaging and strategies (Herndon 2018). He adds:
“There’s some topics I have to talk about differently, and there’s some subjects
where I’m probably not going to be the one to lead on the issue. You have to be
honest and say, ‘You know what, as a young Black man I’m probably not going to
be the flag-bearer for that’ (ibid).”
This personal account demonstrates plainly some of the unspoken yet impactful ways in which
racialized experiences amongst candidates from marginalized and underrepresented9
backgrounds can arise from both internal as well as external processes. The various forms of
systemic oppression that make up the unique racialized histories of Black, Indigenous, Asian,
Latino, and other people of color in the United States inform not only the type of racism they
experience, but the racism they internalize about themselves as well (Alvarez 2020; Hua and Junn
2021). As Allred suggests, the campaign strategies and rhetorical choices available to
underrepresented candidates are ultimately limited when their campaign must be able to
9 The term "underrepresented" is used to signify the underrepresentation of conventionally marginalized
groups in governing institutions following longstanding traditions of gatekeeping and other systemic
forms of racism and sexism. At the time of this writing, nonwhite racial minorities and women remain
underrepresented in Congress despite making notable gains in recent elections.

39
successfully demonstrate their viability not only as an elected representative of a particular
district, but as a legitimate political actor as well.
Pointing explicitly to his race and gender identities as a "young Black man", Mr. Allred
highlights how social positioning is a foundational force in shaping the perception and strategic
assessment of what messages and types of rhetoric to use, and just as importantly, to avoid. This
provides further insight into how candidates’ experience and perception of the electoral
environment are formed at the intersection of race and gender. Facing the unique challenges that
come from operating without the assumed legitimacy and familiarity granted to those with
privileged social positioning, underrepresented candidates encounter additional electoral risks
that their campaigns must strategically navigate. These candidates must not only meet standard
political expectations but also transcend voter prejudices and stereotypes that are inextricably
tied to the intersectional nature of their racial and gender identities.
Race and gender stereotypes10 often serve as cognitive shortcuts for voters. Facing voter
expectations that often reflect prevailing identity-based stereotypes, meaning underrepresented
candidates and their campaigns must tailor their strategies to present themselves in ways that:
conforms to voters’ stereotypical beliefs (Kahn 1996, Dittmar 2015, Ellemers 2017, Sanbonmatsu
2002, Plant et al. 2000), meets their expectations or established norms of officeholders (Dittmar
2015), and avoids triggering voters’ negative identity-based perceptions (Cryer 2020).
10 See: Sanbonmatsu 2002; Huddy and Terkildsen 1993; Dolan 2014; Hatemi et al. 2012; Bauer 2013; McElroy
and Marsh 2010.

40
Gender Effects
Perceptions of emotionality or gender-emotion stereotypes create inequitable work
environments by imposing constraints on women in both how much and what kind of emotions
they are seen to display (Brescoli 2016). Prior studies underscore the salience of gender in shaping
emotional expressiveness in both nonverbal responses (Fischer and LaFrance 2015) as well as
communicative patterns (Cormack 2016; Fridkin and Kenney 2014; Gershon 2008; Dolan 2010).
Other studies have demonstrated a gendered divide regarding emotional rhetoric where women
leverage language to forge voter relationships, while men utilize it to reinforce their social
standing (Brunel and Nelson 2000; Kemp, Kennett-Hensel, and Kees 2013).
Women candidates are also subject to certain gender stereotypes of emotions that present
a fundamental barrier to their ability to successfully ascend to leadership positions (Brescoli
2016). Scholarship has described a greater proclivity for women political actors to express
sadness, fear, or shame, whereas men have been more inclined to express anger and other
confrontational emotions (Fischer et al. 2004; Hess, Blairy, and Kleck 2000; LaFrance and Banaji
1992; Plant et al. 2000). Furthermore, other studies have found that incongruent expressions of
gender-stereotyped emotions by organizational leaders affected assessments of their
effectiveness - i.e., male leaders receiving lower effectiveness ratings when expressing sadness
and female leaders receiving lower ratings when expressing either sadness or anger (Lewis 2000).
This body of work underscores the notion that established gender identities serve as key
drivers for such emotional appeals (Johnson 2013), linking emotional rhetoric closely to
prevailing gender stereotypes (Bauer 2015). Otherwise making clear that stereotypes and
socialized expectations about an individual’s perceived identity underlie how emotional
expressions are perceived and understood in society.

41
Given the gender-stereotyped emotions that may affect potential voters and electoral
success, women political actors are expected to be constrained in their online public discourse in
a way that conforms to gendered society standards.
In light of the corpus of literature describing gender-stereotypes with regards to
communication and emotion, several studies have found conflicting gender attitudes across
various political components (Burrell 1994, 2005, 2008, 2014; Fox 2018, Hayes and Lawless 2015,
2016) that also puts into question the emotive gap described earlier in this section (Bystrom 2006;
Sapiro et al. 2011). These mixed findings suggest an incomplete capture of gender’s effect on these
variations. Notably absent from extant scholarship is the incorporation of racial identity as an
intersecting factor with gender - which I suspect, if incorporated, could provide novel insight for
at least some of the controversial gender-identity scholarship.
Race Effects
Historically, underrepresented candidates have grappled with inequitable challenges
during certain electoral procedures such as ballot inclusion (Phillips 2021), a reality inherently
intertwined with the hierarchical structures in the broader societal context. The racial identities
of these candidates, coupled with societal racial attitudes and prejudices, play a pivotal role in
how voters receive, interpret, and engage with information related to them. Prior research has
documented that even subtle racial cues can lead white voters to harbor biases against Black
candidates, indirectly affecting the perception of the candidate’s political leanings (Berinsky et al.
2011). Beyond political attitudes, race is also shown to influence behavioral outcomes like vote
choice but this is dependent on various contextual factors including a candidate’s campaign style
and previous record (Citrin, Green, and Sears 1990).

42
Extant literature has revealed that racial minorities are evaluated by different criteria and
face harsher electoral repercussions (Cryer 2019; Krupnikov and Piston 2015; Piston et al. 2018)
that may precipitate as a result of the multiple axes of racial discrimination that can occur on a
single nonwhite minority candidate (Zou and Cheryan 2017). In light of the disparities in how
marginalized candidates are perceived and their messages received, compared to the "expected"
candidate (i.e., white male), racial minority candidates must negotiate racial stereotypes amongst
the electorate while still appealing to the partisan identity of their base.
The use of race as a heuristic affects how voters make political decisions and assess
minority candidates (Hutchings and Valentino 2004), which often lead to incorrect and racialized
interpretations of the candidate's qualifications (Huddy and Feldman 2009; Tesler 2012), issue
stances (Schumaker and Burns 1988), and even ideological leanings (Lerman and Sadin 2016).
Black candidates in particular are judged and penalized more harshly by non-Black respondents
for employing what are traditionally viewed as unsavory campaign strategies, such as negative
campaigning (Reeves 1997; Terkildsen 1993; Piston 2010; Krupnikov and Piston 2015). This ‘race
tax’ is especially sensitive when levied by white constituents to the point of punishing Black
candidates’ neutral stances on issues (Piston et al. 2018).
Emerging research in social and political psychology offers valuable insights that could
help elucidate racially stereotyped emotions and actions. Davin Phoenix (2020) uncovers a
differential reception of expressed anger based on the race of the sender. Phoenix describes how
anger expressed by white individuals is often met with responses aimed at mitigating the emotion
appropriately, whereas anger expressed by Black individuals elicits different, and sometimes
unforeseen, reactions (Phoenix 2020). This underscores the salient role that race plays at the
individual communication level, critically affecting reception and processing of emotions. Given
the profound impact of race on sender and receiver dynamics, we would expect nonwhite

43
minority actors to be more likely to adopt a constrained approach in the use of provocative
appeals in their online rhetoric than white political actors as a product of the heightened burden
of electoral risk.
In navigating the added hurdles of having to attenuate prejudicial views among voters,
political actors from underrepresented race and gender backgrounds operate with greater social
constraints, prompting a constrained approach to avoid triggering electoral demise through a
more sparing use of higher risk strategies, namely negativity, relative to white male candidates.
In other words, strategic messaging—and what it means to be strategic—is informed by the
structural level risks and incentives individually perceived by the candidate.
Going negative on Twitter may not be a strategy that is universally beneficial for all
candidates considering women and racial minorities often face greater costs in employing such
tactics. However, as politics does not occur in a vacuum, isolated social identity effects are far less
likely than the simultaneous, intersecting effect of the racial and gender identities. Given the
contrasting findings from gender-identity literature, there is inadequate capture of effect in
previous models isolating this factor. I argue that integrating the racial identity vector would
provide a more inclusive framework of rhetorical constraint. Specifically, I posit racial and
gendered identities to have a meaningful intersection that would underlie distinct strategic
choices amongst political elites, such that the pattern of discursive rhetorical constraints from
BIPOC11 women would be distinct from all other race-gender combinations. In a similar sense,
11 Specific racial or ethnic group(s) data was also collected for representatives in the 113th to 116th
congressional tweets data set, with categorical indicators coded for the four largest groups—i.e., White (n=
400); Hispanic or Latino (n= 40); Black or African American (n= 49); Asian (n= 14). Given the overall lack
of racial variation in Congress that has historically been and remains a predominantly White institution,
the sample for each racial subgroup was notably still limited despite the relatively large time frame of the
study. Specifically, the 113th to 116th congressional tweets dataset contains a total of 400 unique White

44
the single identity disadvantaged statuses of BIPOC men and white women would produce
distinct rhetorical constraints for candidates classified as either RGID.
This theoretical framework offers a nuanced and comprehensive account of how race and
gender shape what representatives say that also captures the complexities in which these effects
play out in governance and campaigns with varying levels of competitiveness. These mixed
findings point to a more complex system at work, in which the effect of candidates’ racial or
gender identity on their strategic messaging is not necessarily homogeneous.
Moderating Influence of Partisan Norms
Political party affiliation has been the prevailing and often the most impactful dynamic in
much of extant congressional communication scholarship. As previous academic reasoning may
predict legislator behavior to hold constant between individual members of the same political
party. I contend that since social positioning is determined at birth for most individuals, the
privileges or constraints disbursed to a given candidate likely shaped their developing social
acumen in accordance with the lens corresponding to their individual racial and gender identities.
Thus, given the lifetime of societal constraints for minoritized candidates, individual candidates
and elites coming from underrepresented nonwhite racial minorities, these constraints would
likely continue to exert a noticeable effect even after elites have established themselves as
members of a given political party.
MCs, which is nearly four times the number of BIPOC MCs from all racial minority groups combined (n=
103 total). Considering the challenges for inference that arise when further specifications of race are
included in the analysis, the binary indicator ‘BIPOC’ is the primary race measure used to estimate the
effect of nonprivileged social positioning on rhetorical strategies in Congress.

45
The influence of partisan norms on congressional communication is crucial to
understanding the strategies and risk-taking behavior of candidates. Grossman and Hopkins
(2018) highlight that the two major parties, Democrats and Republicans, operate in distinct ways,
emphasizing the unique importance of conservative self-identity within the Republican Party.
The Democratic base comprises an aggregation of conscious social groups mobilized by identitybased interests, while the Republican voting base is characterized by a perception of being
mainstream Americans defending individual liberty and traditional morality against left-wing
ideas (Grossman and Hopkins 2018).
These partisan differences have significant implications for the rhetoric and appeals that
resonate with each party's base. Prior literature notes that Democratic campaign messaging tends
to focus on policy agenda (Benoit 2004, Sanbonmatsu and Dolan 2009, while Republican discourse
emphasizes character traits, particularly sincerity and morality (Benoit 2004). It is expected that
there are substantial disparities in the persuasive strategies employed by partisan elites and the
risks they are willing to undertake in their public outreach messages. Democrats, with a more
diverse voter base, can afford to differentiate themselves and adopt a more expressive approach.
In contrast, Republicans prioritize presenting themselves as co-partisans, making their party
affiliation the primary focus of their self-presentation and image priming. Deviating from party
norms and expectations can be considered high-risk, as it may not align with the norms and
expectations of their intended target audience, namely the Republican base (Grossman and
Hopkins 2018).
Although both Democratic and Republican parties have historically comparable
approaches in candidate-recruitment procedures that disadvantage women, the level of support
and networks available to woman candidates within these parties exhibits variability
(Sanbonmatsu 2002). The heightened focus on conservative ideology as a cohesive rallying point

46
for Republican voters, in particular, poses a distinct obstacle for women in the Republican
network. Prevalent gender-based ideological stereotypes suggest that women lean less
conservative than their male counterparts (Koch 2002; McDermott 1998). These stereotypes are
especially accentuated within the Republican Party, conferring a notable disadvantage, especially
considering that perceptions of a candidate's conservatism significantly influence Republican
voter inclinations (King and Matland 2003).
In today’s Congress, there remains a visibly partisan imbalance in the representation of
women---as Thomsen (2015) writes, “contemporary patterns of women’s representation have a
distinctly partisan flavor… [in which] the number of Democratic women in Congress has
increased dramatically since the 1980s, while the number of Republican women has barely
grown” (295). Unique party cultures that could nurture Democratic representation while
impeding Republican women's progress could have a significant influence not only on election
results but also on communication strategies (Elder 2012). Therefore, since incentives guiding
Democratic and Republican women's decision to run for office differ (Reingold and Harrell 2010)
in addition to discordant partisan ideologies, women candidates from Democratic and
Republican parties are expected to have different strategic approaches to their public rhetoric as
these women would aim to meet the ideologic standards of their respective parties. Specifically,
by adopting a more expressive approach, Democratic women elites would likely be more liberal
in their appeals to negativity than their Republican counterparts. Conversely, due to their
considerable party-driven constraints, Republican women elites could be expected to try and
embody the more traditional gender expectations of their political party by shirking appeals to
negative sentiment in favor of making public discursive appeals to positivity more frequently to
illustrate their commitment towards conservative ideology to their voter base.

47
My theory proposes a multimodal approach that involves the unique intersection of a
candidate's racial and gender identities, their social positioning within societal hierarchies (as per
social dominance theory), and their individual perception of their electoral environment
(informed by cognitive appraisal theory), significantly influences their strategic choices (drawing
on rational choice theory). My framework provides an explanatory basis for candidate behavior
in different contexts and across identities, which will be tested using a dataset comprising
candidates' public outreach messaging during their service in Congress and their strategic
messaging during electoral campaigns. My dissertation applies this theoretical framework to the
2020 campaign tweet data, extending its reach to include electoral contexts where candidates with
varying degrees of institutional vulnerability, i.e., challenger vs. incumbent, face heightened
threats.
A Dynamic Model of Congressional Communication
As the diversity of congressional candidates increases and social media becomes a key
element in political campaigns, there is a need for new perspectives on models of strategic
communication that consider the intersectional influence of social identities on shaping different
incentive structures and corresponding outcomes for individuals who benefit from systems of
privilege compared to those who are conventionally excluded from them. I propose a dynamic
model that applies intersectionality as an analytical framework to explain how race and gender
influence candidate-actors’ rhetorical strategies as a function of their relative social positioning
within group-based hierarchies. This positioning structures the perceptual lens through which
candidate-actors assess their available social resources and determine what I refer to as their risk

48
ceiling, which is conceptualized here as the capacity to take risks in their rhetorical strategy when
electorally vulnerable.
By integrating various theoretical perspectives, we can understand how intergroup
dynamics, shaped by the underlying systems and structures of the broader social context, inform
individual actors' perceptions of themselves and how they are perceived by others. This, in turn,
leads to distinct strategic approaches in rhetorical appeals and messaging choices when
candidates face perceived electoral threats to their political standing.
This framework incorporates social identity, social dominance, cognitive appraisal, and
rational choice theories to highlight their interconnections and implications for the political
landscape.
Within this framework, candidates' social positioning at the intersection of race and
gender within group-based social hierarchies is crucial. These hierarchies afford privileges or
impose constraints based on the candidate's assigned group. The social dominance theory
suggests that a socially constructed group hierarchy shapes a structure of privilege, categorizing
individuals based on visible attributes such as race and gender (Andersen 2003, 33; Kruks 2005).
This categorization places them in either the privileged majority group or the subordinate
minority group, granting advantages or imposing constraints accordingly (Gordon 2004;
Scheurich 1993; Nieto et al. 2010). Distinct party norms and ideologies further contribute to
unique group expectations, which impact the constraints placed on minority groups.
Understanding their social positioning within systems of structured privilege allows
candidates to assess their risk environment and gauge threats to their political survival through
the appraisal process. Underrepresented candidates estimate the risk in their environment in
relation to their offsetting resources, which are often only accessible to the privileged group.
Being denied these advantages results in a harsher risk environment for political actors in

49
subordinate social ranks. Ultimately, as rational actors, candidates balance potential benefits,
risks, and constraints to make strategic moves that best serve their self-interests and ensure their
political survival. Figure 2 provides a visualization of the processes within the framework.
My framework, as presented in

50
Figure 2, is a multimodal approach that interprets a perceived electoral threat to their
ultimate strategic decision. It involves the unique intersection of a candidate's racial and gender
identities, their social positioning within societal hierarchies (as per social dominance theory),
and their individual perception of their electoral environment (informed by appraisal theory),
significantly influencing their strategic choices (rational choice theory). In other words, it is an
appraisal process in which the candidate assesses their available resources to determine whether
their social positioning affords them the privilege—i.e., high perceived status as a desirable social
currency—to respond and overcome the threat.
Congressional elites motivated by their desire to secure reelection, focus their campaign
and in-office rhetorical strategies to develop a presentational style that primes certain image
perceptions to best appeal to their constituency or perceived audience. I argue that the social
positioning of representatives at the intersection of race and gender identities shapes distinct
perceptions of their risk environment. The racialized and gendered social processes underlying
the electoral and political landscape in which the communicative relationship is developed must
be considered to develop a more nuanced and accurate picture of how communication strategies
are formed between the sender and intended audience/target group. All communication,
particularly in the form of political persuasion, is fundamentally relational, developed with
consideration of how the sender of the message can best relate and connect with the target
audience.
Given the expectations drawn from the theoretical framework previously outlined in this
section, I now summarize the corresponding hypotheses that will serve as the basis for the
empirical chapters that follow.

51
H1. Candidate-actors with subordinate social positioning within intersecting
race-gender hierarchical structures will be more likely to demonstrate a
constrained approach in their discursive strategies, using negative valenced
appeals less frequently than those with privileged social positioning.
H2. Candidate-actors with the same subordinately positioned RGID will display
distinct partisan patterns in their discursive strategies, adjusting their
rhetorical approach to meet party-specific expectations and norms.
H3. Candidate-actors with subordinately positioned RGIDs who are also running
as a non-incumbent challenger will be particularly more likely to display
greater constraint in their discursive strategies and use negative valenced
appeals less frequently than challengers with privileged social positioning,
i.e., white male challenger candidate-actors.
H4. Republican and Democratic BIPOC men, BIPOC women, and white women
will have distinct social constraints on their rhetoric that will lead to
diminished use of negativity by Republican actors with subordinate social
positioning, and increased use of negativity by Democratic actors with
subordinate social positioning relative to one another.
a. -- Political actors with subordinately positioned RGID will rely more
frequently on positive valenced emotional appeals than their white
male counterparts and will serve as a secondary measure of rhetorical
constraint that is inversely correlated to negativity.
H5. In the context of heightened electoral threat, challenger candidates with
subordinately positioned RGID will rely on positive valenced emotional
appeals substantially more than their white male counterparts, with a
positivity gap that is wider than differences seen outside of the campaign
period given the heightened constraints within this electoral context.
H6. Political actors from the same RGID but different political party affiliations
will produce distinct patterns and levels of positive emotive appeals that are
shaped by the similarly distinct partisan expectations and biases that color
the lens through which their corresponding constituencies view and evaluate
them by.

52
Figure 2. Detailed theoretical framework of strategic congressional communication.
CONSTRAINED
CONSTRAINED
ENVIRONMENT
ELECTORAL THREAT
| STRESSOR
IDENTITY GROUP
INTERSECTIONAL
GENDE
R
RACE &
ETHNICITY
POSITIONING
PRIVILEGED
NONPRIVILEGE
D
POSITIONING
I
SOCIAL IDENTITY
II
SOCIAL DOMINANCE
III
APPRAISAL PROCESS
IV
RATIONAL CHOICE
GROUP
SOCIAL
POSITIONING
TAKE RISK
EXPRESSIVE
AVOID RISK
BIPOC
WHITE
MAN
WOMAN
PERCEPTION FILTER
ASSESS
RESOURCES
AVAILABLE
INSUFFICIENT
RESOURCES
RESOURCES
SUFFICIENT
STATUS
PERCEIVED
HIGH
APPROACH
STRATEGIC
VALUE
SOCIAL
POSITIVE
STATUS
PERCEIVED
INFERIOR
VALUE
SOCIAL
NEGATIVE
POTENTIAL
'COPING'
LIMITED
RISK
ACCOMMODATE
IN POSITION TO
Triggers risk appraisal
process
hierarchies.
shaped by positioning within social
Perceived status of individual is
assessed
Significance of threat perceived and
relative to individual
and/or extended self.
+ SOCIAL
CONSTRAINTS
group
Assignment of privilege by
membership
CONSTRAINED

53
Recognizing that baseline differences exist in terms of how an individual sender’s social
status and value is perceived by others within the broader context is essential to understanding
the incentive structure candidates are operating from—i.e., the constraints and amount of agency
afforded by their social positioning—when forming their rhetorical strategies. White male
candidate-actors, as a product of their assigned group membership and social positioning at the
top of both racial and gender hierarchies, are least constrained by social perceptions, affording
them the full range of strategies and take a more expressive approach in their rhetorical appeals
relative to nonprivileged groups. This is particularly relevant in electoral contexts and public
office more generally—”an arena where men [white] have always been active” (Mandel 1983,
165), thereby reinforcing gendered expectations of candidate roles that ultimately remain
“consistent with established social patterns” (Dittmar 2015). Seeking to develop constituent trust,
congressional representatives make strategic choices in their self-presentation styles, ranging in
substantive emphasis, frequency of contact, and tone.

54
Chapter 3
Data and Methods
In this chapter, I describe the design and methodological approach to conducting this
research. Leveraging two original and comprehensive sets of congressional social media data, I
used computational text analysis methods to examine the relationship between social identity
and the use of emotive appeals in congressional rhetorical strategies on Twitter. Collectively, the
two sets of data contained over three million unique tweets (combined N= 3,042,200 tweets) from
1,467 distinct congressional actors. The first data set contained 551 incumbent representatives
serving in the 113th to 116th Congresses and 916 challenger candidates running for Congress in
2020. With data spanning across four separate congresses beginning at the start of the 113th
Congress (January 3rd, 2013) when Twitter was first widely adopted in Congress), up until
Election Day of the 116th Congress (November 3rd, 2020), I was able to examine congressional
outreach over-time and across varying institutional and electoral contexts in addition to
intergroup differences.

55
Congressional Tweets as Data
To undertake an empirical evaluation of my theoretical framework and the associated
expectations, I chose Twitter as the primary data source. Given the broad-based incentives on
Twitter for elected officials to strategically shape their public outreach in a manner that captures
attention and fosters engagement, the communication from Congress members on this platform
provides an apt and intriguing information environment for investigating the entire gamut of
candidates' rhetorical strategies, especially those pertaining to emotive rhetoric and appeals.
Another key benefit of using tweet data is that the character limitation forces political
actors to make tradeoffs in what rhetorical appeals and strategies they use in their messaging so
we can infer from patterns of emotive rhetoric that they are making the choice to focus on taking
an expressive approach (using negative valence emotions) or a constrained approach (positive
valence emotions), which the latter would suggest an overall avoidance to negative emotive
appeals.
While there are numerous web-based information environments that would also be
interesting and relevant to look at in this study (e.g., official and campaign websites), being
limited in both time and resources prompted me to narrow the focus to a single platform, given
the varied expectations that would form from different platforms and information environments.
Communication occurring in these static web environments typically does not confer notable
electoral benefits for candidates, as social media platforms often do. Thus, candidates are
presented with little anticipated reward for embracing risky and attention-drawing rhetoric as a
strategy in these environments, since it is more likely to generate adverse effects. I choose to not
focus on them in this study due to their inherent static nature. These platforms have limited
interactive or engagement components, and likely perform a function for campaigns that differs
from that of interactive social media platforms like Twitter.

56
According to Druckman et al. (2009), employing interactive platforms for campaign
messaging is a strategy used by disadvantaged candidates to generate news coverage. This
indicates that the spectrum of rhetorical strategies and the level of risk undertaken in public
communications are likely to vary on Twitter. If candidates were to experiment with diverse
campaign messaging strategies and deploy more emotional appeals, they would likely do so on
Twitter's interactive and dynamic platform. These characteristics suggest that the unique
functionalities and information environment of Twitter as a communication platform make it an
ideal source to comprehensively examine across varying conditions and contexts the range of
strategies and rhetorical choices representatives make in their public outreach.
This enables me to comprehensively test my framework examining the extent to which
the structure-agency dyad inherent to representatives’relational communications are constrained
by their disadvantaged political and/or social positionalities.
The Curious Case of Twitter
The unique interactivity of Twitter offers platform-specific incentives for users to curate
messaging that garners greater engagement and interaction as a means to ensure their content is
seen on algorithm-ordered newsfeeds while also potentially extending the reach of their message.
This creates systematic incentives on Twitter for office-seeking politicians to employ attentiongrabbing tactics in their strategic communication as they compete for virality in a crowded
information environment.
While the communication platform itself has a structural influence on representatives’
strategic considerations and public outreach, this influence may not affect all candidates in the
same way. Specifically, candidates and elected officials from historically underrepresented and
marginalized groups face more constraints in what they are able to say and do, suggesting

57
differences in the way they communicate and facilitate engagement on Twitter. Indeed, the battle
for coverage incentivizes attention-grabbing rhetoric—but for many underrepresented
candidates, the perceived risk of triggering racialized and/or gendered prejudices held by the
public can often far outweigh any potential reward.
The emergence and subsequent prevalence of social media, especially Twitter, as a
conduit for public discourse, have transformed it into an invaluable resource for studying
congressional communications. Prior studies using social media as data have notably focused on
politicians’ issue emphasis (e.g., Hemphill and Roback 2014; Shapiro and Hemphill 2017) and
responsiveness to constituents (Barbera et al. 2019). The granular nature of this data facilitates the
discernment of rapidly evolving temporal shifts in topic prominence. In addition to that an
overwhelming majority of the U.S. Congress members are avid Twitter users.
While acknowledging the constraints inherent in dictionary or bag-of-words models for
quantitative text analysis, these models, when anchored within a theoretically informed
framework, can yield profound insights (Grimmer, Roberts, and Stewart 2022). This is especially
pertinent when dealing with highly detailed data sources such as social media.
The methodological contribution of this research lies in the compilation of a vast dataset,
encompassing various district-level variables over an extended timeframe. Through this
expansive and comprehensive data collection, this study harnesses the wide-reaching and
immediate nature of social media communication. It allows for a nuanced analysis of the
intersection of race, gender, and political communication, both within and across electoral cycles.
(1) 113th to 116th House Representatives’ Office Tweets
For the purposes of my analysis, the primary dataset encompasses a comprehensive
compilation of public communications disseminated by elected officials via their formal House
member Twitter profiles spanning the 113th to 116th Congresses, from 2013 to the end of 2020.

58
Anchored by a theoretical impetus to probe the nuanced individual-level variations exhibited by
office-seeking political figures in their deliberate communications, this study narrows its lens to
spotlight the in-office communication endeavors undertaken by members of Congress (MCs)
within the U.S. House of Representatives. This specific focus on House MCs' official accounts is
strategic, designed to harness the richer individual and contextual variations afforded by the
chamber’s notably expansive—and more diverse—legislative body, complemented by its shorter
term-length.
Methodically, I have amassed all public-facing missives relayed on these official House
Twitter platforms between 2013 and 2020, encapsulating the 113th through the 116th Congresses,
culminating in a dataset enriched with over two million tweets (n= 2,015,659). Chronologically,
for each congressional tenure, the tweet timeline initiates on the inauguration of the congressional
session (specifically, January 3rd of the inaugural year) and extends through to the designated
midterm election in November of the succeeding year. Please refer to the appendix for a more
detailed description of Tweet selection criteria.
Figure 3 plots the monthly average number of tweets per MC over time from 2013 to 2020
in which each row of panels displaying the frequency of tweets over the course of each congress.
From the facet grid plot (Error! Reference source not found.), there is a slight general uptick in
the average number of tweets during election years within each Congress with the largest
difference seen in the 116th Congress between 2019 and 2020—i.e., MCs tweeted approximately
67.3 times a month compared to 78.1, respectively.

59
Figure 3. Monthly average number of tweets per MC (113th to 116th Congresses).
Additionally, Figure 3 makes clear that the use of Twitter for congressional
communication has been steadily increasing over time with the average frequency of tweets per
MC during the 113th Congress being 49.6 tweets per month compared to 72.4 in the 116th
Congress.
There is a total of 1,535 distinct MC+Congress observations in the 113th to 116th
congressional tweet data. Error! Reference source not found. displays the number of tweets and
individual MCs by race and gender identity group. The table also includes the average amount
of tweets per MC for each subgroup. Despite never making up more than 20% of the House,
women were notably tweeting much more frequently than their male counterparts—on average,
congresswomen had an overall average of 5,040.2 tweets per representative compared to 3,685.4
tweets per representative amongst men in the 113th to 116th Congresses. The difference was

60
especially stark when looking at BIPOC women who had the highest average number of tweets
per MC amongst the groups (5408.8 tweets per MC), further highlighting the heterogeneity in
communication styles even within gender.
Table 1. Number of tweets and MCs by race and gender (113th to 116th Congress).
n Tweets MCs Avg. per MC
White Men 1,257,885 387 3250.3
BIPOC Men 271,945 66 4120.4
BIPOC Women 205,535 38 5408.8
White Women 280,294 60 4671.6
N 2,015,659 551* 17451.1
* Refers to the number of total unique MCs in data.
I divided each Congress by period to control for how time and proximity to reelection
may affect the types of emotional appeals elected officials share online. I created a categorical
variable that classifies the period (governing/ campaign) in which a tweet was posted. The first
annual session or the full first year after being elected was defined as the governing period. The
campaign period entailed the second session within each Congress, or the year leading up to each
respective reelection date. The dates of each period time frame by Congress12 (along with the total
number of tweets and tweet average per MC) can be seen in Table 2. After the tweet data set was
cleaned and merged with each MC's Twitter account bio information, I then added my other
variables of interest at the member or district level, which will be discussed further in a later
section.
Table 2. Time frame and average number of tweets per MC by Congress and period.
Period Start End Tweets Avg. /MC
113th Governing 01-03-2013 12-31-2013 226,304 559.9
Campaign 01-01-2014 11-04-2014 220,887 546.7
114th Governing 01-03-2015 12-31-2015 235,393 611.4
Campaign 01-01-2016 11-08-2016 266,030 663.4
12 Tweets outside of the congressional time periods were excluded from the sample.

61
115th Governing 01-03-2017 12-31-2017 326,640 796.7
Campaign 01-02-2018 11-06-2018 284,617 668.1
116th Governing 01-03-2019 12-31-2019 227,186 770.1
Campaign 01-01-2020 11-03-2020 228,602 769.7
(2) 2020 House Candidates’ Campaign Tweets
The second dataset contains all public tweets from incumbents’ official House and
campaign Twitter accounts and challengers’ campaign accounts from January 1st, 2020, to
election day on November 3rd, 2020 (total N= 1,263,353 tweets).
The analytical procedure unfolded in several stages, focusing on the aggregation of tweets
made by both challenger and incumbent candidates during the 2020 election for sentiment appeal
analysis. I assembled the 2020 campaign tweets by month and candidate, resulting in a
comprehensive dataset comprising 1,263,353 documents and 389,671 features. The small handful
of non-major party challenger candidates (i.e., not affiliated with Democratic or Republican Party)
in the initial scraped tweet data were excluded from the analysis given the limited sample (58
candidates; 52,316 tweets) and missing data issues. This adjustment brought the dataset to 1,343
candidate accounts and 1,568 tweets.
The data was then aggregated by candidate and month, culminating in 12,243 candidatemonth observations. In the final stage, I aggregated the data by account and month for incumbent
candidates for a comparative analysis between campaign and office tweets. This process yielded
a dataset with 13,874 observations, as shown in Table 3.

62
Table 3. Number of candidates, accounts, and tweets in 2020 congressional campaign tweet
dataset by race and gender (RGID).
Data Collection
To create the data set of congressional tweets used for the following analyses, I compiled
over 2 million distinct tweets from elected representatives serving in the 113th to 116th
Congresses from 2013 to 202013. With the help of a research assistant to collect the user handles of
115th and 116th congressional members’ public Twitter account(s), I used the list to scrape the
tweets of all accounts still accessible. Following this, I merged the data with tweets from the 113th
and 114th Congress that were kindly shared with me by close colleagues.
Prior to cleaning, the initial compiled data contained a total N of 2,212,334 tweets collected
from the public Twitter accounts of elected congressional representatives serving in the 113th to
116th Congress. There was a total of 551 unique MCs in the data across a span of four separate
congresses, which may serve as a reminder of the longevity and resilience that many incumbent
legislators must have to stay in office.
13 I am grateful to Pablo Barberá and Maggie Macdonald for sharing their data. Tweets from the 115th and
116th Congress were captured through Twitter’s REST API.

63
The original data included tweets from both official government Twitter accounts (N=
2,015,659 tweets), and additional campaign or personal accounts (n= 196,675 tweets)—the latter
of which being notably less prevalent in the sample, particularly in the earlier congresses. For
representatives with two separate Twitter handles found in the data, each account type was
evaluated and manually coded as either "office" or "campaign" in its functionality.
For the following set of analyses, I limited the sample to only official House government
accounts, excluding any congressional outreach shared from campaign or personal accounts,
which are less reliable as a standardized source of data across the study’s time frame. By focusing
the analysis to only public messages shared from congressional members’ office Twitter accounts,
this helped to establish a uniform base for comparison with regards to the specific context and
message function of all tweets in the sample. Additionally, standardization of account type also
allowed for general assumptions to be made about the intended function and use of official
government Twitter accounts as a tool for public outreach and governance more broadly.
This data allowed me to study how congressional outreach from specified office accounts
that may vary systematically during campaign periods, i.e., election years, when congressional
members’ capacity as elected representatives may take a back seat to their capacity as political
candidates.
Operationalizing Key Terms and Concepts
Threat Perception in Competitive Electoral Contexts
In the context of the proposed theoretical framework, perceived electoral threat serves as
the trigger that facilitates the risk appraisal process. Competitive electoral environments in which

64
campaign considerations are salient and engaged are often contextual drivers of distinct
rhetorical strategies for those in electorally vulnerable positions relative to those with greater
electoral security and safety.
In this dissertation, I use the term ‘campaign salience’ broadly to encompass various
conditions that capture contextual variation in legislators’ attentiveness and responsiveness to
the electoral environment while in office. Campaign salience represents the context that will
amplify candidates’ focus on electoral incentives and perception of threats to their political safety.
The model posits that campaign salient contexts will be the trigger for identities to become salient,
setting the strategic considerations underlying group-based appraisals and subsequent strategic
behavior.
Election years (‘campaign periods’) when representatives' campaign considerations are
more salient contribute to heightened perceived electoral threat and a more competitive electoral
context that incentivizes the greater use of campaign-oriented messaging strategies in
representatives’ public outreach official government public accounts (non-campaign) public
accounts.
Explanatory Variables
MC and District Level Indicators
The key independent variables included in the regression models presented in the results
section are representative-level characteristics such as gender, party affiliation, political ideology,
and race. I created factor variables for state (1 for Alabama,... 50 for Wyoming), Congress
(113,...116), and month (1 for January,...11 for November), to represent categories in the
population.

65
In addition to a member’s party affiliation, I also included controls for their ideological
extremity and leadership status. Existing research underscores the relationship between
heightened ideological extremism in congressional elites and the increased usage of partisan
rhetoric and appeals (Gentzkow, Shapiro, and Taddy 2016). I created the variable for ideological
extremity (‘id_extreme’) by taking the absolute value of DW-Nominate first dimension scores for
ideology14, which is originally coded with values between -1 (extremely liberal) to +1 (extremely
conservative). This transmuted the scores into a cleaner continuous scale ranging from 0 to 1
which is agnostic to partisan orientation. MCs with values closer to 1 are interpreted as being
more ideologically extreme while values closer to 0 are considered relatively more moderate (min
= 0.09; mean = 0.4445; max = 1). To measure ideology, DW-Nominate scores were used for each
MC within a specific Congress. By employing all unique MC-DW-Nominate combinations, the
"id_extreme" variable was generated to delineate the relative bounds of ideological "extremity" and
moderation. In order to capture any ideological shifts over time, each MC within a Congress was
assigned their DW-Nominate score for that specific year, considering that an MC's ideological
positioning might shift if they serve in more than one Congress. In total, there were 1,535 unique
combinations of individual MC and ideology scores within each congress. Most representatives
(i.e., 1,488 MCs) remained consistent across their respective congressional sessions, receiving the
same score in the first and second year. Yet, 47 MCs from the 113th to 116th Congresses have
separate scores within the same congress, suggesting ideological shifts at the individual level.
Previous work suggests that MCs in leadership positions are in a more prominent position
to be more openly critical and may use social media for agenda setting (Gilardi et al. 2022; Russell
2021). Given that congressional leaders have distinct responsibilities and corresponding strategic
14 Lewis, Jeffrey B., Keith Poole, Howard Rosenthal, Adam Boche, Aaron Rudkin, and Luke Sonnet (2023).
Voteview: Congressional Roll-Call Votes Database. https://voteview.com/.

66
incentives shaped by their formal responsibility to build successful and winning coalitions
(Russell 2021), I expected that they communicate and use Twitter differently in comparison to
rank-and-file members 15.
Campaign Oriented Indicators
After the tweet data set was cleaned and merged with each member’s Twitter account bio
information, I then added the other key demographic variables to the data. In addition to the
primary variable of interest (i.e., [1] period: which classifies whether a tweet is sent during the
[governing/ campaign] period within a congressional session), I also created variables for the
following member- or district- level characteristics: [2] incumbent: whether the MC is a
(freshman/ incumbent); [3] gender: whether the MC identifies as (female/ male); [4] minority:
whether the MC identifies with an underrepresented racial minority group (nonwhite/ white)16;
[5] competitiveness: whether the MC is in a competitive district (0/1); [6] party: whether the MC
is (Democratic/ Republican); [7] leadership: whether the MC holds a leadership position (0/1);
and [8] extremity: whether the MC is more or less ideologically extreme (scale of 0 to 1).
The cookOrder variable signified the competitiveness of a district based on Cook Political
Report's Partisan Voter Index. This measure is delineated on a scale from 1 to 4, with 1 signifying
'safe' districts where electoral outcome is nearly certain due to the pronounced advantage of one
party, and 4 signifying a 'competitive toss-up' district, where neither candidate holds a definitive
15 Congressional leadership positions that are included are: House Minority Leader, House Majority Whip,
House Majority Leader, House Majority Whip, and Speaker of the House. There are 6 unique MCs total in
the full congressional tweet data set coded as holding one of the aforementioned leadership positions (N =
28,012 tweets).
16 I also created a race/ethnicity categorical variable (African American/ Asian American/ Hispanic or
Latino/ White) for more nuanced analyses regarding the specific ways in which representatives’ racial
identities can shape their rhetorical strategies and representational styles in distinct ways. The current
sample size for each racial and ethnic subgroup, however, is notably limited– there are 400 unique White
MCs and 103 unique nonwhite MCs in the full data set (Black or African American, n= 49; Asian, n= 14;
Hispanic or Latino, n= 40).

67
advantage. This stratification allowed for nuanced analysis of how competitive electoral
landscapes influence the communication strategies employed by candidates seeking to secure the
favor of their constituents.
Meanwhile, the partyFit variable quantified the alignment between a candidate's party
affiliation and the political leaning of their district, as measured by the 2020 Cook's partisan lean
measure. This variable is derived after eliminating URLs from texts using tokenizers and utf8
packages, and subsequently creating a word count variable, denoted as WC, that quantifies the
number of words per tweet. To code the indicator for district party fit, a candidate is classified as
a 'fit' (yes) if their party affiliation aligns with the district's specific party leaning, and as 'not a fit'
(no) when it does not. This measure provided insight into the social positioning of representatives
vis-à-vis their party affiliation and the partisan leaning of their districts, thereby furthering our
understanding of the complex interplay between perceived electoral risk and communication
strategies.
Seat Competitiveness
Seat competitiveness was accounted for in the following manner: by incorporating the
Cook Partisan Voting Index (PVI) scores (2017) that measure the partisan lean of each district,
factoring in the results of the 2016 and 2012 elections. The Cook PVI quantified how each district's
presidential performance compared to the national average, considering the preceding two
presidential elections. The PVI score, updated in 2017, included the results from the November
2016 presidential election. Notably, the number of competitive districts or “swing seats” has
diminished substantially in the last 20 years—in 1997 the Partisan Voter Index categorized 164
districts with a rating between D+5 and R+5, accounting for more than a third of the House. This
figure was larger than the number of both strongly Democratic and strongly Republican seats.
However, after the extremely polarized 2016 election, only 72 districts fell between D+5 and R+5.

68
This constituted less than one sixth of the House and a 56 percent decline since 1997. This figure
also indicates a 20 percent reduction from just four years ago when there were 90 swing seats.
Control Variables
Within the regression models presented in this dissertation, certain factors were controlled
for, though they are not explicitly displayed within the tables. The encompassed controlled
variables are as follows: month and state fixed effects, account type, average followers, monthly
average tweet word count, logged total n tweets omitted from the table.
It is notable that the "monthly average followers" served as a proxy variable, conceptualized
to represent the name recognition associated with a particular candidate. Additionally, to account
for the variance in verbosity among candidates, the average tweet word count was incorporated
as a baseline control, ensuring the models adjust for those representatives who exhibited a
proclivity for more extensive tweet compositions.
Analytic Methods
Measuring Emotion-Associated Rhetoric
To address my hypotheses, I made use of natural language processing methods to
computationally classify both collections of congressional messages shared on Twitter spanning
four Congresses over seven years (i.e., 2013 to 2020). The use of automated text analysis was
appropriate given that I sought to examine race and gender differences in aggregate levels of
emotive expressiveness in congressional outreach messages online. Previous studies have
effectively employed computational methods to measure a wide range of rhetorical appeals

69
across a variety of contexts and applications such as moral17 and moral emotions (Brady et al.
2017), ‘grandstanding’ messages in committee hearings (Park 2021), and emotive rhetoric in
various domains of political discourse, such as parliamentary debates (Rheault et al. 2016) and
populist communications (Widmann 2021). A robust literature stemming from computer science
focus on developing computational approaches and tools for affective computing, that is, the
detection and measurement of both verbal and non-verbal expressions of sentiment and
emotional states (Cowie and Cornelius 2003; Ishizuka, Neviarouskaya, and Shaikh 2012; Picard
1997)
I used computational text analysis methods to analyze a collection of congressional
communications online, studying how office-seeking representatives and candidates make use of
emotional appeals in their governing and campaign strategic outreach. In contrast to manually
coding texts, computational approaches enabled me to work with all relevant texts across
multiple congresses, allowing for a more comprehensive yet nuanced look at systemic and
individual patterns in congressional communication.
The NRC Word-Emotion Association Lexicon includes keys for eight discrete emotions in
addition to valence categories for positive and negative sentiment. It remains the largest lexicon
of its kind at the time of this writing, and notably deviates from other approaches that largely
focus on the use of words that denote emotion or limit delineation to the broader sentiment
categories. The NRC lexicon that is relied upon for this research instead used crowdsourcing to
validate a larger set of words that people can associate with an emotion. I have included a more
detailed explanation of generation of the outcome variables in the appendix.
17 E.g., Sagi and Dehghani 2014; Sterling and Jost 2018; Brisbane, Hua, and Jamieson 2023; Johnson and
Goldwasser 2018.

70
The establishment of the dictionary object that was used in my analyses was informed by
how specific words correlated to Plutchik’s (2001) proposed set of basic emotions (Figure 4). The
wheel can be interpreted as follows. Similar emotions appear next to each other in the wheel while
contrasting emotions are diametrically opposite to each other; the radius indicates emotive
intensity; and the white spaces between the basic emotions indicate the complex emotions that
are formed by the combination of the adjacent basic emotions. Presented as a two-dimensional
circumplex model, the emotions wheel as seen in Figure 4 provides a useful framework for
understanding the range, intensity, and interrelations among the different emotions we
experience.

71
Figure 4. Diagram of Plutchik’s (2001) proposed set of eight basic emotions and corresponding
emotional states.18
Figure 4 visualizes Plutchik's theorized set of foundational primary emotions, encompassing joy,
trust, fear, surprise, sadness, disgust, anger, and anticipation. These human emotions are thought
to be evolutionary survival responses and are represented along a spectrum detailing both
intensity and their affinities with adjacent emotions. Beyond their inherent similarities, proximate
18 Based on figure 1 in Mohammad and Turney (2013).
optimism love
aggressiveness submission
contempt awe
remorse disapproval
annoyance ANGER rage
vigilance
ANTICIPATION
interest
ecstasy
JOY
serenity
admiration
TRUST
acceptance
terror FEAR apprehension
amazement
SURPRISE
distraction
pensiveness
SADNESS
grief
boredom
DISGUST
loathing

72
emotions can also converge to manifest complex emotional states. This schematic provides a
comprehensive framework for grasping the breadth, fervor, and interconnectedness inherent
among diverse human emotions.
The negative sentiment dictionary key contains 958 unique words while the positive
category has a total of 1,031 unique words. The resulting lexicon contains a total of 12,153 kfeatures (’kFeats’), or total word combinations. Table 4 displays the number of k-features by
emotion category.
Table 4. Number of k features by NRC sentiment or emotion category.
category kFeats
1 anger 1245
2 anticipation 837
3 disgust 1056
4 fear 1474
5 joy 687
6 negative 3316
7 positive 2308
8 trust 1230
12,153
Once the dictionary object was constructed, I began the process of preparing the tweet
text data for analysis, which involved cleaning and standardizing all tweet texts prior to applying
the dictionary. Given the generally limited amount of text a tweet can have to begin with (i.e., not
to exceed 280 characters), all preprocessing steps were considered and decided on with the
priority of maintaining message clarity and preserving as much original text as possible.
A more detailed and comprehensive description of all preprocessing steps taken to
prepare the text corpus for text analysis is included in the Appendix along with any additional
considerations weighed during the preprocessing stage.

73
Multivariate Regression Analysis of Rhetorical Strategies
Given the theoretical emphasis on the individual level, I estimate multivariate OLS
regressions of representatives’ average use of emotive and issue-related rhetoric on their official
House Twitter accounts from 2013 up to 2020 during the 113th to 116th Congress. A more detailed
description of my regression analyses and interpretation can be found in the appendix. To do
this, I converted the subset of office account tweet data so that the unit of analysis is at the
representative level within each congress. To preserve and leverage some degree of the wide
temporal variance afforded with the original tweet data, I aggregated the data by congress, MC,
and month, generating a total N of 32,788 unique MC+month observations. Specifically, I
calculated aggregate measures from each MC’s set of tweets, grouped by congress, year, and
month. Within a given congress, there were up to 22 distinct MC+month observations across the
two-year session.
Applying the analyses discussed in this chapter onto the 113th to 116th congressional
cohort and the 2020 campaign applied a rigorous and well controlled regression protocol that
evaluated online emotive rhetoric of political elites and approximated their perceived risks and
constraints when engaging in online public discourse based on their relative use of negative vs.
positive emotive appeals in their tweets. Multiple regression models were be run for each racegender RGID to test my hypotheses outlined in chapter 2 with interest as to their relative monthly
emotive appeals as it compared to the other groups. Possible correlations with the risk a House
representative appreciated was averaged over spanning three congresses as well as during a
setting of heightened electoral threat in the 2020 campaign year.

74
Chapter 4
Racing on a Staggered Track: Race-Based Constraints on
Negative Emotive Appeals
The communication strategies political actors use offers a rich subtext with substantive
insights into the systemic biases and challenges pervasive within society. This chapter aims to
dissect the intricacies of political rhetoric through the analytical prism of intersectionality.
Specifically, it is expected that MCs from underrepresented RGID backgrounds will navigate the
online discursive spaces with more constraint than their white male counterparts who benefit
from privileged social positioning. Operationally, I hypothesize that legislators from
underrepresented RGID backgrounds to be more sparing in their use of the comparatively riskier
negative emotive appeals on Twitter relative to white male legislators whose privileged social
positioning grants them latitude to adopt a more freely expressive approach.
By delving into these expectations and corresponding hypotheses, this chapter seeks to
enrich understanding of how intersectionality impacts the strategic calculus in political

75
communications, thereby contributing valuable insights to the broader discourse on political
communication.
Expectation: I expect that MCs from underrepresented RGIDs are likely to exhibit more constrained
strategic approaches in their political communications compared to more privileged groups.
Hypothesis: I hypothesize that MCs from the different underrepresented RGID backgrounds will differ from
the white males, who occupy the privileged social position of society, in the extent to which they engage in
the riskier negative emotive appeals within their public rhetoric.
Navigating Risk: Strategic Communication in an Evolving Congress
The public actions and communications of political elites are seldom spontaneous, but
rather the end products of meticulous strategizing by teams that weigh the risks and benefits
associated with various communicative avenues. Existing scholarship underscores that
candidates are acutely sensitive to the balance of risks and rewards, and this incentive
architecture profoundly influences their chosen communication strategies (Druckman, Kifer, and
Parkin 2009). The contexts from which these strategies arise, however, are multifaceted and often
contingent upon an array of variables. A core aspect of public outreach strategy is the emotional
tenor of a given message (Marcus 2002; Loseke 2009), with particular emphasis on campaign
negativity and negative valenced emotions (Lau et al. 2007) is the strategy of whether or not to
go negative.
While extensive research has examined the role of emotional appeals in congressional
campaigns (Maier and Nai 2020; Marcus, Neuman, and MacKuen 2000; Brader 2005; Ridout and
Searles 2011; Valentino et al. 2008; Gervais, Evans, and Russell 2020), with many studies
particularly focused on negativity as a key political strategy (Haselmayer 2019; Ansolabehere and
Iyengar 1995; Lau, Sigelman, and Rovner 2007; Druckman, Kifer, and Parkin 2010; Evans,

76
Cordova, and Sipole 2014), little is known about the conditions under which candidates may
systematically differ in their use of negativity as a rhetorical strategy, particularly when there is
good reason to believe that individual candidates are evaluated inequitably.
To discern the circumstances under which political actors opt for risk-laden negative
strategies, a historical survey examining past instances of 'going negative' and reflecting on what
may underlie the efficacy or failure of such appeals serves as a foundational bedrock for
constructing a behavioral framework. the visual representation of the framework that is a
redemonstration of the initial framework presented in chapter 1. It is followed by Figure 5, which
is the simplified version focusing on the non-privileged arm of the framework that I will focus on
in the next chapters as it is upon this framework that I introduce the disparate risk perception
that subordinate groups experience.

77
Figure 5. Diagram of individual perception and appraisal process of risk environment for
minority RGIDs from framework19.
19 This is an abridged arm of the larger theoretical framework (see:
Figure 2) that specifically highlights the risk appraisal process leading to rhetorical constraint that political
actors with subordinately positioned social identities adopt in response to perceived electoral threats.
+ SOCIAL
CONSTRAINTS
ELECTORAL
THREAT
ENVIRONMENT |
STRESSOR
III
APPRAISAL PROCESS
AVOID RISK
CONSTRAINED |
LOW RISK
PERCEPTION FILTER
ASSESS AVAILABLE
RESOURCES.
LIMITED 'COPING'
POTENTIAL
INSUFFICIENT
NONPRIVILEGED
POSITIONING
INFERIOR
STATUS
NEGATIVE
SOCIAL
VALUE
WOMAN
BIPOC
WHITE
MAN
BIPOC WOMAN
MINORITY
INTERSECTIONAL
IDENTITIES

78
Who Gets to Go Negative?
In 2013 when social media was catching steam as a new communication paradigm for
information dissemination20, Stieglitz and Dang-Xuan (2013) conducted a study to examine the
relationship between emotion and news diffusion in a social media setting—specifically, Twitter.
They found that emotionally charged messages were likely to be retweeted more quickly and
more frequently, leading them to suggest that companies and brands should design content that
attends to the priority of triggering emotions. Since then, the relationship between emotion and
information has been studied in politics across varying communication platforms, forming a
general consensus that negative emotions in particular spread faster than positive ones (Ferrara
2015)
Existing scholarship accentuates various merits of employing negative rhetoric: the
capacity to capture increased public attention (Druckman, Kifer, and Parkin 2010), extend the
reach of messages across digital social platforms, elicit and sustain stronger reactions (Soroka and
McAdams 2015) greater resonance of messaging (Derryberry 1991; Ito et al. 1998; Soroka 2014),
enhanced retention of information (Kanouse and Hanson 1972; Kanouse 1984; Ito et al. 1998;
Meffert et al. 2006), and heightened receptivity to persuasion under certain conditions (Vaish,
Grossmann, and Woodward 2008).
Additionally, the impetus to go negative in campaigns has commonly been associated
with candidates who are marginalized or disadvantaged in some manner—be they challengers,
underdogs, outsiders, populists, members of minority parties, or women (Kahn and Kenney 1999;
20 See Jungherr (2016) for a comprehensive review of Twitter use in electoral campaigns.

79
Druckman, Kifer, Parkin 2009; Lawless and Pearson 2008; Parsneau and Chapp 2017). These
studies elucidate that the phenomenon of embracing negativity often serves as a "desperation
strategy" predominantly employed by candidates who face steep electoral odds. For example,
incumbent candidates benefit from “incumbency advantage” (Fiorina 1989; Freejohn 1974;
Gronke 2000; Jacobson 2004; Lazarus and Reilly 2010) that is founded in greater voter familiarity,
thereby disincentivizing any moves that could jeopardize this stable footing by attracting undue
attention (Druckman, Kifer, and Parkin 2009; Ensley, Tofias, and de Marchi 2009). With less to
lose as the non-incumbent candidate, challengers are expected to take larger rhetorical gambles
in hopes of forcing opponent engagement and greater public scrutiny of the candidates—this
subsequently creates a significant gap between candidates in the amount of negative rhetoric is
used in their public communications (Druckman, Kifer, Parkin 2009). Yet, the dynamics of this
challenger-incumbent dichotomy are influenced by multiple variables that introduces variability
to this relationship, such as seen in competitive races, where incumbents must engage in negative
discourse to defend their seats (Auter and Fine 2016; Druckman, Kifer, and Parkin 2010). This
suggests there is more nuance towards understanding the incentive structures of political elites
that guide how individual representatives appeal to emotion that extends beyond institutional
dynamics.
Studies have also shown gender’s association in shaping the tenor of political rhetoric.
Evidence suggests that woman candidates are inclined to adopt more ideologically polarized
positions during campaigns, partly because they face a higher likelihood of attracting primary
challenges (Lawless and Pearson 2008; Parsneau and Chapp 2017). This suggests that social
identities can also influence political elite behavior, although beyond gender there has been a
relative dearth of literature incorporating the component of racial identity into strategic political
communication literature.

80
Understanding political elite rationale for adopting specific public communication
strategies is founded on the risks and incentives that these actors perceive. Aimed at maximizing
voter turnout, these rewards are intrinsically linked to the ability to persuade the electorate,
grounding the incentive structures of messaging within the dynamics between the speaker and
receiver. In this framework, the associated risks encompass variables that could precipitate
misunderstandings or misinterpretations among the audience, be it through semantic discord or
emotionally charged reactions elicited by elite discourse. Given that ineffectual communication
can precipitate the abrupt end of a political career, the judicious calibration of rhetoric becomes
an invaluable tool for political actors in traversing their unique risk environments.
Negativity’s Thorns
The widespread use of negativity in online political discourse may, in fact, belie the
electorate's overall distaste for such rhetoric (Geer 2007). Public attitudes surrounding negativity
in addition to the general uncertainty of public reactions to appeals to negativity by
unconventional candidates can make going negative a risky strategy to many candidates.
However, given the seemingly prolific use of this rhetoric in politics today, do these actors not
appreciate the burden of risk associated with this communicative strategy that may otherwise
debilitate others? Understanding risk environments within a political construct necessitates
understanding the origins of these socially constructed risks.
A considerable volume of extant literature explores the prevalence of negative sentiment
in political communication, the predominant focus has been on the conditions that catalyze the
employment of negativity. There has, however, been considerably less scholarly attention
allocated to understanding the circumstances in which candidates opt for a more tempered, nonnegative approach—particularly if said actor is in a disadvantaged position (i.e., not an

81
incumbent). Often extant literature assumes homogeneity between elites from the same political
party, however as gender research has shown, disparate social identities may influence on an
individual’s behavior in Congress.
Studies on the intersection of racial and gender identities in political communication have
largely been underdeveloped, with studies focusing on racial differences in this arena are scarce
at best. Scholarship in this area imply that going negative may not be a strategy that is universally
favorable across all RGIDs, particularly those excluded from the privileged group (Piston et al.
2018; Dittmar 2015; Averill 1982; Marcus et al. 2000; Phoenix 2020; Bauer 2017; Krupnikov and
Bauer 2014). As expressions of emotions by an individual may be processed in a different manner
by an outgroup than if the speaker-audience relationship was between ingroup members, so
surfaces the tendency to form negative impressions of actions taken by outgroup members
(Phoenix 2020; Hogg and Reid 2006; Tajfel and Turner 1986). Studies of racial cues note their
potential to activate negative racial attitudes in audiences that precipitate inequitable evaluations
and punishments by voters for candidate misdeeds (Tokeshi 2021; Piston 2018, Piston 2015).
These findings suggest that candidate, or sender, attributes have varying effects on voter behavior
so should not be considered in isolation to studies of campaigns and strategic communications
given that they are mutually reinforcing (Krupnikov and Piston 2015). Findings of gender and
race dependent effects on messaging underscores the saliency of intersectionality, specifically the
intersection of a political actor’s racial and gender identities, in fully appreciating how the
interplay of these identities assigned at birth influence how political elites engage in with the
public.
Drawing upon the theoretical framework (Error! Reference source not found., Figure 4),
I argue that elected officials originating from historically marginalized racial and gender
identities operate without the societal privileges afforded to the dominant RGID—white men.

82
According to the principles of social dominance theory, these privileges manifest as higher
perceived status and positive social value. Since these advantages are selectively allocated rather
than universally distributed, this perpetuates inequities in setting varied social baselines and
biases embedded in public perception, which in turn produces a spectrum of outgroup
perceptions of minoritized candidates by audiences that are accompanied by potentially unseen
rhetorical risks to an unconventional candidate based on their minoritized identities. As such, the
primary theoretical expectation that is empirically investigated in this dissertation is that political
actors from marginalized racial and gender backgrounds will be more likely to adopt a more
constrained strategic approach in their public communications, particularly when juxtaposed
against their privileged counterparts.
This constrained approach is fundamentally shaped by the perceived risk environment
inherent to the speaker. Within the realm of online political discourse, specifically on Twitter,
topics that command attention frequently appeal to negative sentiments. Such rhetorical choices
can enhance a candidate's memorability, but they also introduce risks, particularly if misapplied.
Rooted in social dominance theory, societal biases tend to shield the dominant group while
disproportionately disadvantaging subordinate groups. Given that members of dominant groups
serve as the normative baseline for political officeholders, they are often subjected to greater
societal leniency in contrast to the heightened scrutiny faced by candidates from non-dominant
racial and gender backgrounds. Varied audience expectations and evaluative criteria further
amplify this complex risk environment, rendering it replete with latent hazards for candidates
who diverge from the dominant white male archetype.
Given the theoretical expectation that political elites from historically marginalized
backgrounds—specifically, at the intersection of their race and gender identities—will be more
likely to adopt a constrained approach in their rhetorical strategy, I hypothesized that

83
congressional representatives from underrepresented race and gender identity backgrounds
would be less likely to take risks in their rhetorical strategy, using less appeals to negative
sentiment and negative discrete emotions of anger, disgust, and fear than white male
representatives with privileged social positioning.
In this chapter I test my hypothesis through analysis of RGID differences in congressional
strategic messaging on Twitter, focusing on the extent to which members from historically
marginalized racial and gender backgrounds are more constrained in the risks they take with
their rhetorical choices to present themselves to voters. I focus on the use of negative emotionassociated rhetoric as one of the primary outcome variables of the study as it represents a
comparatively higher risk strategy to the alternative, positivity. Additionally, extant literature
has thoroughly investigated negative emotions from which reliable measurements of effect have
been derived and included in the models I present in this chapter.
Leveraging two data sets of over 3 million unique congressional tweets, I examine the use
of negative emotion-associated rhetoric in the public messages of political elites shared across
electoral contexts with varying levels of perceived threat. Models were run to test my hypothesis
on tweets from House representatives from the 113th to 116th Congresses—across a time-period
evenly split by Republican and Democratic presidential administrations as well as alternating
periods of campaign and governance.
The Road Less Privileged
As a measure of rhetorical constraint that resulted from appraised electoral threat, risk
taking behavior/risk tolerance was estimated based on a given MC’s willingness to engage in a
more uninhibited, “expressive approach”. Operating under the premise that negative sentiment
is traditionally deemed riskier than positive sentiment, the use of negativity is conceptualized

84
here as an expressive strategic approach that I expect will be more frequently adopted by those
powerfully positioned relative to other intersectional identities, freeing them from the social
constraints imposed by racialized and gendered processes in the broader social context. Male
representatives who are powerfully positioned in the dominant race occupy the top of the social
order and are more expressive in their approach (will use more negativity) whereas the avoidance
of such riskier sentiment is classified as a constrained approach (will use less negativity), which
is expected to be championed by historically underserved populations.
To examine how this paradigm operates across intersecting racial and gender identities, I
conduct multivariate OLS regressions that compare how each underrepresented race/gender
background influences how MCs makes use of emotion-associated rhetoric relative to the
dominant white male category. As seen in Table 5, the race/gender coefficient labels with ‘(fct)’
each signify a non-baseline level of a factor variable in which the RGID of being white and male
is set as the baseline for comparison. The corresponding results allow us to see how varying
intersectional identities associated with nonprivileged positioning (by race, gender, or, both)
constrain representatives’ risk-taking behavior, shaping distinct rhetorical strategies that reflect
greater constraint in their public messaging. Since going negative has disproportionately higher
risks for BIPOC candidates and women candidates (e.g., Piston et al. 2018, Dittmar 2015), I
hypothesized that BIPOC men, BIPOC women, and white women would use negativity less than
white male MCs.

85
Table 5. OLS regressions of all 113th to 116th MCs’ monthly average use of negative valenced
emotions on Twitter (RGID as factor levels).
*p< .1; **p< .05; ***p< .01
Note: Congress, month, and state fixed effects, controls for leadership status, age,
monthly avg. tweet WC, and logged total n tweets omitted from table. SEs were
clustered by MC and Congress.

86
Figure 6. Plot of minority RGID (factor levels) effects compared to baseline level (white male)
from regressions of all 113th to 116th MCs’ monthly average use of negative valenced emotions.
Note: Models fit with OLS for each emotive category (4 total models plotted). Error bars
show the 95% confidence interval of SEs clustered by MC and Congress. Other model
specifications, controls, and fixed effects were omitted from plot.
In support of the hypothesis, the results shown in Table 5 show that BIPOC
representatives, regardless of gender group, significantly differed in their use of overall negative
sentiment and disgust-associated rhetoric from the established white male baseline that each of
the other three RGID levels are directly compared to. More specifically, the reported coefficients
for the first model in Table 5 shows that ‘BIPOC Man (fct)’ and ‘BIPOC Woman (fct)’ identities
were associated with using on average approximately 0.7% and 1.3% fewer negative-associated
words per tweet significant at the p< .05 and p< .01 level, respectively, than white male MCs. This
pattern is also seen for the use of disgust (Table 5), in which BIPOC men and BIPOC women are
Model: (1) (2) (3) (4)

87
also less likely to adopt disgust appeals in their public office tweets than white men, using about
0.6% and 1% fewer disgust-associated words per tweet on average (p< .01). These differentiated
patterns suggest the critical impact of legislators’ relative social positioning informed by race and
gender identities on shaping their political communication tactics.
Conversely, white women exhibited a statistically significant increase in their fear appeals
having approximately on average 0.7% more fear-associated words per tweet (p< .05). The
remaining coefficients for white women did not have statistical significance, indicating a
comparable utilization of overall negative sentiment, anger, and disgust relative to white men.
This result aligns with existing literature that has documented a rising trend in the employment
of negative sentiment by white women during campaign periods (Evans, Ridout and Searles 2011,
Lawless and Pearson 2008; Parsneau and Chapp 2017).
Fear as an emotion is inhibiting and is situationally associated with greater uncertainty
and less control, which can be a steppingstone by political actors towards offering their policies
as the protection against a potential threat21. To better understand why MCs from certain
intersectional identities may appeal to fear more than others, the highest scoring fear tweets were
inspected and provided in Table 6. The tweets from Rep. Clark—D, who is a white woman, and
from Rep. Gosar—R, a white man. Both make appeals to fear in their narratives as supporting
evidence for the stances that they elect to take on certain issues.
21 On the discrete emotion of fear, see: Marcus and MacKuen (1993); Marcus et al. (2000), Brader (2005,
2006); Valentino et al. (2008); Wagner and Morisi (2019); Lupia and Menning (2009); Clifford and Jerit
(2018); and Nadeau et al. (1995).

88
Table 6. Examples of fear-associated rhetoric in 113th to 116th MC office account tweets. 22
member text score
GOSAR, Paul “Please, take a moment to read this thread of tweets by my friend,
colleague and fellow Arizonan, @RepAndyBiggsAZ. This tragedy is real,
and the illegal immigration in our state and country is real. QT
@RepAndyBiggsAZ Yesterday was the three-year anniversary of Grant
Ronnebeck's murder. He was gunned down in Mesa on January 22,
2015, by an illegal alien. The illegal alien, a convicted felon, was free on
bond while facing deportation. "Grant’s Law" would have prevented this
terrible tragedy.”(2018-01-23)
Fear: 11
FLORES, Bill “This morning, USBP agents in Portland arrested a violent criminal for
assault. When searched, agents found a pipe bomb, fused explosive
device, machete & knife. This person wasn’t there to protest. They were
there to cause destruction and incite violence.”
(2020-07-06)
Fear: 7
CLARK,
Katherine M.
“This isn’t about politics, this is about solutions. Gun violence is a
problem in America. Repeat: Gun violence is a problem in America. If
we can prevent even one death by investing in more research, isn’t that
worth it? #Enoughisenough QT @axios Republicans on the House
Appropriations Committee blocked a proposal on Wednesday that
would allocate $10 million to fund the Center for Disease Control and
Prevention's gun violence research.” (2018-07-12)
Fear: 11
These eye-catching appeals draw in the attention of audiences effectively and help draw
out the question: what else can be done? These tweets reveal that making an appeal to fear can
catch audiences’ attention but are ineffective in politics without associating it with potential
solutions within audience control—in the realm of negativity and as seen in these sample tweets
as in the others, this is often done by pairing fear with anger.
The analysis using the race/gender factor variable served as a pivotal tool in isolating
divergent patterns of emotive appeals between the MCs with underrepresented intersectional
identities of the study and the dominant white male cohort. Although invaluable for such direct
comparisons, this analysis did not extend its scope to contextualize these identity driven
differences within a more expansive framework that will be discussed in the coming sections.
22 URLs were removed from tweet text for clarity. Unedited and complete texts for the included tweets are
available in the Appendix.

89
Examining the Interaction of Race & Gender
To probe the nuanced interplay of race and gender in the strategic communication
patterns among elite actors, the subsequent regressions were conducted with the aim of homing
in on the individual influences of gender and race while holding all other variables constant (see
Table 7). Significance in either the race or gender variable would indicate a clear demarcation
along gender (women vs. men) and racial (white vs. nonwhite) lines. Included in this analytic
model was an interaction term, designated as ‘woman*white’, aimed at discerning potential
moderating effects of one variable in relation to the other.
In the results shown in Table 7, the coefficient for the interaction term "woman*white"
captures the additional change in the dependent variable for white women compared to nonwhite
(BIPOC) women. It also represents the difference in the effect on the dependent variable of being
a woman between white and BIPOC individuals. A regression coefficient that is greater than zero
for this interaction term would signify that white women have a stronger association with the
outcome in question than do BIPOC women. Conversely, a coefficient less than zero would
indicate that either BIPOC women or white men display a stronger association with the outcome
compared to white women.

90
Table 7. OLS regressions of all 113th to 116th MCs’ monthly average use of negative valenced
emotions on Twitter (woman*white).
*p< .1; **p< .05; ***p< .01
Note: Congress, month, and state fixed effects, controls for leadership status, age,
monthly avg. tweet WC, and logged total n tweets omitted from table. SEs were
clustered by MC and Congress.
Statistically significant greater-than zero regression coefficients for the 'white' variable
point to an elevated inclination to employ overall negative sentiment, disgust, and fear appeals

91
among white candidates as compared to racial minorities. Conversely, non-significant
coefficients at the 'woman' variable suggest that gender in isolation does not serve as a significant
factor on the usage of negative rhetoric. Additionally, the statistically significant regression
coefficient at the interaction term 'Woman*white' corresponds to a 1% heightened propensity for
white female MCs to make fear appeals as opposed to their BIPOC female counterparts. This
aligns with my theoretical expectations by which MCs from underrepresented racial backgrounds
are less reliant on the riskier negative emotive rhetoric than their white counterparts.
The results of the described analysis revealed differences between white women and
BIPOC women. However, to understand how these different intersectional identities may impact
how MCs appeal to negativity compared to one another, a broader scope is required. This was
done by examining the effect of each RGID as a binary variable in separate regression models,
holding all else constant.
Race & Gender Constraints through Partisan Lens
To examine how different RGID backgrounds may distinctly influence patterns of
negative emotive appeals, separate regression analyses were conducted in which the group levels
of the RGID variable (i.e., white men, white women, BIPOC men, BIPOC women) were coded as
individual binary variables to signify group membership (coded as 1) or not (coded as 0). For
each emotive category (negative sentiment, anger, disgust, fear), model estimates (see Table 8)
are obtained from four separate models, each specified with a binary indicator of a single RGID—
that is, reporting the coefficient for the respective RGID included in the individual model.
Statistically significant regression coefficients that are greater than zero indicate a higher
propensity than the overall cohort to appeal to negative emotions, while coefficients lower than

92
zero indicate a lower propensity to go negative. Statistically non-significant coefficients indicate
equivocal negative emotive appeal behaviors.
Since candidates from racial minorities are graded upon a harsher scale with
disproportionate electoral punishment from an electorate that is predominantly white (Piston et
al. 2018), I hypothesized that BIPOC men and women MCs would appeal to negative sentiment
and negative valenced emotions at a lower rate than their white congressional counterparts.
Based on extant literature suggesting a shifting gender norm amongst Democratic voters
(Ridout and Searles 2011), I hypothesized that Democratic white women, whose discourse
involves an electorate whose party ideologies have more equitable views of gender, would be
more expressive in their rhetorical approach and would rely on negative sentiment appeals and
discrete negative emotive appeals to a similar or greater degree as Democratic white male MCs.
Since conservative ideology has more traditional gender views that recognizes certain behaviors
and traits as either masculine or feminine, I hypothesized that Republican white women MCs
would use less negative sentiment and negative valenced emotive appeals to their white male
copartisans due to the classification of negative emotions as more masculine in general. With
divergent political ideologies, white women from the Democratic and Republican parties were
also hypothesized to have a notable gap between the MCs of opposing parties who share RGID,
specifically, Democratic white women should use more negativity than their Republican RGID
counterparts. This partisan difference between white women would be represented by a
statistically significant coefficient that is below zero at the interaction term ‘white woman*
Republican’23.
23 A value less than zero at this term means Republican MCs of a given RGID uses negativity LESS than
Democratic MCs of the same RGID

93
Table 8. Model estimates of RGID*Republican effects from separate regressions of 113th to 116th
MC monthly average use of negative valenced rhetoric.

94
In this model, the statistically significant (p< .05) regression coefficient at ‘White Woman’
signifies their increased use of negative sentiment appeals on Twitter. In contrast, BIPOC women
who had a significant regression coefficient for overall negativity (p< .01), indicated that they
were the group who used negative sentiment appeals the least. Similar inverse relationships
between white women and BIPOC women exist at disgust and fear, which is suggestive that these
valenced emotions are points of constraint for BIPOC women and otherwise drive the differences
between divergent races in overall negativity measurements (Table 8).
The disparities between white and BIPOC women in their appeals to disgust illustrate a
primarily racial emotive constraint. White women share space with white men as the RGID
groups who used the highest amount of disgust-associated rhetoric in the 113th to 116th
congressional tweets, while BIPOC women tweeted with disgust the most sparingly (Table 8).
While it has not been described explicitly in literature, appeals to disgust were expected to be a
high-risk appeal for minority RGID representatives given its unsavory and judgmental tone that
taps into audience morality and may also bear the risk of potentially uncovering negative
outgroup attitudes that may backfire. A further exploration of disgust will take place in the
following paragraphs to further contextualize this difference in appeals between MCs from
disparate RGID backgrounds.
Appeals to disgust are often triggered through judgments of morality (Clifford 2019;
Clifford and Wendell 2016; Horberg et al. 2009), which can be used as a tool to characterize
outgroups. Hence, if a person harbors a preconceived negative bias against an outgroup, an
appeal to disgust is more likely to inspire indignation and political action against the said
outgroup (Gadarian and van der Vort 2018; Jost 2021; Kam and Estes 2016).
As disgust carries high moral impact (Gadarian and van der Vort 2018; Jost 2021; Rozin
and Fallon 1987; Kam and Estes 2016; Clifford and Jerit 2018; Aarøe et al. 2020), effective appeals

95
must come from individuals who garner their audience’s respect as a guide of moral behavior,
which is a facet that would logically privilege the dominant group regardless of party—who
benefit from the social resources and societal perception as one who can influence moral
judgements. Consequently, appeals to disgust can be perceived to be a high-risk rhetorical
strategy for political actors who do not fit the mold of a prototypical politician. This was
corroborated by the results presented in Table 8 use of disgust by BIPOC representatives in
comparison to white representatives of all genders. This is not to say that BIPOC representatives’
tweets are devoid of disgust, however when taking a closer look at each RGID’s use of disgust
laden rhetoric, clear differences are apparent in the way these appeals are used. Take Democratic
Rep. Maxine Waters’ (CA-43) tweet in Table 9, note how Rep. Waters’ tweet is tied to a certain
issue or situation as opposed to her counterpart who tweets directly to another person’s character
without association with any particular issue or political event. This relationship seems to be
consistent amongst MC tweets and can further contextualize this racialized constraint—this
warrants further quantitative analysis in future studies.
Table 9 presents some illustrative examples of how the generally sparing use of disgustassociated rhetoric was used in congressional outreach messages on Twitter during the 113th to
116th Congresses. The five example tweets shown in Table 9 capture interesting differences in
how underrepresented legislators, particularly those from marginalized nonwhite racial and
ethnic backgrounds, are more so integrating disgust-laden language to supplement and place
greater emphasis on their message (e.g., demanding that something needs to change), as opposed
to expressing disgust in a more explicit and direct way (e.g., targeting an opponent) that places
the emphasis more on the emotion itself than on the message.

96
Table 9. Examples of disgust-associated rhetoric in 113th to 116th MC office account tweets.24
member text score
BROOKS, Mo
(AL-5)
Brooks, Mo. (@RepMoBrooks). “.@realDonaldTrump legal team
defense brief reveals how egregious, illegal and flawed the Socialist
Dem #impeachment scam is. Socialist Dems should be ashamed of
fraud and election theft they foist on America. America: Defend and
fight for Constitution!” 19 January 2020.
Disgust: 6
WATERS,
Maxine
(CA-43)
Waters, Maxine. (@RepMaxineWaters). “Trump, now that the truth
has been exposed about the horrific conditions of children and
families seeking asylum at the Southern border, will you continue to
idly watch sickness, disease, death and disaster? Will you use this as
political fodder for your reelection? Bad Bad Bad.” 8 July 2019.
Disgust: 8
PASCRELL,
William J. Jr.
(NJ-9)
Pascrell Jr., William J. (@BillPascrell). “Thoughts and prayers to the
republicans who've excused every atrocity corruption crime cruelty
deception defamation disgrace felony fraud guilty plea hypocrisy
impeachment indignity insult lie libel scandal slander slur theft
treachery They own all of it” 2 July 2020.
Disgust: 10
McSALLY,
Martha
(AZ-2)
McSally, Martha.(@MarthaMcSally). “The #BornAliveAct builds on
an important bill recently signed into law in AZ. After the horrific
events of abortion practitioner Kermit Gosner who has was found
guilty of murder and manslaughter, we must take necessary steps
against the murder of live babies.” 19 January 2018.
Disgust: 5
BROWN,
Anthony
Gregory
(MD-4)
Brown, Anthony Gregory. (@RepAnthonyBrown). “The NRA's
surrogates blame everything for gun violence. They blame Ritalin,
but no, it's guns. They blame doors, but no, it's guns. They blame
godlessness, but no, it's guns. They blame video games, but no, it's
guns. They blame abortion, but no, it's guns. And
#EnoughIsEnough.” 22 May 2018.
Disgust: 8
These minoritized representatives more so weave in disgust to act as a supplement to their
tweets and are amongst the highest scoring in terms of disgust appeals— are examples of tweets
that are among the highest scoring in disgust coming from MCs from both parties and different
RGIDs. As demonstrated by a tweet shared by Republican Rep. Mo Brooks (AL-5) in Table 9, Rep.
Brooks uses disgust-associated rhetoric to stir intergroup partisan sentiment by explicitly linking
the “Socialist Dems” to what he describes as immoral and unjust actions, which in this case
includes the “#impeachment scam”, “fraud”, and “election theft”. Disgust-laden messaging as
24 URLs are removed from tweet text for clarity. Tables containing the unedited and complete text for each
referenced tweets are available in the Appendix.

97
this tweet artfully exemplifies is likely to be an effective appeal particularly amongst
conservatives who are shown to have higher disgust sensitivity than liberals (Inbar, Pizarro, and
Bloom 2009). The attention-grabbing and contagious nature of rhetoric associated with moral
emotions in social networks online (Brady et al. 2017) underscore the potential reward that the
use of disgust appeals can yield when able to activate approach- and action-motivating emotions
amongst voters such as moral outrage. While the precise intent behind Rep. Brooks' tweet remains
undetermined from the text itself, its provocative nature to those even beyond his own party is
palpable.
Notably, the incorporation of the binary race/gender variable facilitated the use of the
interaction term '[RGID]*Republican'. This served as an initial move towards understanding the
effect of political party affiliation within this framework. The analysis presented in Table 8
includes ‘[RGID]*Republican’ to evaluate if party affiliation impacts rhetorical appeals within a
specific RGID. The results from this analysis corroborated the hypothesis that white women from
disparate parties would have divergent patterns of appeals to negativity. Republican-affiliated
white women MCs tweeted fewer appeals to overall negative sentiment (-1.6%), anger (-0.8%),
and disgust (-0.8%) than Democratic white women, as informed by the statistically significant
regression coefficients at the interaction term “white woman*Republican”(Table 8).
Moreover, given that white women MCs from the 113th to 116th were generally more
reliant on negative emotive appeals than the other RGIDs, white women MCs from the
Democratic Party can be inferred to appeal to negative sentiment and valence emotions just as
frequently to white men MCs if not more, demonstrating a substantial divergence between white
women who differ based on their party affiliations. This greater proclivity to appeal to negativity
by white women that depends on affiliation with the Democratic Party aligns with party ideology
that is rooted in tenets of acceptance and equity, meaning speakers from race and/or gender

98
minorities to certain democratic audiences may be incentivized to be more expressive, and
occasionally, more negative to fulfill their audiences’ expectations.
Party Moderating Effect – Separate RGIDs Subset by Party
Elites from underrepresented RGIDs must carefully maneuver their partisan expectations
in relation to their social identities whose disparate expectations guided by different ideologies
of the Democratic and Republican parties establish distinct incentives for elites (Elder 2012;
Reingold and Harrell 2010; Bauer 2015; Krupnikov and Bauer 2014; Osborn 2012; Thomsen 2015).
These distinct incentive structures result in patterns of constraints and rhetorical appeals specific
to their intersecting social identities. Within the Republican Party, grounded in a conservative
ideology that often upholds traditional gender norms, one would anticipate a propensity towards
signaling conventional feminine qualities. In contrast, the Democratic Party's ideology, which
emphasizes acceptance and equity, offers incentives for deviation from traditional gender roles
(Krupnikov and Bauer 2014; Osborn 2012; Thomsen 2015; Russell et al. 2023).
To better delineate party-driven rhetorical tendencies among political elites with varying
intersectional identities, I also run separate OLS regression models estimating the individual
effect of each intersectional race and gender identity on the risks they take in their strategic
outreach messages on Twitter. To specify, for each model race and gender are specified as a single
binary variable corresponding to one of the four intersectional identities (i.e., ‘whiteMan’,
‘BIPOCMan’, ‘BIPOCWoman’, ‘whiteWoman’), in which values of ‘1’ indicate that a representative
is a perceived member of the particular intersectional race and gender identity group. This
produces four separate models per outcome category, meaning that the coefficient estimates
reported for each table shows the results from 16 distinct OLS regression models of either

99
Democratic or Republican MCs’ use of emotion-associated rhetoric in their office account tweets
from the 113th to 116th Congress.
Visualizing the Effects of MCs’ Intersecting Race & Gender Identities by Party
Table 10 presents the outcomes of models that evaluate each race/gender intersectional
category in relation to a binary variable, compared against the full cohort segmented by their
political party affiliation. Figure 7 is a visual representation of this data. Statistically significant
regression coefficients (p< .05) exhibited in the table delineate the distinct impact that
membership in a particular RGID exerts within each of the major political parties.
To clarify, while the models presented in Table 8 provide the interaction term
“[RGID]*Republican” to compare Republican and Democratic MCs of the same RGID, the models
in the following figures compare how a given RGID compares to the rest of their affiliated
political party in their use of negative emotive rhetoric.
For Republican representatives, their relative monthly negative sentiment appeals on
Twitter are represented by red plots and tails depicted in the graph. The nonsignificant p-values
reveal that Republican Members of Congress (MCs) within this analytical model indicate that
Republican representatives generally tweet with negativity similar to one another across RGID
groups (Figure 7). Such uniformity in rhetorical strategies could be indicative of the more
homogeneous race and gender profile characteristic of Republican elites, as gleaned from the
dataset at hand.
Table 11, Table 12, and Table 13 presents the regression data as it pertains to appeals as
they pertain to the discrete negative emotions of anger, fear, and disgust. Figure 8, Figure 9, and
Figure 10 are the respective visualizations of the aforementioned tables.

100
Within the Republican Party, significant differences in the use of fear by white women
and BIPOC men at opposite bookend extremes of fear may reflect the nuanced gendered and
race-based expectations that drive or inhibit decisions to appeal to fear. the only variances with
statistical significance (p< .1) among RGIDs in my analysis related to appeals to fear, as presented
in Figure 9. Specifically, Republican white women in Congress were 1.5% more likely than the
average Republican MC to engage in fear appeals, as demonstrated by the statistically significant
coefficient greater than zero. On the other side of the spectrum, Fear appeals were also
significantly associated (p< .05) with Republican BIPOC men, who were found to use less fearassociated rhetoric than the remainder of Republican MCs (-3.7%). The findings at the bookend
extremes of fear within this specific cohort may reflect the nuanced gendered and race-based
expectations that drive or inhibit decisions to appeal to fear.
On the contrary, a non-significant coefficient for white male Republicans indicates that
their utilization of fear in Twitter discourse aligns closely with the party average. Outside of the
diminished use of fear amongst BIPOC men, nonwhite minority Republican representatives did
not exhibit any statistically significant regression coefficients, a phenomenon potentially
attributable to the relatively lower powered Republican BIPOC cohort, in other words, a scarcer
social and partisan identity pairing. Given the considerably lower count of MCs fitting the BIPOC
Republican criteria, the data points in this subgroup were limited, leading to extended tails on
these plots that indicate a lack of statistical power to draw robust conclusions from this
subgroup's data.
For Democrats, negativity is represented by blue plots and tails for each corresponding
RGID group. Notably, negative sentiment within the Democratic Party varied significantly along
racial lines (Table 10). Statistically significant coefficients (p< .01), both greater and less than zero,
corresponding to white women MCs and BIPOC women MCs respectively, reveal a racial gap in

101
the utilization of expressive online rhetoric. This finding lends empirical support to my
hypothesis concerning the role of social identities in shaping rhetorical choices, particularly as
they diverge along racial lines. The gender effects appear more nuanced, especially concerning
white women's negativity, evidenced by a 1.9% greater inclination for white women from the
Democratic party to employ negative rhetoric on Twitter relative to the remainder of the
Democrats.
Similarly, in the Democratic Party, distinct racial patterns emerged in the use of discrete
negative emotions like anger and disgust. Statistically significant coefficients greater than zero
correlate with white women's 2.1% and 1.1% higher likelihoods to appeal to anger and disgust
respectively when compared to their party averages (Figure 8, Figure 9). Likewise, white male
Democrats demonstrated a 0.8% higher likelihood to employ disgust and fear appeals, supported
by significant coefficient values (Table 12, Table 13).
Conversely, both BIPOC men and women in the Democratic Party made fewer appeals to
the discrete emotions of disgust and fear. This observation aligns with existing literature, which
has previously noted an uptick in negative rhetoric among woman candidates during campaigns
(Evans et al. 2014; Evans and Clark 2016; Gervais et al. 2020). However, this study extends that
body of research by suggesting that the propensity for negative rhetoric among white women in
the Democratic Party may endure even during governance phases.

102
Table 10. Model estimates of RGID effects (by party) from separate regressions of 113th to 116th
Dem. & Rep. MCs’ monthly average use of negative sentiment.
NEGATIVE
term party estimate std.error p.value
White Woman Democratic 0.019*** 0.009 0.030
White Man Republican 0.013 0.009 0.166
White Man Democratic 0.012 0.008 0.110
White Woman Republican -0.010 0.011 0.374
BIPOC Man Democratic -0.013 0.010 0.174
BIPOC Man Republican -0.014 0.017 0.407
BIPOC Woman Republican -0.016 0.032 0.623
BIPOC Woman Democratic -0.026*** 0.009 0.003
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 7. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep. MCs’
monthly average use of negative sentiment per tweet.

103
Table 11. Model estimates of RGID effects (by party) from separate regressions of 113th to 116th
Dem. & Rep. MCs’ monthly average use of angry appeals.
ANGER
term party estimate std.error p.value
White Woman Democratic 0.021*** 0.007 0.004
White Man Democratic 0.005 0.006 0.408
White Man Republican 0.004 0.007 0.565
White Woman Republican 0.002 0.008 0.838
BIPOC Woman Republican -0.002 0.020 0.908
BIPOC Woman Democratic -0.013** 0.007 0.041
BIPOC Man Democratic -0.015** 0.007 0.035
BIPOC Man Republican -0.016 0.013 0.210
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 8. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep. MCs’
monthly average use of anger-associated rhetoric per tweet.
Note: Models fit with OLS for each RGID by party subset (8 total models plotted). Error
bars show the 95% confidence interval of SEs clustered by MC and Congress. Other
model specifications, controls, and fixed effects were omitted from plot.
Model: (1) (2) (3) (4) (5) (6) (7) (8)
White Man BIPOC Man BIPOC Woman White Woman
Democratic
Republican
**

104
Table 12. Model estimates of RGID effects (by party) from separate regressions of 113th to 116th
Dem. & Rep. MCs’ monthly average use of disgust appeals.
DISGUST
term party estimate std.error p.value
White Woman Democratic 0.011** 0.005 0.033
White Man Democratic 0.008** 0.004 0.046
White Man Republican 0.002 0.005 0.642
White Woman Republican -0.001 0.005 0.916
BIPOC Woman Republican -0.002 0.013 0.887
BIPOC Man Republican -0.006 0.010 0.586
BIPOC Man Democratic -0.010* 0.005 0.051
BIPOC Woman Democratic -0.014*** 0.005 0.003
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 9. Model estimates of RGID effects (by party) from separate regressions of 113th to 116th
Dem. & Rep. MCs’ monthly average use of disgust rhetoric.
Note: Models fit with OLS for each RGID by party subset (8 total models plotted). Error
bars show the 95% confidence interval of SEs clustered by MC and Congress. Other
model specifications, controls, and fixed effects were omitted from plot.
Model: (1) (2) (3) (4) (5) (6) (7) (8)
White Man BIPOC Man BIPOC Woman White Woman
Democratic
Republican
**

105
Table 13. Model estimates of RGID effects (by party) from separate regressions of 113th to 116th
Dem. & Rep. MCs’ monthly average use of fear appeals.
FEAR
term party estimate std.error p.value
White Woman Republican 0.015* 0.009 0.082
White Woman Democratic 0.011 0.008 0.141
White Man Democratic 0.009 0.006 0.123
White Man Republican 0.003 0.008 0.699
BIPOC Man Democratic -0.011 0.007 0.140
BIPOC Woman Democratic -0.015** 0.007 0.036
BIPOC Woman Republican -0.027 0.023 0.251
BIPOC Man Republican -0.037** 0.015 0.014*
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 10. Model estimates of RGID effects (by party) from separate regressions of 113th to
116th Dem. & Rep. MCs’ monthly average use of fear rhetoric.
Note: Models fit with OLS for each RGID by party subset (8 total models plotted). Error
bars show the 95% confidence interval of SEs clustered by MC and Congress. Other
model specifications, controls, and fixed effects were omitted from plot.
Democratic
Republican
Model: (1) (2) (3) (4) (5) (6) (7) (8)
White Man BIPOC Man BIPOC Woman White Woman
**

106
Within the Democratic Party, BIPOC MCs had statistically significant regression
coefficients indicating lower reliance on appeals to anger, disgust, and fear by BIPOC legislators
of both genders in addition to lower use of appeals to overall negative sentiment by BIPOC
women MCs (Figures 7, 8, 9, 10). These differences point to a significant constraint in the use of
these appeals, likely due to the potential risks of employing negative rhetoric as a nonwhite
minority (Piston et al. 2018). Observed constraints on the use of negative appeals within the
Democrats only subset is primarily associated with race, reinforcing the perception that party
expectations for nonwhite racial minorities hold considerable sway over their rhetorical choices.
For Republican BIPOC members, the results suggest a possibly lower reliance on fear appeals by
BIPOC men relative to the remainder of the Republican party (p< .1). Otherwise, the absence of
significant coefficients associated with Republican BIPOC women regarding their negative
appeals or at the interaction term ‘BIPOC woman*Republican’ does not necessarily mean that
Republican BIPOC women do not share similar racial constraints to their Democratic
counterparts; rather, the degree of difference cannot be conclusively drawn from Republican
BIPOC women owed to the fact that their cohort is much more limited and therefore more prone
to producing a wide range of variance that would otherwise mask notable effects. Party specific
patterns may arise if their cohort was adequately powered.

107
Analysis of the 113th to 116th MC cohort subset by party revealed one of the rare instances in
the study where anger was significantly increased and decreased in RGID comparisons to the
average of their overall subsetted cohort (Table 11, Table 11. Model estimates of RGID effects
(by party) from separate regressions of 113th to 116th Dem. & Rep. MCs’ monthly average use
of angry appeals.
ANGER
term party estimate std.error p.value
White Woman Democratic 0.021*** 0.007 0.004
White Man Democratic 0.005 0.006 0.408
White Man Republican 0.004 0.007 0.565
White Woman Republican 0.002 0.008 0.838
BIPOC Woman Republican -0.002 0.020 0.908
BIPOC Woman Democratic -0.013** 0.007 0.041
BIPOC Man Democratic -0.015** 0.007 0.035
BIPOC Man Republican -0.016 0.013 0.210
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 8. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep. MCs’
monthly average use of anger-associated rhetoric per tweet.
Model: (1) (2) (3) (4) (5) (6) (7) (8)
White Man BIPOC Man BIPOC Woman White Woman
Democratic
Republican
**

108
Note: Models fit with OLS for each RGID by party subset (8 total models plotted). Error
bars show the 95% confidence interval of SEs clustered by MC and Congress. Other
model specifications, controls, and fixed effects were omitted from plot.
). Of the Democratic representatives in the sample, BIPOC women and men are shown to
be more sparing with their use of angry appeals, having approximately 1.3% to 1.5% less angerassociated words per tweet, respectively, than their white co-partisans (p< .05). BIPOC men and
women representatives’ appeals to anger were both outpaced by Democratic white women—a
finding that potentially reflects the racialized differences noted by Phoenix (2020) in how anger
is received and responded to when associated with individuals from different race and ethnic
groups. Anger when told from the perspective of a Black individual, for example, is not received
similarly nor comparably to white-associated anger, which is typically met with a desire to quell
or abate the emotion. Existing literature suggests that emotional responses to emotional appeals
from an outgroup differ for subordinate race/gender groups compared to members of the
dominant group. The observed reluctance of BIPOC male MCs to employ anger-associated
rhetoric may be indicative of experiencing greater constraints in being able to leverage anger as
a strategy given the uncertainty and greater likelihood of negative public perception of such
appeals.
These observations also corroborate earlier studies that emphasize the profound influence
of political parties on MCs' behavior (Lindstädt and Vander Wielen 2012; Arbour 2014) and
demonstrate the moderating effect that political party affiliation can have on appeals to
negativity. The stark contrast in the use of negative appeals between white women in the
Democratic and Republican parties may be attributed to the differing ideological expectations for
women within these political entities. Adhering to more traditional gender stereotypes as
conservative ideological norms dictate (Cassese and Holman 2018; Banda and Cassese 2022;
Sanbonmatsu 2002; Huddy and Terkildsen 1993; Dolan 2014; Hatemi et al. 2012; Bauer 2013;
McElroy and Marsh 2010), Republican women attempting to navigate the gendered expectations

109
of their partisan ingroup may exhibit more constraint in their rhetorical choices to avoid
perpetuating detrimental stereotypes. Democrats, on the other hand, are shown to place greater
emphasis on the individual-oriented moral foundations of equality and fairness (Graham, Haidt
and Nosek 2009) typically associated with liberal ideological leanings. This suggests that
Democratic women—particularly when already in office—may have greater latitude to go
negative. Having to navigate gendered expectations that are less established and uniform as they
are within the true ideological vehicle that the Republican Party embodies (Grossman and
Hopkins 2016), Democratic women leverage more negative associated emotions in their rhetorical
strategies, which may reflect fewer partisan constraints on their expressive freedom.
Traditional academic discourse has typically presented an image of homogeneity among
elites within a singular party. However, my research counters this narrative, highlighting that
political actors' appeals are intricately influenced by the intersection of their racial and gender
identities. The present study demonstrates constrained use of certain emotive appeals, which are
conventionally deemed riskier, is not merely dictated by party ideologies that promote balanced
platforms. Instead, it is significantly impacted by the externally perceived identities of these
political figures.
Political party affiliation has extensively been cited as a primary driving force behind
many political communication observations with institutional positioning differences often cited
as the other keystone factor to explain political behavior. My findings up to this point
demonstrated how social identities have a key role within the theoretical foundation of this
discipline. Political parties continue to have substantial influence on behavior that interacts with
distinct social identities to determine how exactly these effects would manifest. It was therefore
imperative to determine how social identities may factor into extant knowledge of institutional

110
power dynamics, specifically how disparate RGIDs factor into the framework of the incumbentchallenger dynamic.
House of Anger
Contrary to what previous findings in campaign communication may suggest, were
minimal differences appreciated in the expression of anger across all analyses25, potentially
representative of its widely recognized effectiveness as a constituent motivator. This likely
reflects the pervasive usage of anger appeals by all MCs in the online domain and may reflect
shifting online norms that see anger appeals as business-as-usual in political communication.
The shared prevalence of angry emotive rhetoric seen across all types of messages often
meant that anger-laden tweets would also include other emotive appeals. To examine how anger
associated rhetoric was used, particularly in relation to other categories of emotive rhetoric, I
compute the Pearson correlation coefficient between scored category included in the analysis
(negative, anger, disgust, fear, positive, joy, anticipation, trust), as shown in the correlation matrix
plot presented in Figure 11. Each cell in the matrix is shaded according to the strength of the
correlation between the two emotion categories—specifically, coefficients further from 0 are
assigned darker shades of purple to indicate a strong correlation given that a perfectly linear
correlation has a coefficient value of 1. From Figure 11, we see that anger highly correlates with
fear associated words and overall negativity associated words, adding some support for the
notion that anger serves as a common resource and foundation political actors can build upon
and individualize through other rhetorical appeals.
25 A notable exception within the Democratic party as white women championed anger appeals for the
party and created a clear gap in anger appeals between themselves and Democratic BIPOC men and
women.

111
Figure 11. Plot of correlation matrix coefficients between emotion categories.
Note: The limits interval used for assigning color to the plotted coefficients is set to (0,
1) in which darker shades indicate greater correlation between variables and lighter
indicating less.
It is important to also note that as discrete negative emotions may evoke completely
different responses amongst voters and in a similar way to further complicate things, discrete
negative emotions may have a popular utility in terms of delivering a certain message, but certain
emotions may be used for several general end goals. In examining the ubiquity of Anger thus far,
I ask—does the ubiquity of the discrete emotive rhetoric produce different results when

112
examining high scoring tweets for anger? Or is there a more consistent pattern that these Tweets
generally follow?
Table 14 provides examples of anger laden tweets for further qualitative inspection. Take,
for example, the tweet by the female BIPOC representative, Rep. Valez Butler Demings, D-Fl 115
that is displayed in Table 14. Rep. Demings certainly uses anger associated words in this, as a
whole, appeal with the intention of evoking anger amongst her electorate and encourages them
to take action against gun violence. Representative Demings’ tweet was among the highest
scoring anger tweets (10) in the entire sample. Compare Rep. Deming’s tweet to Rep. Lee Zeldin,
R-NY / white man/ 116th congress/ whose focus on Israeli foreign policy as an evocation of
anger with first the presentation of disparate behavior of his Democratic colleagues, then a
specific call to Speaker Pelosi, and finally an appeal for to the public to take action against this
bill. These tweets are generally representative of the word content of appeals to anger—primarily
as a vehicle to discuss policy stances and to specifically call to the audience to take action and
affect the bill or issue as it pertains to the speaker’s preference. The data points to another
important predictive factor that influences an elite’s decision to appeal to anger, see the tweet
above by Eliot Engel, D-NY/white man/116th/ in which he raises attention to the issue of
silenced sexual assault victims while raising direct connection with Trump, R—this tweet not
only illustrates the presumed basic structure and function of appeals to anger, but also
emphasizes the tweeter’s political party affiliation and if it opposes the president. A tweeting elite
who is in the opposing party of the president would expectedly be predisposed towards making
appeals to anger in their calls to take action to supplant the current power seat.

113
Table 14. Examples of anger-associated rhetoric in 113th to 116th MC office account tweets.
member text score
DEMINGS, Valdez Butler
BIPOC Woman
Democratic
2018-07-20
Can you imagine, surviving a massacre and then a few
months later lose your father to #gunviolence? Two children
went from nearly being shot dead in their classrooms to
losing their father to senseless gun violence. We’re allowing
our children to deal w/ stacked trauma. QT @thehill Father
of Parkland survivors shot to death in robbery
Anger: 10
ENGEL, Eliot Lance
White Man
Democratic
2019-04-23
It is outrageous that the Trump Admin. would threaten to
vote against a resolution condemning rape as a weapon of
war. This is callous and an insult to every woman who has
been the victim of sexual assault. Survivors of rape deserve
our fullest support.
Anger: 9
ZELDIN, Lee M
White Man
Republican
2019-02-02
House Dems were falling over each other racing to House
floor to vote against white Supremacy in Jan. Only 1 NO
vote in House, but now House Dems are denying a vote to
condemn anti-Israel and anti-Semitic hate.
!@SpeakerPelosi #PassHRes72 and; condemn this hate as
well. Let’s vote!
Anger: 9
Chapter 4 Summary
My study fortifies the extant literature on the transformative role of Twitter in politics,
providing an unparalleled platform for candidates who have historically existed outside of
conventional political paradigms to gain notoriety and even electoral success. While the utility of
negative campaigning has been long established in political science literature for its attentioncapturing capacity while propelling engagement and partisan heuristic processing (Rydell and
Mackie 2008, Valentino et al. 2008, Brader 2005), my study extends the framework by providing
context as to how disparate social identities may not equitably have access to negativity’s
beneficial campaign attributes. Social identities, particularly at the intersection of racial and
gender identities, factor into varying online public discursive strategies in establishing distinct
incentive structures.

114
Drawing from the social dominance theory, I proposed that the privileged group in a
society as the only group occupying the top position of the social order (i.e., white men) benefits
from exclusive privileges derived from positively skewed societal biases. Conversely, all other
subordinate groups within the social dominance theory are casualties of negatively skewed
stereotypes harbored by the general public that perpetuates negative outgroup biases and
attitudes against these unconventional political participants, which is comparatively obstructive
relative to the benefits of the dominant group. Such intrinsic barriers start at birth and fortify
these implicit barriers that may cement subordinate social positioning for individuals from
underrepresented RGIDs, potentially limiting career progression, political campaigning, and
public messaging (amongst others).
These societal structures are ways in which individuals, including political candidates,
can triangulate how they may fit into the perceptive framework of potential voters. Risks and
rewards are heavily influenced by their distinct RGIDs, producing corresponding incentive
structures for each RGID that is shaped by audience expectations. The varying expectations of
unconventional candidates introduce greater risks associated with such uncertainty, that
scholarship has also demonstrated to result in disproportionate punishment of the candidate
(Piston et al. 2018). Broadly speaking in this respect, appeals to negativity are generally
considered to be riskier than positive appeals given many voters find this strategy distasteful
(Geer 2007).
In general, the political actors whose tweets were analyzed in the present study used
negative associated words less than positive associated words, which may reflect a general
acknowledgement of the risks that are inherent to negativity. The inequitable risks of negativity
stacked against underrepresented elites does not entirely exclude these legislators from going

115
negative all together. Rather, the results entail a more judicious deployment of these emotions in
their online discourse.
Summary: Representative Negativity?
With anger the norm, political elites may try to further differentiate themselves and reap
the benefits of ‘going negative’ beyond anger by possibly going more negative with other discrete
negative emotions as suggested by my results. Variations in the use of overall negative sentiment
appeals and appeals to specific negative valence emotions—disgust and fear—were evident
along racial lines. BIPOC political actors, bound by social constraints informed by social
dominance theory and their demographically unconventional and subordinate status, may have
some awareness of the corresponding differences in how certain emotional appeals are received
by audiences (Phoenix 2020). Therefore, because of the risks of triggering detrimental outgroup
biases and electoral punishment (Piston 2010; Reeves 1997; Terkildsen 1993), within a given
discursive situation BIPOC MC’s, constrained with a lower risk ceiling informed by their RGID
positioning on the order of privilege, adopt an approach more judicious in their use negative
emotional appeals. Adopting a constrained approach to online rhetoric was not limited to
nonwhite racial minority elites as white women were revealed to adopt this strategy if certain
political party conditions are met.
The distinct behavior of white women, as presented in this study, serves as a vivid
example of partisan moderation driven by gender expectations both from the party and its
constituents. The Republican Party, leaning towards a more conservative ideology, bears
expectation for a woman to represent gender in a traditional sense that may view a woman
appealing to negativity poorly. Republican white women therefore constrained their approaches
relative to their copartisans and Democratic counterparts likely in service to embodying these

116
norms as possible. Conversely, Democrats, with a more diverse electorate in terms of race and
gender, may reward women and minorities who emphasize individuality, morality, and fairness,
thus cultivating an environment where said female elites may have the shared “greater piece of
mind” with the public. Democratic white women fulfilled these expectations as they engaged
more extensively in negative sentiment appeals, exhibiting levels of disgust comparable to their
white male party peers and higher levels of fear-based rhetoric. The results of these analyses
illuminate how these partisan expectations shape the behavior and strategic rhetoric of MCs with
minority RGID backgrounds.
The outcomes of my research provide salient insights into the variability in how elected
representatives employ appeals to negativity on Twitter, encompassing both governing and
campaign periods. Contrary to the often-overgeneralized assumptions prevalent in political
communication scholarship that categorize House members strictly based on party affiliation, my
findings discern a noticeable divergence grounded in intersecting racial and gender identities.
Such divergence imposes constraints on Members of Congress (MCs) with regard to the safe
deployment of negative emotive rhetoric on Twitter.
Furthermore, the campaign season emerged as a period consistently correlated with an
uptick in emotional rhetoric. This likely signifies strategic rhetorical shifts among political elites
in anticipation of an imminent election. In the context of this elevated electoral tension, the risk
environment could become increasingly stringent for political actors from minority racial and
gender backgrounds, which may lower their risk ceilings even further relative to white male
candidates.
Given the risks associated with going negative and people’s general aversions to feeling
negative emotions, a minoritized RGID actor’s capacity to appeal to negativity may be
conceptualized to be capped by a risk ceiling informed by their intersecting social identities. To

117
better understand how social identity dynamics furnish into the incentive structures that shape
political elites’ strategic rhetoric, these constraints were examined under the heightened electoral
duress of the campaign period. This period of time leading up to an upcoming election raises
public awareness of the state of politics, thereby raising the stakes for political actors. This would
reasonably elevate an individual’s discomfort and potentially their constraints while navigating
political spaces online and may produce competing incentive structures in the process. In the next
chapter I present my investigation of social identity constraints as they pertain to the 2020 election
candidates in both challengers and incumbents alike.

118
Chapter 5
Feeling the Heat: Electoral Threat, Institutional Positioning, and
Challenger Negativity
To further assess how race and gender influence distinct rhetorical strategies for
underrepresented candidates in response to perceived threat in the electoral environment, I
examine congressional communications on Twitter in the year leading up to the 2020 U.S. House
of Representatives elections (held on November 3, 2020). This analysis involved tweets from both
incumbent and non-incumbent challenger candidates’ campaign accounts. As members of
Congress expect public attentiveness towards politics to be highest during election years
(Macdonald 2020), this raises the stakes for what they say and how they say it. The heightened
scrutiny during election years transforms the campaign season into a period of amplified electoral
risk. While the impetus to go negative is high for many candidates to mobilize votes, perhaps
among the strongest motivators to go negative is a challenger candidate hoping to overcome
incumbency advantage.
To fully grasp variations in campaign rhetoric, it is crucial to consider how social identity
effects fit with established theories on incumbent and challenger strategic communication. Per
my proposed theoretical framework, underrepresented RGIDs are subject to a risk ceiling whose

119
height is determined by their relative social positioning and acts as a constraint on their appeals
to negativity. This dynamic, in essence, relies on how polarizing a candidate is willing to become,
which is limited by the perceived degree of risk appraised with their public facing actions.
Therefore, the pressurized atmosphere of an impending election would expectedly apply
electoral duress onto all candidates. Increased electoral threat, in turn, may shift the risk equation
a candidate calculates that heightens the risk of each public discursive move given the heightened
attention the public gives to races in campaign years. Conceptually speaking, a candidate’s RGID
risk ceiling may hold constant but a faster rate of risk accrued with each move limits the latitude
afforded to an underrepresented RGID candidate to adopt an expressive approach.
The benefits of incumbency are well-documented, stemming from attributes such as voter
familiarity and demonstrated competency in office, which candidates can leverage to showcase
the value they bring to their districts (e.g., Fiorina 1989; Ferejohn 1977; Gronke 2000; Jacobson
2004; Lazarus and Reilly 2010). Incumbents, due to their advantageous positioning within the
political institution and generally reduced electoral vulnerability, lessen the need for voters to
rely on categorical cues like gender or race, thereby mitigating potential voter bias (Druckman,
Kifer, and Parkin 2009; Fiorina 1989; Fowler 2016; Fowler and Hall 2014; Gronke 2000; Jacobson
2013; Mann and Wolfinger 1980). This mitigated reliance on cues subsequently discourages
incumbents from employing attention-seeking rhetoric, such as negative appeals.
Conversely, challenger candidates face an information asymmetry that leads voters to rely
more heavily on simplistic cues—such as gender or race—for assessment (McDermott 1998). This
voter behavior is substantiated by existing literature; for instance, Kahn (1996) finds that voters
often resort to gender stereotypes, which are typically more robust, when evaluating challengers
due to the limited available information (507). To counteract this informational disadvantage,
challengers may employ negative emotional appeals to shift voter focus on the incumbents' track

120
record (Druckman, Kifer, and Parkin 2009). Some scholars have also attributed the use of
negativity to candidates' perceived vulnerability (Gainous and Wagner 2014; Lassen and Brown
2011; Evans, Cordova, and Dipole 2014), which may be due to factors such as limited campaign
resources, running as a non-incumbent challenger, or in an open seat race (Lau and Pomper 2002).
The institutional dynamics delineating incumbents and challengers have been extensively
scrutinized in existing scholarly discourse, with the onus consistently placed on challenger
candidates to escalate public awareness of a race through potent, often negative, rhetorical
strategies. Within this competitive interplay, the unyielding strength of incumbency advantage
has been long acknowledged, reinforcing the widespread perception that challengers must
engender public focus and provoke incumbents into dialogical platforms where they are subject
to voter reevaluation as a viable strategy for overcoming this formidable obstacle.
Yet, this body of research conspicuously underemphasizes the integral role of social
positioning, particularly candidate RGID, in shaping a candidate's inclination and willingness to
appeal to negativity online. While perhaps understandable, given the historically monolithic
composition of elected bodies—most notably a predominantly white, male House of
Representatives—an intersectional analytical lens could yield markedly divergent outcomes,
thereby challenging entrenched assumptions about the utility of negative emotive rhetoric in
campaign settings.
As my findings on appeals to negativity have elucidated, the choice to "go negative" is a
more complex affair for candidates from underrepresented racial and gender groups.
Accordingly, it stands to reason that as these candidates' perceived electoral risks escalate, I
expect that they would, being rational actors, adopt an increasingly constrained communicative
approach. Given that appeals to negative sentiment represent a higher-risk strategy particularly
for candidates from historically marginalized racial and gender backgrounds (Piston et al. 2018;

121
Dittmar 2015; Averill 1982; Marcus et al. 2000; Phoenix 2020; Bauer 2017; Krupnikov and Bauer
2014), these constraints can be empirically assessed.
Therefore, I hypothesize that candidates who do not have dominant or privileged social
positioning will adopt a constrained rhetorical approach characterized by a lower reliance on
appeals to negative sentiment and specific negative emotions like anger, disgust, and fear, when
compared to theirrespective white male challenger or incumbent counterparts. Specifically, a risk
ceiling is formed for these emotions as a limit to the degree and breadth to which they appeal to
negative emotions. These constraints, which are rooted in societal cues based on race and gender,
manifest even before these individuals embark on their political careers. As a result, this dynamic
is expected to influence both challenger and incumbent candidates alike in their use of negativity.
In this chapter, I test if rhetorical constraints, associated with disparate, subordinate social
identities, persist in the campaign period, whose electoral goals greatly depend on going negative
by those who lack institutional advantage. Like chapter 4, several OLS regression models are
performed to determine the impact of intersecting racial and gender social identities on candidate
capacity and amenability to go negative.
2020 Incumbent and Challenger Campaign Tweets
To examine the use of negative overall sentiment and negative discrete emotions (i.e.,
anger, disgust, fear) in candidates’ rhetorical strategies across different electoral contexts with
varying degrees of competitiveness and perceived threat, I leverage a second collection of
congressional tweets from representatives in the year leading up to the 2020 election (January 1,
2020, to November 3, 2020), but now with the inclusion of non-incumbent, or challenger,
candidates as well. Covariates that potentially influence strategic rhetoric in a campaign setting,

122
such as district party fit and competitiveness of race, were included into the models and
controlled for. The cohort of 2020 candidates were additionally stratified based on incumbency
vs. challenger status to better define the effects stemming from institutional advantage vs.
disadvantage. These stratified groups were then evaluated as subsets to elucidate if RGID impact
behavior of challengers or incumbents specifically.
Using multivariate regression models specified with RGID (white woman, BIPOC man,
BIPOC woman) as factor variables that are denoted with ‘(fct)’ in the model, I estimate
underrepresented candidates’ monthly average use of negative sentiment and the discrete
emotions of anger, disgust, and fear in their 2020 campaign tweets relative to the baseline RGID
set to white male candidates. The data in Table 15 shows how each underrepresented RGID
differs from white male candidates in the patterns of appeals to negative sentiment and discrete
negative emotions that they adopt. Incumbency status was also included as a covariate in the
model to evaluate for significant differences against challengers with respect to these emotional
appeals.

123
Table 15. OLS regressions of all 2020 House candidates’ monthly average use of negative
valenced emotions on Twitter (RGID as factor levels).
Comparing tweets from MCs from underrepresented RGIDs directly to the white male
2020 candidates produced substantial differences in negative rhetorical choices between all
subordinate RGIDs and the dominant, white male identity. Notably, and expectedly, white male
candidates relied on negative rhetoric to a greater extent than candidates from other RGIDs (p<

124
.01), which is in line with theoretical expectations for the dominant group to be the more
expressive group and therefore be the most accommodating for higher-risk strategies (Table 15).
Also in line with expectations was the lower propensity to appeal to negativity by
incumbents (p< .01) across all negative emotive appeals in the analysis (Table 15). These
coefficients highlight the decreased use of negativity in their Twitter communications of
incumbents in relation to challengers and is consistent with existing literature describing this
institutional advantage (Druckman, Kifer, and Parkin 2010; Druckman and McDermott 2008;
Marcus, Neuman, and MacKuen 2000) that underscore incumbents’ incentive to avoid discursive
negativity as to keep attention away from their respective races.
Challenging Expectations in Challenger Rhetoric
To further explore the implications of institutional (incumbency) positioning on the
rhetorical strategies used by candidates with varying social positioning, I subset the 2020 data by
incumbency status and run separate multivariate regression models of each set of candidate’s use
of negativity and negative discrete emotions (i.e., anger, disgust, fear) in their campaign tweets.
The distribution of challengers and incumbents separated by party and sorted into respective
RGIDs is presented in Table 16. Extant literature detailing incumbent and challenger relationships
suggest preferential deployment of higher risk negative emotive appeals by institutionally
disadvantaged candidates, such as challengers (Lassen and Brown 2011; Evans, Cordova, and
Dipole 2014; Gainous and Wagner 2014; Gervais et al 2019). The Prospect Theory has been
proposed as the behavioral model behind challenger candidates’ strategic rationale that explains
their adoption of higher risk negativity as a desperate attempt to close the gap of incumbency
advantage (Druckman, Kifer, and Parkin 2009, McDermott, Fowler, and Smirnov 2008). By
drawing more attention to their race along with more incumbent scrutiny, negativity is an

125
instrument for challengers to establish themselves in public discourse and influence the scope of
the debate.
My findings corroborate previous studies’ observations of heightened challenger
negativity relative to incumbent negativity with statistically significant (p< .01) coefficients
indicating incumbents’ lower propensity to appeal to overall negative sentiment and the discrete
emotions of anger, disgust, and fear relative to their challenger counterparts (Table 15). However,
as Congress becomes more diverse from a race and gender standpoint, social identities are
becoming more salient in explaining candidate behaviors of both incumbents and challengers.
Would the rhetorical constraints seen in the historically minoritized RGID representatives in the
113th to 116th Congresses also apply for challengers and incumbents? As challengers are known
for their starkly negative rhetoric, would challengers from disparate social identities be
incentivized by their institutional disadvantage to go negative in a similar manner, or would
constraints from their subordinate identities take precedence?

126
Table 16. Total number of unique tweets, month observations, accounts, and candidates by
incumbency status, race/gender, and party.
Status Race/Gender Party Total Tweets Obs. Accounts Cands
Challenger BIPOC Man Democratic 94142 825 92 92
Incumbent BIPOC Man Democratic 66576 782 86 56
Challenger BIPOC Man Republican 40486 558 70 70
Incumbent BIPOC Man Republican 3217 66 7 5
Challenger BIPOC Woman Democratic 77559 723 76 76
Incumbent BIPOC Woman Democratic 57911 582 62 42
Challenger BIPOC Woman Republican 53080 397 44 44
Incumbent BIPOC Woman Republican 1676 29 3 2
Challenger White Man Democratic 189095 1318 153 152
Incumbent White Man Democratic 93698 1275 135 87
Challenger White Man Republican 166287 2154 262 260
Incumbent White Man Republican 126822 2319 269 175
Challenger White Woman Democratic 173698 1361 145 145
Incumbent White Woman Democratic 54408 623 65 47
Challenger White Woman Republican 49332 657 77 77
Incumbent White Woman Republican 15366 205 22 13
Challenging With and Without Privilege
To test if social identity impacts challenger negativity, I performed multivariate
regressions on the subset cohort of challengers from the 2020 campaign tweets. The first model
compared white women, BIPOC women, and BIPOC men against the white men in their average
monthly tweets scores for challengers (Table 17). These results are in Table 17, with ‘(fct)’ label
adjacent to each race/gender variable name and correspond to analyses comparing different
minoritized RGIDs within the challenger cohort against a white male challenger. The analysis

127
yielded significant negative coefficients (p< .01) for underrepresented subgroups in terms of
overall negative sentiment and specific emotions like disgust and fear. Women and BIPOC male
challenger candidates appealed to overall negativity and disgust less than white male
challengers. Additionally, female candidates of all races appealed to fear less than their white
male counterparts, while BIPOC male challengers appealed to fear in similar capacities to the
control challenger group. These findings align with Druckman et al. (2009), in that white male
challengers significantly increased their use of negative emotive rhetoric. However, with the
introduction of intersectional social identity strata within the dataset, I find my data to
correspond with my expectations with a clear gap in the use of negativity between challenger
candidates that is suggestive of significant constraints to negativity expressed by challengers with
minoritized social identities relative to white male challengers.

128
Table 17. OLS regressions of 2020 House challenger candidates’ monthly average use of
negative valenced emotions on Twitter (RGID as factor levels).
Challengers Analysis - Democrat vs. Republican Models with RGID and Party
Interaction Term
To evaluate how each RGID of challengers compares with the broader cohort of 2020
challengers, individual regression models were employed, treating each RGID as a discrete
binary variable. The resulting data is presented in Table 18. This analysis replicates the
methodology used in the study of negativity across the 113th to 116th Congresses outlined in

129
chapter 4, incorporating an interaction term ([RGID]*Republican) to assess any differential impact
of party affiliation—Republican versus Democratic—on the proclivity for negative campaigning
within specific RGIDs of 2020 challengers.
Statistically significant (p< .01) coefficients in this model highlight a pronounced
divergence in the appeals made by white male challengers relative to other 2020 challengers.
Specifically, white male challengers were 5.1% more likely to resort to overall negativity in their
tweets than the rest of the challengers. These appeals involved contributions from appeals to
discrete emotions of disgust and fear, with white males employing such rhetorical tactics 2.9%
and 3.4% more frequently than other challengers, respectively (Table 18).
Conversely, challengers who identified as either white woman or BIPOC woman were the
RGIDs associated with the lowest reliance on appeals to overall negativity and fear among the
challenger candidates. White women further demonstrated a statistically significant coefficient
indicative of the lower propensity of white women challenger candidates to appeal to the discrete
negative emotion of disgust relative to their challenger peers. Specifically, these significant
regression coefficients translate to reduced propensities relative to the challenger cohort by white
women challengers to employ overall negativity, disgust, and fear in their campaigns by 2.5%,
2.1%, and 1.9% respectively. For BIPOC women, the propensity to engage in appeals to overall
negativity and fear was reduced by 3.1% and 3% respectively (Table 18).
BIPOC male challengers did not exhibit any statistically significant regression coefficients,
suggesting their appeals to the studied negative emotions largely aligned with the averages of
the entire 2020 challenger cohort.
To investigate if political-party affiliation impacts challenger candidate rhetoric, the
interaction term ‘[RGID]*Republican’ was included in the model. Statistically significant
regression coefficients were indicative of differences between Republican affiliated challengers to

130
appeal to the corresponding emotive appeal relative to Democratic MCs with matching RGID
demographics.
Statistically significant interaction terms in Table 18 were exclusive to white men, whose
coefficients in overall negativity (-2%) and fear (-2.8%) point to a decreased propensity for these
emotional appeals among white male Republican challengers compared to their white male
Democratic counterparts. This distinction may reflect the effect of political party alignment with
the sitting president at the time of the 2020 campaign.26
The examination of the challenger cohort in my study reveals discernible variations in
their utilization of negative emotional appeals in online spaces. Contrary to the conventional
assumptions that suggest a uniform approach among challengers, the data demonstrates notable
exceptions. Specifically, the uptick in group negativity among challengers is predominantly
propelled by white men, who constitute the dominant RGID atop the social hierarchy.
Additionally, when party affiliation is accounted for, the differences between challengers in their
recourse to negative emotional appeals appear to be largely inconsequential.
Challengers represent a disadvantaged group given their outsider status with respect to
the political positioning and are incentivized to go negative to attract attention and scrutiny of
their race and the incumbent. The results of my analysis contribute to the growing body of
research on congressional online rhetoric by showing a variance amongst subordinate RGIDs and
their capacity for negativity appeals. Given the risks associated with going negative and people’s
general aversions to feeling negative emotions, a candidate’s willingness to go negative is honed
by balancing the potential rewards with their burden of risk. Thus, to capture a fuller
understanding of these various competing forces with the ultimate goal of election, an
26 The sitting president at the time of the 2020 campaign was Donald J. Trump, Republican.

131
examination of incumbent candidates in the 2020 election is the logical next analytic step and will
be discussed in the next section.
Table 18. Model estimates of RGID*Republican effects from separate regressions of 2020
challenger candidates’ monthly average use of negative valenced rhetoric.

132
Incumbents Analyses
Incumbents, as expected, used less negativity relative to their challenger opponents. This
difference is representative of their advantage as an incumbent and their incentive to maintain
such a gap between challengers. Perpetuation of this advantage entails mitigation of constituent
attention onto their race, with a messaging goal to maintain the status quo at the polls. While
challenger candidates have been established in scholarship as the candidate most incentivized to
adopt negative emotive rhetoric, Incumbents, though incentivized to minimize attention
grabbing rhetoric, must make negative emotive appeals when responding to challengers
(Druckman, Kifer, and Parkin 2009) my findings presented in chapter 4 notes a consistent
statistically significant association with the campaign period and an increase in emotive appeals
that include negativity appeals as House representatives transition into incumbent candidates
during a campaign. In the evaluation of the incumbent cohort, similar analytical techniques were
employed as those used for the challenger cohort. A multifactor RGID approach was applied in
which the RGIDs (indicated by ‘(fct)’ nomenclature) were compared to a white male incumbent
control. The results of this analysis are presented in Table 19. This analysis revealed statistically
significant regression coefficients for white women and BIPOC men in comparison to white male
incumbents, pointing to their comparatively restrained use of negative emotional appeals relative
to the incumbent cohort. Such moderation may indicate ongoing constraints experienced by these
incumbent candidates from these minoritized RGIDs. Intriguingly, BIPOC women incumbents
did not manifest statistically significant differences in their rhetorical strategies compared to
white male incumbents. This raises the possibility that the advantage associated with incumbency
may embolden BIPOC women incumbents to resort to negative appeals at rates akin to white
male incumbents.

133
Table 19. OLS regressions of 2020 House incumbent candidates’ monthly average use of
negative valenced emotions on Twitter (RGID as factor levels).
Evaluation of each RGID treated as a binary variable was also performed for incumbents
to better assess how incumbent negative rhetoric during a campaign may vary across the different
RGIDs within the 2020 incumbent cohort with corresponding results presented in Table 20. White
men incumbents were the only incumbent candidates with statistically significant (p< .05)
increases in their appeals to disgust relative to the cohort of incumbents. Incumbent BIPOC men
appealed to negativity more sparingly than the other incumbent candidates with statistically
significant regression coefficients associated with appeals to overall negative sentiment (-2.4%),

134
disgust (-2%), and fear (-2.8%). BIPOC women and white women did not have significant
regression coefficients in this model, meaning female incumbents represent the group average of
the incumbent cohort.
Table 20 also incorporated an interaction term, ‘[RGID]*Republican’, to evaluate partisan
variations within specific race/gender subsets. A notable partisan deviation emerged with white
female incumbents, who consistently displayed negative coefficients (p< .01) across overall
negative sentiment and the array of analyzed discrete negative emotions (anger, disgust, fear).
This divergence in rhetorical strategy across partisan lines echoes patterns previously noted in
analyses of the 113th to 116th Congresses, suggesting that the observed rhetorical variations
within the incumbent cohort may be attributed to partisan expectations shaping distinct incentive
structures for different RGIDs.

135
Table 20. Model estimates of RGID*Republican effects from OLS regressions of 2020
incumbents’ monthly logged average use of negative valence rhetoric in office tweets.

136
As challengers are relatively unknown in comparison to incumbents, they must build up
their ‘brands’ to get the vote. While white male challengers may use negative rhetoric as a means
to draw attention and awareness (Druckman et al 2009), this tactic is likely ineffective for women
candidates unless they become linked to a party whose ideologies encourage them to do so (i.e.,
Democratic Party). It seems constituents do not draw the association between a candidate and
their political party until said candidate is actually elected, given the dramatic shift in form from
a more undifferentiated challenger to a seasoned incumbent. Therefore, establishing oneself as a
white woman challenger entails a risk environment of volatile constituent expectations and high
political vulnerability. While the difficulty of predicting audience expectations may abate as
Congress continues to diversify, within the scope of this study and authoring of this dissertation
the safer approach for unconventional challengers likely remains constrained.
To understand the divergent rhetorical choices of incumbents and challengers more
comprehensively, especially within the context of various RGIDs, grouped point plots were
employed. These visualizations serve to consolidate the data on the use of negative sentiment
appeals as well as targeted emotional triggers like anger, disgust, and fear.
Figure 12, Figure 13, Figure 14, and Figure 15 illustrate the findings previously outlined
in Table 18 and Table 20 by presenting plotted coefficients as they pertain to Democratic
candidates’ appeals to overall negative sentiment and to the discrete negative emotions of anger,
disgust, and fear, respectively. These coefficients estimate the relative impact of each specific
race/gender identity group on the emotive appeals deployed within a party’s set of campaign
tweets by both incumbent and challenger candidates. Importantly, only the coefficients relating
to the binary race/gender indicators are reported and plotted, providing a focused view of the
individual effects of each RGID category (comprising eight total models).

137
Overall evaluation of the Democratic candidates in the 2020 campaign found distinct
patterns within the different RGIDs of challengers and incumbents. Challengers of all races and
genders aligned with the expectation of increased reliance on negativity by challengers over
incumbents, however my data further contextualizes this in that white-male challengers appeal
to negativity the most amongst challengers by a substantial margin. Furthermore, female
challengers appeal to overall negativity at a similar rate to female incumbents of the same race,
only significantly appealing more to disgust amongst both white women and BIPOC women
(Table 21). This also aligns with expected challenger negativity but shows a more abated
divergence between challenger and incumbent – suggesting a race and gender dependent
constraint for challengers or expressivity for incumbents.
It is important to note that due to the underrepresentation of racial and gender diversity
among Republican candidates in the 2020 election, they were intentionally omitted from the
models shown in the figure. Additionally, the models generate relative estimates based on each
corresponding in-group average (i.e., challenger or incumbent). Therefore, the graphs depicting
challengers and incumbents are not directly comparable. Nonetheless, the contours of the graph
accurately reflect the nuanced variations in emotive rhetoric within each in-group.

138
Figure 12. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of negative appeals.

139
Figure 13. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of angry appeals.

140
Figure 14. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of disgust appeals.

141
Figure 15. Model estimates (w/ 95% CIs) of separate race/gender effects on 2020 Democratic
challengers’ and incumbents’ use of fear appeals.

142
Who Runs Lowest? Challengers, Incumbents, and Social Identity
Navigating the intricacies of the challenger-incumbent dynamic presents a complex
landscape for the application of the analytical protocol delineated in the preceding chapter.
Specifically, the embedded incentive structures encourage disadvantaged challenger candidates
to adopt negative strategies while simultaneously disincentivizing incumbents from pursuing
similar tactics (Druckman, Kifer, and Parkin 2010).
To investigate whether this relationship is consistent across all race/gender subgroups, a
targeted analysis was conducted using the Democratic candidate subset, further stratified by race
and gender. Serial regression models estimated relative use of negative sentiment, along with
discrete emotions such as anger, disgust, and fear as it pertains to challengers and incumbents
within a given RGID. Table 21 summarizes these results. The presented regression coefficients in
this table are the values derived from each individual model run using the subset Democratic
race/gender cohort.
Generally, challengers within a given race/gender category exhibited a higher usage of
negativity-associated terms compared to their incumbent counterparts of the same RGID (Table
21). BIPOC men had statistically significant regression coefficients across all negative sentiment
and discrete negative emotions that indicates BIPOC male challenger candidates outpaced BIPOC
male incumbents in every negative emotive category within this analysis. White male incumbents
were outpaced by white male challenger candidates in overall negative sentiment, disgust, and
fear but used anger-associated words to similar extents. Female candidates of both BIPOC and
white backgrounds demonstrated statistically significant coefficients solely in appeals to disgust,
indicating challenger women appeal to disgust 0.5% more than incumbent women. Beyond this
discrepancy at disgust, female challengers and incumbents appealed to the other negative
emotive parameters at similar rates.

143
Table 21. Effect of incumbency status on 2020 House candidates’ monthly average use of
negative sentiment, anger, disgust, and fear appeals by RGID.
The findings presented in Table 21 substantiate previous research on the long-recognized
incumbency advantage (Druckman, Kifer, and Parkin 2010), underscoring challengers' incentives
to “go negative" as a strategy to attract attention and prompt reevaluation of incumbents by

144
voters. My analysis of the 2020 candidates corroborates this incumbent-challenger dynamic and
broadens its applicability across various race and gender intersections. Granted, this tendency is
far from uniform across disparate RGIDs and comes with a variety of implications. Notably
evident is the widened gap between male challengers and incumbents, in contrast to the narrower
discrepancy among woman candidates.
This gendered discrepancy can be conceptualized by the separate influences of
institutional positioning and social identity. The asymmetry in institutional positioning between
incumbents and challengers establishes a set of incentives that spurs challengers to employ
negative appeals in a bid to stimulate voter engagement. Alternatively, a candidate’s RGID, as
informed by my findings in the previous chapter, would constrain subordinately positioned
candidates in their use of negativity by virtue of an uncertain risk environment for
unconventional candidates.
The conspicuous variance in negative appeals between male challengers and incumbents
implies that the incentive structures for male candidates are predominantly shaped by
institutional dynamics. In this regard, both white and BIPOC male challengers seem less inhibited
by their respective social identities, thus displaying a marked tendency to engage in negative
tactics relative to their incumbent counterparts.
Conversely, the incentive structures of woman candidates are likely driven by a balance
of social identity and institutional identity effects. The data suggest that female challenger
candidates are incentivized to appeal to negativity, thus manifesting the difference in disgust
appeals between challengers and incumbents. Given that this represents the lone differential
between female challengers and incumbents, it likely signifies an inherent constraint imposed by
their gender identity that carries more weight in their decision in this context than what is seen
in men. This limitation on female challengers' engagement with negativity is corroborated by the

145
significantly lower frequency of negative appeals noted in the binary RGID variable regression
of challenger candidates examined earlier in this chapter (see Table 18). In this instance, the
gender identity of woman challenger candidates appears to act as a limiting factor on the extent
to which they can resort to negativity.
By evaluating differences in negativity between challengers and incumbents, my results
support a less homogeneous landscape of negative rhetoric online than what extant literature
may previously suggest. This rhetorical terrain is shaped by intersecting social identities, as well
as contexts of institutional positioning which are instrumental in molding incentive structures.
This places emphasis on the necessity of adopting thorough approaches examining various axes
of effects and the interplay between these axes as opposed to previous monolithic notions of
political communication largely focusing on a single race/gender type of candidate.
Claiming a Seat, Finding a Voice: Rhetorical Shifts After Electoral Victory
“Sexism isn't new to Congress or our country. The verbal assault wielded against my
colleague was disgusting, inappropriate, and endemic of the toxic masculinity that
permeates our culture. I stand with Rep. @AOC. This is our House. We are claiming our
space. So, get used to it.”
- @repkclark, 2020-07-23
Representative Katherine Clark (MA-5)
The results of the present study illustrate a notable difference in the rhetoric adopted by
candidates outside of the political institution as opposed to incumbents who have established
themselves within their respective parties. Existing research provides evidence that elected
officials' rhetorical strategies evolve after they assume office, allowing them greater flexibility to
deviate from party lines and construct a more personal and nationally resonant brand (Towle and
Collier 2010; Amsalem 2019). This trend is particularly evident among representatives who

146
leverage social media platforms such as Twitter to maximize their political influence, even as
freshman representatives traditionally not afforded such a platform. This shift, in part, is fueled
by the need for these politicians to develop a brand identity that transcends partisan divides and
resonates with a broader spectrum of voters. It also offers an opportunity to cultivate a more
prominent national profile, often a prerequisite for future political ambitions. In the analyses
discussed in the previous sections, I demonstrate such a paradigm shift that is particularly notable
amongst congresswomen and candidates such that the social constraints placed upon
underrepresented race/gender individuals are particularly potent before the candidate has
established themselves as a member of the institution (i.e., winning a house seat).
As challengers, white women show a level of negativity in their tweets that is largely
indistinguishable from other challengers of underrepresented RGIDs or even those of the
opposing party. This suggests a homogenizing effect, possibly rooted in the shared commitment
among challengers to employ negativity as a rhetorical tool. However, once elected, these women
exhibit a discernible shift in the tenor of their emotive rhetoric, a pattern that bifurcates along
party lines. Specifically, Democratic white women incumbents display an uptick in negativity in
their tweets, while their Republican counterparts curb their reliance on such rhetorical strategies.
This trend mirrors the party-driven divergence observed in this RGID group in the analysis of
the 113th to 116th congresses from chapter 4, providing a degree of longitudinal consistency.
To clarify, this shift in negativity should not be seen as an isolated change in emotional
rhetoric; rather, it likely signifies a broader strategic transformation in communication that
extends beyond emotive appeals. This change potentially involves adopting a less confrontational
tone more suitable for governance and legislative processes.
Furthermore, the consistent divergence in negativity between parties underscores the
notion that these shifts in rhetorical style may be fulfilling partisan expectations. It serves as an

147
indicator that, although candidates may run under a specific party banner, their public
communications do not fully align with party elite expectations until after they secure electoral
victory. This suggests that the challenger phase serves as a platform for new political hopefuls to
gain visibility and introduce themselves to potential voters, while the incumbent phase is when
these actors’ partisan identities become major determinants in discursive strategies.
Her Disgust, His Disgust: Negativity’s Qualitative Gap
See the examples of disgust tweets provided in Table 22, note how tweets from
Democratic white women and Democratic BIPOC women imbue disgust into their rhetoric while
also discussing a salient issue. In a similar way to anger, it seems that at least in this subset of
women, framing disgust within issue discourse is acceptable as a way to signal their partisan
ideology (Table 22). Looking at the disgust appeals made by Democratic women also shows the
diversity of appeals in terms of issues and approach. It seems that making such a variety of
appeals with disgust may reflect the constraints Democratic women shed by getting elected to
the political institution, thereby allowing them to show more of themselves to their constituents.
However, applying such a qualitative approach towards the candidate tweets also shows notable
differences between tweets from women incumbents vs. white men.
Despite this closure of the quantitative gap between negativity appeals by women vs. their
white male counterparts, there appears to still be a qualitative gap between the natures of these
tweets from the white men vs. women. As these tweets from women have generally shown, many
of them use disgust as a supplement towards the larger goal of presenting their stance on issues.
Note how Jackie Speier’s higher scoring disgust tweet is used to help frame her stance on an issue.
Compare this to the challenger Madison Cawthorn, who scored lower for disgust associated
words in his tweet than Rep. Speier but features the appeal to disgust as the primary focus of his

148
tweet, which could stand to argue that these more direct appeals to an emotion may have a more
potent effect. This qualitative discrepancy suggests that while women have been matching their
white male incumbent counterparts in the use of negative emotional appeals, there is still a
distinction between how these emotions are applied between white men and women. Future
research perhaps focused on the moralizing effect of tweets could be a helpful step to resolve the
quantitative and qualitative differences that have been demonstrated.
Considering the year of the woman and in tandem with Rep. Clark’s tweet, the
diversification of Congress brings in a growing census of new voices and perspectives into the
political sphere who inherently face differing risk terrains and constraints, thereby producing
distinct communication behaviors based on their presented social identities, partisan identity,
and position within the political power structure.

Table 22. Examples of disgust-associated rhetoric in 2020 House candidate tweets.
Member Party Text Score
BIPOC
Women
LEE, Barbara
(CA-13)
Democratic “Dr. King named three evils in America: the evil of racism,
the evil of poverty, and the evil of militarism. Racism and
white supremacy still plague our nation, too many
Americans remain trapped in poverty, and our country is
caught in a cycle of violence and endless war.”
Disgust: 6
OMAR, Ilhan
(MN-5)
Democratic “We live in a sick, sick nation - where the contagion of
greed and the epidemic of Wall Street politicians is
forcing Flint, Michigan residents to choose between going
into debt for poisoned water or die of #coronavirus”
Disgust: 6
White
Women
CLARK,
Katherine
(MA-5)
Democratic “Sexism isn't new to Congress or our country. The verbal
assault wielded against my colleague was disgusting,
inappropriate, and endemic of the toxic masculinity that
permeates our culture. I stand with Rep. @AOC. This is
our House. We are claiming our space. So, get used to it.”
Disgust: 5
MALONEY,
Carolyn
(NY-12)
Democratic “Rush Limbaugh's most infamous act was calling a young
grad student a slut for testifying in favor of contraceptive
coverage. A terrible diagnosis does not erase a lifetime of
racism, homophobia and misogyny. A sick man deserves
healthcare; a bigot does not a medal of honor. #SOTU”
Disgust: 5
MALONEY,
Carolyn
(NY-12)
Democratic “Asian-Americans have been viciously targeted and
subjected to violent attacks, discrimination, and
intolerance during the #COVID19 crisis. Racist violence is
NEVER okay and will not be tolerated in NYC. This
disgusting persecution must stop.”
Disgust: 5
SCHAKOWS
KY, JAN
(IL-9)
Democratic “Ignoring the mass death and suffering while lying to the
people, is monstrous in my view. Donald Trump has been
attacking Black women and women of color his whole
political career. His gross misogynoir will not stand.
We’re with @KamalaHarris. #Vote #BidenHarris2020”
Disgust: 5
SPEIER,
Jackie
(CA-14)
Democratic “While @POTUS’s shameless corruption is on full display
over his resistance to meaningful election reform, my
colleagues and I are fighting for no-excuse vote by mail
and $ so that Americans don’t have to risk their lives to
vote in an election under attack from malign foreign
actors!”
Disgust: 5
White
Man
CAWTHORN,
Madison
(NC-11)
Republican “Violent Domestic Terrorists are burning down buildings
and ending innocent lives Sadly, many spineless
Politicians (including my opponent) refuse to denounce
this terrorism. Help elect an Outsider with a Titanium
Spine”
Disgust: 5
CAWTHORN,
Madison
(NC-11)
Republican “Unlike my opponent, I will not tolerate career politicians
dictating our lives from Washington. I swear to you, until
my dying breath, I am your representative and beholden
solely to the people of this district. Let me be your weapon
in Washington”
Disgust: 5
149

150
Chapter 5 Summary
In this chapter I examined the negative rhetoric candidates of the 2020 election use in their
online discourse on Twitter. Building upon the conceptual framework articulated and empirically
supported in the preceding chapter, the results from the present analysis substantiates existing
theories about the role of institutional dynamics in shaping the behavior of challengers and
incumbents. While challenger candidates across all RGIDs were shown to be more expressive in
their strategic approach, opting to rely more frequently on negative emotional appeals than
incumbents, I find that this trend does not necessarily manifest uniformly across all challenger
candidates. Importantly, the results uncover significant differences between candidates from
underrepresented backgrounds and white male candidates, even when controlling for partisan
differences. Specifically, challenger candidates relied more heavily on negative emotional appeals
compared to incumbents across all RGIDs. This deviation challenges conventional wisdom that
often presumes behavioral homogeneity among challengers, lending credence to the nuanced
ways in which the impending election might differentially shape a candidate's incentive
structures and, consequently, their chosen rhetorical strategies.
Extant literature has correlated the propensity to 'go negative' as tied to Prospect theory
(Druckman, Kifer, and Parkin 2009; McDermott, Fowler, and Smirnov 2008), that explains why
an individual in a disadvantaged position would take on negativity out of desperation. However,
given that the strategy to go negative is generally seen in poor taste by the public (Geer 2006),
going negative is often labeled as the riskier emotional appeal option given the general dislike of
voters to occupy this emotional space for too long. This duality of intentions behind going
negative thus complicates analysis during the campaign period but for a given occurrence of
negativity this relationship may be reduced to two main drivers: 1) subordinate RGIDs that

151
constrains negativity and establishes a risk ceiling, or 2) institutional position disadvantage
(challenger) that incentivizes negativity.
Challenger negativity exceeded incumbent negativity in all RGIDs but was more
prominent amongst male candidates compared to woman candidates. This suggests that the
institutional position dynamics separating challengers and incumbents is the primary driver
behind male candidates’ decisions to go negative during campaigns. Among male candidates, the
divergence in negativity between challengers and incumbents was most pronounced. Both
BIPOC and white male incumbents were markedly less negative in their appeals compared to
their challenger counterparts across races and genders. This suggests that the institutional
position dynamics associated with being a challenger are likely the key drivers informing male
challengers’ rhetorical decisions.
Conversely, woman candidates, both incumbents and challengers, demonstrated a more
similar use of negative appeals to one another with the exception of disgust appeals. This suggests
a risk appraisal and strategic decision-making process that strikes a closer balance between the
institutional disadvantage of a challenger that would propel negativity vis-à-vis the social
constraints of a subordinate socially positioned candidate whose unconventional race/gender
identity would apply greater constraints against negativity.
Additionally, my data underpins the well-documented advantage incumbents possess
(Alford and Hibbing 1981; Ansolabehere, Brady, and Fiorina 1992; Cox and Katz 1996; Erikson
1971; Ferejohn 1977; Fiorina 1977; Jacobson 1987; King and Gelman 1991; Krehbiel and Wright
1983; and Mayhew 1974), evidencing a near-uniform aversion to deploying negative sentiment
appeals across different race and gender RGIDs. House representatives’ comparatively prudent
use of negativity vs. their challenger counterparts that was demonstrated in my analysis aligns
coherently with the conventional wisdom that incumbents prefer to minimize undue attention to

152
the electoral race, thus strategically avoiding the high-stakes arena of negative campaigning.
However, this does not mean that incumbents are completely devoid of negative emotional
rhetoric in their public discourse. As the analysis in chapter 4 demonstrated, representatives
consistently increased their negative emotional appeals on Twitter when they become incumbent
candidates during campaign years.
Moreover, the data also reveals that social identity exerts a discernible influence on the
variation in the use of negative appeals among challengers. White male challengers appear to set
a sort of upper limit in the study concerning the employment of negativity on Twitter. Given their
societal status as the dominant RGID, their boldness in leveraging negative appeals is consistent
with what would be expected, given they have the most to gain with relatively less risk.
In addition to the patterns observed among challengers, incumbents also exhibited
distinct patterns in their use of negative rhetoric, varying across different RGIDs. Remarkably,
BIPOC male incumbents were the only subset to significantly diverge from the incumbent
average in terms of negativity, while female incumbents employed levels of negativity
comparable to their dominant white male counterparts. This uniformity in incumbent behavior
across RGIDs supports the notion that incumbents generally shy away from attention-seeking
negativity, thereby reinforcing their institutional advantage.
In examining the interaction between party affiliation and RGID, the study unearthed
further nuances. Republican white male incumbents were more inclined to utilize negative
sentiment, as well as specific appeals to anger and fear, compared to their Democratic
counterparts. Moreover, the results corroborate earlier findings from the analysis of the 113th to
116th congresses, illustrating that white women across parties differ significantly in their use of
negative appeals; Democratic white women showed greater propensity for negative sentiment

153
and specific negative emotions, thereby establishing clear demarcation lines between themselves
and their Republican counterparts.
The results of my analysis reveal the competing influences of a candidate’s social identities
and institutional positioning onto the incentive structures that are formed regarding negative
rhetoric. When examining patterns of negative emotive appeals based on candidate type, it is
clear that while all challengers are incentivized to appeal to more negativity than their incumbent
counterparts, the latitude by which they can do so can be constrained by subordinate social
positioning. My analyses thus far have provided substantial evidence that demonstrates how
political actors from subordinate RGIDs are restrictive in what they say with little regarding what
they may be incentivized to say. Therefore, to gain a more comprehensive understanding of the
intersectional impact of race and gender on rhetorical strategy, an evaluation of positive
emotional appeals is warranted.

154
Chapter 6
On a Positive Note
Playing it Safe and Clean in the House
Considering the substantial research dedicated to negative emotional appeals in political
communication, the study of positive valenced emotions is rather scant in comparison, and
typically studied in conjunction with negativity. While negative messages are frequently shown
to spread more rapidly online, prior scholarship offers some insight into the potential of positive
messages in reaching larger audiences, finding that people are more inclined to share and favorite
positive content (Ferrara and Yang 2015). Emotions such as enthusiasm and hope have been
recognized as powerful tools for promoting political engagement (Brader 2005; Just, Crigler, and
Belt 2007), with recent research uncovering gendered differences in how congressional elites
make use of positive discrete emotions in their public rhetoric. The expression of joy-associated
emotive rhetoric, for example, are found to be more frequently relied upon by women candidates
(Russell, Macdonald, and Hua 2023), suggesting that further investigation is warranted to better
understand the ways in which social identities may also shape distinct patterns of positive
emotive rhetoric. Despite the recognition of the unique role positive valenced emotions play in

155
political messaging, there is relatively limited attention afforded in extant scholarship allocated
towards theoretical development and measurement of discursive positivity in elite rhetoric.
The complexity of measuring this set of emotions may explain the disparities in research
(Eglo et al. 2003). Positive emotions have been described to a far lesser extent perhaps because of
their association with maintaining the status quo (Marcus and MacKuen 1993). Barbara
Fredrickson (2009) helps contextualize how the duality of positive and negative emotions is not
so cut and dry. From a neurobehavioral and developmental perspective, the origin of negative
emotions is to alert an individual to environmental discomforts or obstacles that in turn activate
said individual to adjust their behavior to adapt. Positive emotions however, per Barbara
Fredrickson, are not as linear as negative emotions for when they are triggered. Fredrickson
instead describes positive emotions to have developed as a reversal instrument for the effects of
negative emotions. Triggers for positive emotions, such as a desired result, are often in service to
a preceding state furnished by emotions that are comparatively more negative. This means
triggering a positive emotion is inherently a more complex process than triggering negative
emotions. Consequently, the reflexive urgency that negative emotions can evoke, is a scarcer
result from positivity.
In general, positive emotions indicate to individuals to keep doing as they currently do
(Brader 2005; Brader and Valentino 2007; Marcus and MacKuen 1993; Ridout and Searles 2011),
promoting such stability thus classifies many of the positive valenced emotions as deactivating.
The exception to this is enthusiasm. Enthusiasm is rooted in interest and optimism that can be
fuel to mobilize political action (Brader 2005, Brader 2006, Civettini and Redlawsk 2009; Ridout
and Searles 2011), which requires higher order thinking than say anger activated by physical
harm. Essentially, when trying to connect the dots of appealing to an audience’s emotions in
pursuit of the action driven outcome of voter-turnout, the shortest distance within a closed

156
system is always through negative emotion rather than a more roundabout and gradual positive
emotive route (Young 2021).
In the realm of political discourse, the strategic use of positive emotional appeals emerges
as a risk-mitigating approach for most political actors, a tactic devoid of potential pitfalls for the
speaker. Operating within the parameters defined by the brevity of Twitter's character limitations
and the discernment of leaning towards positive or negative emotional appeals, the deliberate
choice to harness positive sentiments presents itself as the more cautious alternative vis-à-vis
going negative. This strategic orientation aligns with the theoretical framework I have
established, wherein subordinate race and gender groups, aware of their marginalized status and
the accompanying uncertain audience expectations, rationally opt for the safer communication
strategy relative to the privileged group white men, who sit atop the race/gender social
hierarchy. This safer approach in this respect is sparing in appeals to negative emotions and may
bridge the gap with more positive emotional appeals in their public online discourse relative to
white men. While appeals to positive sentiment and discrete emotions can be used to produce a
variety of outcomes, I hypothesized that in settings with higher electoral threat that may
otherwise incentivize negative rhetoric to incite change, relative amounts of positive sentiment
and discrete emotive appeals can be an alternative representation of the constraints minoritized
RGID actors must face.
This paradigm of constraint is observable in the candidate tweets during the 2020 election,
a period underscored by amplified electoral threat across all candidates. Higher electoral threat
perception presumably makes a more hazardous perceived risk environment that may escalate
constraints appreciated by those with subordinate relative social positioning. Evidently, BIPOC
men, BIPOC women, and white women (p< .01). This significance underscores a greater than 1%

157
higher use of appeals to overall positive emotion relative to their white male counterparts (Table
23, Figure 15).
Table 23. OLS regressions of all 2020 House candidates’ monthly average use of positive
valenced emotions on Twitter (RGID as factor levels).

158
Figure 14. Model Estimates (w/ 95% Cis) of Race and Gender Effects to Baseline Group Set to
white Male MCs on Use of Positive valenced Appeals in Office Outreach Tweets.27
Note: Models fit with OLS. Error bars show the 95% confidence interval of SEs clustered
by MC and Congress. Other model specifications, controls, and fixed effects were
omitted from plot.
27 Plotted coefficients estimate the effects of RGID using a race/gender factor variable (contains four levels
total with the baseline group set to White male MCs) on MCs' monthly average use of emotion-associated
rhetoric per tweet (4 models total). Each set of RGID that do not correspond w/ the baseline group are
estimated directly in relation to the set baseline--i.e., White male MCs, which is represented by the dashed
line set to 0 to illustrate the difference between groups.

159
Positively Constrained: The Challenge of Challengers
In light of these findings, it becomes pertinent to scrutinize minoritized challenger
candidates who are doubly disadvantaged—first by their outsider status in political structures
and secondly by their social positioning. Within the dimension of political institutional
disadvantage, challengers are incentivized to draw attention to the election typically through
negativity (Druckman and McDermott 2008; Marcus, Neuman and MacKuen 2000). However, as
demonstrated in the previous chapter, the effects of institutional power dynamics on rhetorical
strategy may be opposed by social positioning effects if an actor is disadvantaged at the
race/gender intersection. Subordinate social positioning therefore manifests greater electoral
risks and a rational political actor employs a constrained communication strategy that is avoidant
of positivity. In the study of Twitter discourse that limits word count and forces a decision on
political actors to go negative or positive, avoiding the risky option of negativity is balanced by
adopting a constrained approach, which can thus be quantified by appeals to positivity.
Statistically significant regression coefficients (p< .01) across all underrepresented challenger
RGID (fct) variables indicate a >2% higher frequency of positive emotional appeals when
compared to white male challengers. This gap is also notably larger than that observed in the
overall pool of 2020 candidates (Table 23, Table 24).

160
Table 24. OLS regressions of 2020 House challenger candidates’ monthly average use of positive
valenced emotions on Twitter (RGID as factor levels).
To assess whether specific RGIDs among the challengers employed more positive
emotional rhetoric relative to their challenger peers, OLS regression analysis was done, the results
of which are presented in Table 25. Within this analytical frame, the statistically significant
coefficient for white male challengers underscores their 3.4% diminished use of positive emotive
appeals, relative to the rest of the challenger cohort. Conversely, BIPOC male and female
challengers both revealed significant regression coefficients implicate a proclivity for appeals to
positive emotions within these subgroups. With the assumption that positive sentiment appeals

161
are the safer approach in comparison to negativity, BIPOC challengers may be motivated to go
positive in response to subordinate relative social positioning and subsequently, their constraints.
White female challengers, however, did not exhibit statistically significant regression
coefficients, signifying that their appeals to positive emotion were commensurate with the overall
average for the challenger group. To delve into partisan differentiation within these findings, the
interaction term [RGID]*Republican was incorporated into the analysis. No substantial
disparities between Republican and Democratic challengers were found across white women,
BIPOC women, or BIPOC men. Republican white male challengers had a higher proclivity for
positive appeals compared to their Democratic counterparts, which might be attributed to the
influence of the sitting Republican president during the 2020 electoral cycle.

162
Table 25. Model estimates of RGID*Republican effects from separate regressions of 2020
challenger candidates’ monthly average use of positive valenced rhetoric.

163
A Positive Spin: Incumbent Messaging
Extending the analytical scope to incumbent candidates reveals another layer of
complexity. Given their inherent advantage in institutional recognition and public familiarity
(e.g., Fiorina 1989; Freejohn 1974; Gronke 2000; Jacobson 2004; Lazarus and Reilly 2010),
incumbents have a greater stake in preserving the status quo. In line with the literature that posits
positive emotions as conducive to maintaining preexisting attitudes and behaviors (Marcus and
MacKuen 1993), incumbents were found to overwhelmingly opt for a risk-averse, positive
emotional approach. According to data, incumbents had statistically significant regression
coefficients (p< .01), demonstrating an 8.7% higher frequency of positive sentiment appeals and
at least a 4.8% increase in the use of positive valenced emotions compared to their challenger
counterparts (Table 19). Notably, the incumbency advantage in adopting this "safer" strategy was
not confined to any particular RGID; it was consistent across all race/gender categories, as
evinced by the non-significant regression coefficients in Table 26. This uniformity suggests that
the calculus for maintaining a positive emotional tone transcends individual social identities
among incumbents. This pattern of appeals is reflective of the potency of incumbent advantage.

164
Table 26. Model estimates of RGID*Republican effects from separate regressions of 2020
incumbent candidates’ monthly average use of positive valenced rhetoric.
The confluence between incumbency and party identification also emerged as a salient
factor. Consistent with previous literature, the data suggests that incumbents, already established

165
within both the institution and their respective political parties, tend to conform to party-specific
expectations in their rhetorical choices (Druckman et al 2009). This parallels findings from the
previous chapter on negative appeals by candidates, indicating that incumbent candidates are
likely to foreground their partisan identities and adjust their race/gender strategies accordingly
to align with overarching partisan ideological frameworks.
The uniformity in adopting an approach rich in positive emotive appeals among
incumbents across different RGIDs raises intriguing questions about behavior under varying
constraints. Without the electoral duress of a campaign, would members of Congress appeal to
positive sentiment differently? One might anticipate that under less stringent constraints, all
underrepresented RGID groups would uniformly adopt safer, positive emotional strategies.
However, this expectation does not hold when analyzing MCs from the 113th to the 116th
Congresses.
113th to 116th Women Stay Positive
OLS regression analyses on these MCs' use of positive emotional appeals revealed a
distinct gender effect in political communication that was apparent in the individual race and
gender variables regression analysis. Analysis of the 113th to 116th cohort using a model
employing individual race and individual gender variables through regression similar to the
analysis in Chapter 4, the results are presented in Table 27. Significance in either race or gender
variables are suggestive of a stronger or weaker association with either of the individual social
identities. An interaction term (Woman*white) was also included to identify any differences
between white women vs. BIPOC women. The significant (p< .01) regression coefficient that is
present at ‘positive’ for ‘Woman’ but is not significant at ‘white’ indicates a higher propensity for
women to appeal to overall positive sentiment than men that is irrespective of race.

166
The discrete positive emotion, ‘anticipation’, has significant regression coefficients at
Woman and at white as well as on the interaction term Woman*white. This indicates that being
a woman and being white imparts a greater likelihood to appeal to anticipation than men and
BIPOC respectively, while BIPOC women have a higher propensity than white women to appeal
to anticipation. While these three relational values can aid in triangulating a valid inference, it
represents a limitation with this analysis alone in confirming which combinations of these
identities make the most impact on tweeting with positivity. Other analyses that involve
comparisons of RGID subgroups are thus necessary.
Table 27. OLS regressions of all 113th to 116th MCs’ monthly average use of positive valenced
emotions on Twitter (woman*white).

167
As was done with negativity, a regression was performed that compared specific
race/gender intersectional subgroups against white men to comparatively evaluate how these
different groups appealed to positive emotions in their public tweets. Table 28 presents the results
of this analysis that show significant regression coefficients for both white women and BIPOC
women (p< .01), suggesting 1.4% and 1.6% higher likelihoods, respectively, of appealing to
positive sentiments compared to their white male counterparts. Conversely, BIPOC men
exhibited emotional appeals that were largely analogous to those of white men (Table 33). This
gendered pattern corroborates earlier research connecting gender identity to positive emotional
appeals (Russell et al 2023), adding empirical weight to the distinct navigational challenges to
meet certain gendered expectations faced by women in legislative settings.

168
Table 28. OLS regressions of all 113th to 116th MCs’ monthly average use of positive valenced
emotions on Twitter (RGID as factor levels).
As the results presented in the previous chapter indicate, elites from underrepresented
RGIDs must carefully maneuver their partisan expectations in relation to their social identities.

169
Different gender expectations guided by different ideologies of the Democratic and Republican
parties establish distinct incentives for elites (Elder 2012; Reingold and Harrell 2010; Bauer 2015;
Krupnikov and Bauer 2014; Osborn 2012; Thomsen 2015), resulting in distinct patterns of
constraints and rhetorical appeals. It is therefore imperative to look at how party ideological
stances manifest in these patterns of constraints and rhetorical appeals.
To further explore the nuanced role of party affiliation in shaping these emotional appeals,
additional within-party OLS regression models were calculated for Democrats and Republicans
serving in the 113th to 116th Congresses. These models also add to our understanding of how
race and gender distinctly influence representatives’ use of positive sentiment and positive
valenced emotions in their in-office public outreach on Twitter, even when bound to the same
core party ideology. TablesTable 29, Table 30, Table 31, Table 32 report these regression
coefficients, which are followed by accompanying dot whisker plots for the coefficient of each
party’s subset model along with its associated 95% confidence interval (see: Figures Figure
16Figure 17Figure 18Figure 19). Republican model coefficients are plotted in red and Democratic
ones in blue. The plots provide visual depictions of how partisan MCs use of positive sentiment
and positive valenced emotions of anticipation, joy, and trust in their public outreach rhetoric on
Twitter, better revealing partisan differences in the relationships between each intersectional race
and gender identity classification and their respective strategic approaches in using safer and
more positive emotive rhetoric.
Among Republicans, whose conservative ideology adheres to more traditional gender
perspectives, one would expect an inclination towards signaling traditionally feminine
qualities, including the use of joy and other positive emotions in their communications. Indeed,
Table 29 and
Figure 16 reveals statistically significant regression coefficients (p< .01) for white women
Republicans, suggesting a 3.2% greater likelihood of employing positive emotive rhetoric relative
to the rest of the party. Conversely, statistically significant regression coefficients were associated

170
with both BIPOC and white Republican men, indicating decreased proclivities for these RGIDs
to deploy positive emotional appeals relative to the Republican party. It is important to note in
this analysis that drawing generalizable inferences for BIPOC women within the Republican
proves challenging due to their sparse representation in the data set.
On the Democratic side, a more equitable distribution of positive emotional appeals across
RGIDs may be expected given the party's more progressive stance on race gender roles. The
data largely aligns with this supposition. As shown in
Figure 16, Democratic RGIDs generally display positive emotional rhetoric more
equitably than their Republican counterparts.

171
Table 29. Model estimates of RGID effects (subset by party) from separate regressions of 113th
to 116th Dem. & Rep. MCs’ monthly average use of positive sentiment.
POSITIVE
term party estimate std.error p.value
White Woman Republican 0.032*** 0.008 0.000
BIPOC Woman Democratic 0.013 0.008 0.108
White Woman Democratic 0.001 0.007 0.839
BIPOC Woman Republican -0.001 0.031 0.972
White Man Democratic -0.003 0.006 0.581
BIPOC Man Democratic -0.006 0.007 0.370
White Man Republican -0.014* 0.008 0.094
BIPOC Man Republican -0.032** 0.016 0.046
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 16. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of positive sentiment per tweet.

172
Table 30 and Table 30. Model estimates of RGID effects (subset by party) from separate
regressions of 113th to 116th Dem. & Rep. MCs’ monthly average use of anticipation appeals.
ANTICIPATION
term party estimate std.error p.value
White Woman Republican 0.019*** 0.007 0.004
BIPOC Woman Democratic 0.017*** 0.006 0.003
White Woman Democratic 0.001 0.005 0.789
White Man Democratic 0.000 0.004 0.988
BIPOC Woman Republican -0.004 0.020 0.830
White Man Republican -0.005 0.006 0.409
BIPOC Man Democratic -0.013*** 0.005 0.007
BIPOC Man Republican -0.029** 0.012 0.013
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 17 present the coefficients associated with anticipatory appeals. A notable genderpreference emerges, with Democratic BIPOC women and Republican white women standing
out as primary exponents of anticipation-oriented appeals. Statistically significant coefficients
(p< .01) for these female RGIDs signify a 1.7% and 1.9% increase in anticipation-associated
rhetoric by BIPOC women and white women, respectively, relative to their affiliated parties.
Notably, BIPOC men were found to use the least amount of anticipation related words with
Democratic and Republican BIPOC men using approximately 1.3% to 2.9% less, respectively,
than the average congressmember in their corresponding parties (Table 30. Model estimates of
RGID effects (subset by party) from separate regressions of 113th to 116th Dem. & Rep. MCs’
monthly average use of anticipation appeals.
ANTICIPATION
term party estimate std.error p.value
White Woman Republican 0.019*** 0.007 0.004
BIPOC Woman Democratic 0.017*** 0.006 0.003
White Woman Democratic 0.001 0.005 0.789
White Man Democratic 0.000 0.004 0.988
BIPOC Woman Republican -0.004 0.020 0.830
White Man Republican -0.005 0.006 0.409
BIPOC Man Democratic -0.013*** 0.005 0.007
BIPOC Man Republican -0.029** 0.012 0.013
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 17).

173
Table 31 and Table 31. Model estimates of RGID effects (subset by party) from separate
regressions of 113th to 116th Dem. & Rep. MCs’ monthly average use of joy appeals.
JOY
term party estimate std.error p.value
BIPOC Woman Democratic 0.028*** 0.007 0.000
White Woman Republican 0.023*** 0.009 0.006
BIPOC Woman Republican 0.017 0.019 0.349
White Woman Democratic 0.009 0.006 0.160
BIPOC Man Republican -0.009 0.014 0.523
White Man Democratic -0.009* 0.005 0.078
White Man Republican -0.016** 0.007 0.024
BIPOC Man Democratic -0.016** 0.007 0.016
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.
Figure 18 highlight the distribution of appeals to the nuanced positive emotion of joy,
which bears resemblance to the patterns observed for anticipation. Democratic BIPOC women,
with a notable increase of +2.8%, and Republican white women, at +2.3%, emerged as the chief
exponents of joy-themed appeals in their respective party landscapes. Statistically significant
regression coefficients are reported for Republican white men (-1.6%), Democratic white men (-
0.9%), and Democratic BIPOC men (-1.6%), suggesting that male representatives are less inclined
to employ joy-associated appeals in their outreach online than their corresponding parties.
Table 32 and Table 32. Model estimates of RGID effects (subset by party) from separate
regressions of 113th to 116th Dem. & Rep. MCs’ monthly average use of trust appeals.
TRUST
term party estimate std.error p.value
White Woman Republican 0.015** 0.007 0.041
BIPOC Man Democratic 0.008 0.006 0.203
White Man Democratic 0.006 0.005 0.235
White Man Republican 0.003 0.007 0.710
White Woman Democratic -0.005 0.006 0.359
BIPOC Woman Democratic -0.014** 0.007 0.040
BIPOC Man Republican -0.028** 0.015 0.024
BIPOC Woman Republican -0.048** 0.024 0.016
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.

174
Figure 19, which delves into appeals associated with trust, reveal that Republican white
women stood at the forefront of their party. A statistically significant regression coefficient (p<
.05) reflects a +1.5% enhanced tendency among Republican white women to incorporate trustassociated words in their Twitter exchanges relative to their Republican peers. Conversely,
Republican BIPOC men registered a significant coefficient denoting a restrained utilization of
trust-related expressions in comparison to their copartisans.
Table 30. Model estimates of RGID effects (subset by party) from separate regressions of 113th
to 116th Dem. & Rep. MCs’ monthly average use of anticipation appeals.
ANTICIPATION
term party estimate std.error p.value
White Woman Republican 0.019*** 0.007 0.004
BIPOC Woman Democratic 0.017*** 0.006 0.003
White Woman Democratic 0.001 0.005 0.789
White Man Democratic 0.000 0.004 0.988
BIPOC Woman Republican -0.004 0.020 0.830
White Man Republican -0.005 0.006 0.409
BIPOC Man Democratic -0.013*** 0.005 0.007
BIPOC Man Republican -0.029** 0.012 0.013
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.

175
Figure 17. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of anticipation -associated rhetoric per tweet.
Table 31. Model estimates of RGID effects (subset by party) from separate regressions of 113th
to 116th Dem. & Rep. MCs’ monthly average use of joy appeals.
JOY
term party estimate std.error p.value
BIPOC Woman Democratic 0.028*** 0.007 0.000
White Woman Republican 0.023*** 0.009 0.006
BIPOC Woman Republican 0.017 0.019 0.349
White Woman Democratic 0.009 0.006 0.160
BIPOC Man Republican -0.009 0.014 0.523
White Man Democratic -0.009* 0.005 0.078
White Man Republican -0.016** 0.007 0.024
BIPOC Man Democratic -0.016** 0.007 0.016
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.

176
Figure 18. Plot of model estimates of RGID effects (by party) of 113th to 116th Dem. & Rep.
MCs’ monthly average use of joy-associated rhetoric per tweet.
Table 32. Model estimates of RGID effects (subset by party) from separate regressions of 113th
to 116th Dem. & Rep. MCs’ monthly average use of trust appeals.
TRUST
term party estimate std.error p.value
White Woman Republican 0.015** 0.007 0.041
BIPOC Man Democratic 0.008 0.006 0.203
White Man Democratic 0.006 0.005 0.235
White Man Republican 0.003 0.007 0.710
White Woman Democratic -0.005 0.006 0.359
BIPOC Woman Democratic -0.014** 0.007 0.040
BIPOC Man Republican -0.028** 0.015 0.024
BIPOC Woman Republican -0.048** 0.024 0.016
Note: Models fit with OLS for each RGID as a binary indicator by party subset (8 models
total). SEs clustered by MC and Congress.

177
Figure 19. Model estimates (w/ 95% CIs) of separate RGID effects of 113th to 116th MCs’
monthly average use of Trust Appeals in office account tweets.
Within the Democratic Party, BIPOC women emerged as the sole RGID group with a
statistically significant regression coefficient. This coefficient represents a -1.4% lower propensity
to use trust-associated rhetoric in their tweets compared to their Democratic peers. A notable
distinction in trust-themed appeals between white Republican women and BIPOC Democratic
women showcases a stark divergence between the two RGIDs that were otherwise the
predominant users of positive emotions in their tweets. This is perhaps reflective of delineation
of trust as an inhibiting emotion while anticipation and joy may represent more activating
positive emotions.
The results from the party subset models find that in terms of positive sentiment and
discrete appeals to positive activating emotions, Democratic BIPOC women and Republican
white women lead their corresponding parties in using such rhetoric. In applying my proposed

178
framework of social identity constraints, this data aligns well with the notion that increased
positive valenced emotive appeals corresponds to increased constraints. Notably true in the
results of the Republican analysis given the gender constraints applied to white women by virtue
of their conservative ideology. Democratic ideology, although more equitable in nature, does not
necessarily aid Democratic BIPOC women in overcoming their rhetorical constraints for the time
being, possibly owed to the relatively novel participation of this RGID within congress. As the
subset models clarified how specific RGID identities influence rhetorical choices relative to their
copartisans, a separate analysis is required to compare how elites who share intersectional
identities compare across party lines.
To interrogate the partisan intricacies influencing positive emotional rhetoric, Table 33
integrates the interaction term ‘[RGID]*Republican’ to contrast the rhetorical strategies of white
women within Republican and Democratic parties. The data reveal a significant ideological
cleavage between Republican and Democratic white women, as manifested by their dissimilar
employment of positive emotional language. Statistically significant coefficients at a level of p<
0.01 for the interaction term '[RGID]*Republican' delineate a greater inclination among
Republican white women to utilize overall positive emotions (3.4%), and more specifically, the
discrete emotions of joy, trust, and anticipation, each at 1.6%, in comparison to their Democratic
counterparts (Table 33).

179
Table 33. Subgroup main effects from separate OLS regressions of all 113th to 116th MCs’
monthly logged average use of positive valenced emotions on Twitter.

180
This empirical evidence is consonant with existing literature that characterizes Republican
white women as adopting more traditionally feminine rhetorical personas, while Democratic
white women embrace rhetorical strategies that more closely align with their male copartisans
(Krupnikov and Bauer 2014; Osborn 2012; Thomsen 2015; Russell et al. 2023). In essence,
Republican white women leverage positivity as a subsidiary tactical maneuver to fulfill partisanideological gender expectations and signal party affiliation through a constrained approach,
whereas Democratic white women possess a broader latitude to engage in negative rhetoric.
In relation to the findings on negativity presented in the preceding chapter, the proclivity
for Republican women to lean towards positive sentiment coheres logically within their more
traditional ideological parameters and is consistent with recent literature (Russell et al. 2023).
Specifically, Republican women tend to embrace positivity more robustly (Table 33), while their
Democratic peers took an approach more imbued with negative sentiment appeals (Table 8).
These findings corroborate my hypothesis such that taking the safer approach of ‘going positive’
can be an indicator for perceived constraint as a Representative of the House if said candidate
presumably constrained on their use of negativity feels the risks of going negative are too high in
each situation. Notably, this strategic divergence in emotional rhetoric develops post-election,
when these women secure congressional seats and become institutionally enmeshed in their
respective parties, given the overall similarity of emotional appeals between Republican and
Democratic white women challengers (Table 34).

181
Table 34. Subgroup main effects from separate OLS regressions of 2020 challenger candidates’
monthly logged average use of positive valenced emotions on Twitter.

182
Anticipation Proclamation—BIPOC Women’s Keys to the House
The application of anticipation within the realm of electoral communication and rhetoric
demonstrates a potent strategy. Politicians frequently invoke anticipation to stimulate excitement
about potential future outcomes, inspire hope, and craft a sense of urgency around their platforms
or policy agendas. Although the idea of a future focused emotion could be applied toward
evoking negative emotion, ‘anticipation’ is a strongly positive emotion as indicated by the
correlation matrix (Figure 11). Like other positive emotions, the breadth of literature investigating
this is limited and it has been investigated in the past under different emotion names such as:
‘hope’, ‘confidence’, or ‘enthusiasm’. Enthusiasm has been associated with garnering votes with
adherence to existing predispositions, while hope has been linked with information seeking
behavior (Brader 2005; Just, Crigler and Belt 2007)—which both provide utility for political
campaigns. Obama’s 2008 ‘Yes We Can’ slogan is a successful example of an appeal to
anticipation whose intention was to inspire change and promote unity. Thus, through
engendering a sentiment of anticipation, political actors can encourage constituents to participate
in action-oriented activities such as voting, campaigning, or lobbying.
As candidates are allowed only a limited number of words per tweet, decisions must be
made as to how they can communicate with their constituency. As demonstrated in Chapter 4,
not all RGIDs have the ability to take on the risks of going negative and potentially face more
serious electoral punishment from the public if they do so (McCormick and Jones 1993). As
expected, BIPOC women, given their subordinate relative social positioning, would
hypothetically adopt the safer, constrained approach of positive sentiment appeals. As certain
positive valenced emotional appeals, such as enthusiasm, can coalesce to get constituent bases
interested and politically engaged, it is not surprising that BIPOC women, who are electorally
incentivized like all other candidates, adopted an approach of future-focused anticipation appeals

183
that incorporates sentiments similar to enthusiasm in a manner distinct to their RGID that is
reflective of their lone designation of doubly disadvantaged in both race and gender identity axes
and therefore most constrained group status. Anticipation as an emotion category embodies
forward focused rhetoric, not unlike enthusiasm or hope, and represents a potential positive
valenced emotion classification that may be replicated in future studies, especially ones utilizing
the NRC dictionary.
Qualitative inspection of anticipation tweets seen in Table 35 revealed several aspects of
the coded emotion. The future oriented nature of anticipation can be seen in all of the following
example tweets. Note the Tweet by Rep. Pramila Jayapal—D, how it orients the audience’s focus
to “opportunity” and then a call to action with “vote”. Also see Rep. Judy Chu—D, calling for
initiation of the Defense Production Act now in anticipation of the rising need of masks during
the COVID pandemic. Rep. Norma Torres—D uses anticipation to inform her constituents of how
they can contribute to the improvement of their local airport with their votes. Rep. Ileana RosLehtinen—R also appeals to anticipation by highlighting her stance on a specific immigration
legislation with a call to action with the “vote tomorrow”. Note that these tweets, all from BIPOC
women from both Democratic and Republican parties, are used to frame their stances on issues
with a call to action. It seems to be used in a similar manner to anger or fear in terms of presenting
elites’ stances on issues, which is understandable given the substantial co-presentations of
anticipation with anger in tweets, especially in BIPOC women, in which 94% of their anticipation
appeals have anger associated words in the same tweet.

184
Table 35. Examples of anticipation-associated rhetoric in 113th to 116th MC office account
tweets.28
member party text score
JAYAPAL,
Pramila
Democratic “ICYMI: Our state has the opportunity to
restore equal opportunity for all in all state
schools, jobs, and facilities. Vote to approve I1000 when you mail in your ballot. (Friendly
reminder: You don't need a stamp to mail your
ballot, but it must be postmarked by
tomorrow!)”. 2019-11-05.
9
ROSLEHTINEN,
Ileana
Republican “I'm proud of my legislative hermano
@MarioDB for staying true to his longtime
commitment to modernizing our immigration
system and providing a pathway to citizenship
for #DREAMers. I'm glad that we may finally
have the opportunity to do so through our vote
tomorrow. QT @MarioDB This bill allows
dreamers to stay here and allows them to
become part of society w pathways to
citizenship, it legislatively stops the separation
of minors from their parents on the border, and
it secures the border. *That* is what this bill
does. Everything else is cheap rhetoric.” 2018-
06-21.
7
CHU, Judy Democratic “Our healthcare workers are putting
themselves at risk to fight #coronavirus and
they deserve our help. That is why President
Trump must use the Defense Production Act to
immediately begin production on masks and
ventilators. We cannot wait.
#DefenseProductionActNOW #MasksNow.”
2020-03-27.
7
TORRES,
Norma
Judith
Democratic “Ensuring that @flyONT can continue to grow
and thrive is critical to the success of the Inland
Empire. Read about my work to help Ontario
International Airport reach its full potential
here: QT @ivdailybulletin How federal defense
funds can help Ontario International Airport
grow”. 2018-07-05.
7
28 URLs are removed from tweet text for clarity. Unedited and complete texts for the included tweets are
available in the Appendix.

185
Within the domain of emotional rhetoric employed by BIPOC women representatives, a
salient disparity emerges in the Democratic Party, particularly in their utilization of anticipation
and trust-related language. Quantitative data reveal a noteworthy discrepancy between BIPOC
women and the remainder of Democrats in the appeal to these specific emotional states—marked
by a +1.7% increase in the use of anticipation-oriented rhetoric and a -1.6% decrease in trustassociated language (Table 33). This divergence can possibly be explained through the lens of
higher-intensity emotional states as denoted in Plutchik’s emotional wheel (Figure 20), which
categorizes anticipation under 'interest' and trust under 'acceptance'.
The intensified focus on anticipation, classified under the emotional state of 'interest,'
suggests a more apprehensive orientation, potentially reflective of the systemic challenges and
anxieties associated with BIPOC women's intersectional non-privileged positioning. Conversely,
the diminished emphasis on trust, grouped under the emotional state of 'acceptance,' intimates a
relative absence of experienced comfort among these representatives. The emotional tenor of
'acceptance,' imbued with connotations of comfort, seems less accessible to BIPOC women, who
may find themselves in positions of systemic disadvantage. Thus, these findings imply that the
differential employment of these positive-valence emotions serves as a nuanced signal of the
unique experiential and institutional contexts faced by BIPOC women within the Democratic
Party. This dynamic presents an interesting and potentially ripe area for future research given
that extant literature has notably less understanding of the form and function of positive emotions
(Frederickson and Levenson 1998).

186
Figure 20. Diagram of Plutchik’s (2001) wheel of emotions (as shown in Figure 4, p. 71)
Chapter 6 Summary
Constrained to Taking the High Road
While extant research has predominantly focused on the impact of negative emotions due
to their greater efficacy in influencing voter behavior, the role of positive emotions remains
optimism love
aggressiveness submission
contempt awe
remorse disapproval
annoyance ANGER rage
vigilance
ANTICIPATION
interest
ecstasy
JOY
serenity
admiration
TRUST
acceptance
terror FEAR apprehension
amazement
SURPRISE
distraction
pensiveness
SADNESS
grief
boredom
DISGUST
loathing

187
comparatively underexplored. This research gap is further exacerbated by the complex variability
within the spectrum of positive valenced emotions, which lends itself to a complex investigative
landscape.
Extant literature suggests that differential voter responses to emotional appeals are
contingent upon the race and gender of the audience (Phoenix 2020). This implies that emotional
cues emanating from varied RGIDs are likely to elicit distinct reactions. My study posits that
underrepresented RGIDs, constrained by their perceived electoral vulnerabilities, would
naturally shy away from employing higher risk communication approaches, such as negative
emotional rhetoric. Instead, these groups may strategically pivot toward the utilization of positive
emotive appeals. Empirical evidence from my analysis corroborates this hypothesis, as
demonstrated in Chapters 4 and 5, where groups that exhibited restraint in their use of negative
rhetoric—such as BIPOC Members of Congress, nonwhite male challenger candidates, and white
female Republicans—displayed a pronounced inclination toward positive communications when
contrasted with their white male counterparts.
Therefore, what emerges is a 'constrained approach,' whereby underrepresented and
unconventional RGIDs are compelled to navigate their public communications through the safer
harbor of positive sentiment appeals. This strategic choice inherently entails a diminution in the
deployment of negative rhetoric, situating these groups in a unique relational context vis-à-vis
the more emotionally expressive white male cohort.
Identities of the Constrained
In the heightened electoral climate characterized by amplified electoral threat, challenger
candidates under the most pronounced social constraints—namely white women and BIPOC
challengers—exhibited a marked preference for positive emotional appeals. This observation

188
complements the existing literature that underscores the incumbent advantage as not only
reducing reliance on negative emotional strategies but also amplifying the usage of positive
appeals, lending credence to the theory that positive emotional rhetoric fosters a sense of stability
and reinforcement of the status quo.
Notably, the role of political affiliation as a moderating variable was also confirmed in the
realm of positive emotional appeals, paralleling the findings on negative emotions discussed in
the previous chapter. white women in the Republican Party, operating within the traditional
gender expectations of their political affiliation, manifested higher levels of positive emotional
appeals compared to their male co-partisans as well as Democratic women. This observation
aligns with prior studies, further illustrating the nuances in the interplay between gender and
party affiliation in the strategic deployment of emotional appeals.
The study also discerned a clear gender bias in the strategic use of positive sentiment
appeals. Women from both political affiliations deployed these appeals more frequently
compared to their male counterparts, with BIPOC men utilizing them even less than the average
rates within their respective parties. In particular, Democratic white women appeared to align
with their party's average use of positive sentiment, presumably exceeding the levels employed
by Democratic BIPOC men.
Discerning Positive Valenced Emotions
My research also contributes to an understanding of the heterogeneity within the
spectrum of positive valenced emotions, particularly in their efficacy in voter mobilization.
Specifically, anticipation emerged as the dominant positive emotional strategy leveraged by
BIPOC women, while there was a noticeable decline in the usage of the emotion of trust. This
reflects the duality of different positive emotions with regards to activation—similar to the

189
varying effects of negative emotions. From a utility standpoint, anticipation in this respect can be
defined as a positive emotion focused on the future. This suggests that positive emotions such as
enthusiasm or hope fall under the anticipation umbrella, and collectively, anticipation represents
an optimistic emotion that may possess superior utility in mobilizing the electorate. According to
Plutchik’s emotions wheel, anticipation can be categorized under 'interest,' thereby further
substantiating its role as an effective voter engagement strategy.
In the examination of the other positive valenced emotions outside of anticipation, such
as joy and trust, these emotions are inherently deactivating. Categorizing joy and trust with
milder emotional states of serenity and acceptance, as per the Plutchik wheel, supports the notion
that these positive valenced emotions that are not anticipation are deactivating. Amongst these
emotions, the emotion of joy emerged as a baseline or normative element in political elite tweets,
akin to my finding with appeals to anger in the previous chapter. Joy exhibited no significant
variations across the categories analyzed, suggesting its foundational role in shaping appeals on
Twitter.
Positive sentiment and positively valenced emotional appeals are generally considered
safe approaches to political elite online discourse given the typically action-abating nature of most
positive emotions (Fredrickson 2009). Additionally, as political actors hailing from
underrepresented racial and gender identities may feel constraints against being fully expressive
given the risks associated with going negative as a subordinate race and gender. The forward
focused discrete emotion of anticipation is a potential nidus of future studies given such a distinct
reliance on this approach.

190
Chapter 7
Conclusion
In politics, talk is often cheap and attention scarce. The overloaded information
environments of new media technologies offer fierce competition yet more opportunities than
before for political elites to broaden their reach—creating incentives to not only tweet more
frequently, but to do so in a way that strikes the right balance between the need to distinguish
themselves while still conveying enough familiarity to importantly connect with voters. While
the advent of social media has certainly opened invaluable entry points that have permitted new
diverse voices to enter the political process that has long been exclusionary of diversity, it remains
to be seen whether social media as a platform for political outreach ultimately helps advance
diverse interests or rather, simply perpetuates inequities borne from existing social structures.
In this dissertation, I investigate the ways intersecting identities of race and gender shape
the rhetorical strategies that legislative representatives employ in their online public outreach.
Focusing on the digital arena, which serves as a burgeoning platform for maintaining the
'electoral connection' (Mayhew 1974), I scrutinize the nuanced challenges that representatives
from underrepresented backgrounds face in the calculated risks they take with emotional
appeals. These online platforms, epitomized by Twitter's ceaseless streams of content and high

191
network reactivity, have democratized public outreach but at a cost. Elected officials increasingly
prioritize strategies designed not to enlighten or inform, but rather to secure heightened public
engagement and visibility.
In the highly reactive and content-saturated environment of platforms like Twitter, the
conventional objective of public outreach has been upended. The primary goal for elected officials
is less about substantive dialogue and more about maximizing engagement and reach. This shift
elucidates new contours of risk and reward for public figures, particularly those from
underrepresented backgrounds.
This chapter serves multiple functions. First, it synthesizes the results from prior chapters,
addressing the significant knowledge gap regarding intersecting identities like race and gender
in political communication. It then elucidates the broader implications of these findings in
relation to my initial research questions. Furthermore, I discuss the limitations inherent in this
study and outline avenues for future scholarly investigations. Importantly, my work reinforces
the theoretical framework I propose, offering a foundational perspective on the influence of
intersecting social identities on elite political rhetoric.
One of the principal motivations for undertaking this dissertation was to explore the
conditions under which political elites employ negative rhetoric—an established strategy for
garnering online attention. This question gained new dimensions in the wake of social
movements such as the Black Lives Matter, when a surge of negative emotions like anger, fear,
sadness, and disgust permeated public discourse. As Phoenix (2020) notes, however, not all
political actors benefit equally from the deployment of such emotional expressions.
By leveraging computational textual analysis on an expansive dataset—comprising over
three million Twitter communications from congressional offices and campaigns from 2013 to
2021—this study advances both theoretical and empirical understandings of the rapidly evolving

192
landscape of congressional communication in the digital age. My findings raise critical questions
about the potentially exacerbating role that these highly competitive online platforms play in
fostering political divisiveness. Moreover, the research suggests inequitable electoral
consequences for underrepresented candidates who, unlike their more privileged counterparts,
cannot capitalize as effectively on the online vitriol.
Focusing on the 113th through the 116th Congress, as well as the 2020 electoral cycle, this
dissertation disentangles the intricate layers in the strategic utilization of emotional appeals in
political discourse. It considers the intersectionality of race and gender as key factors shaping this
landscape. The findings shed crucial light on the enduring inequities in the political process,
inequities that disproportionately burden historically marginalized communities who remain
underrepresented in political offices yet stand in greater need of representation. This dissertation
thereby provides a nuanced understanding of how RGID influence the emotional undertones
framing the political dialogue articulated by members of Congress.
Summary of Findings
My study fortifies the extant literature on the transformative role of Twitter in politics,
providing an unparalleled platform for candidates who historically exist outside the conventional
political paradigms to gain notoriety and even electoral success. Twitter as a platform both
broadens the reach of engagement while also creating a finite communication environment by
setting word-count limits. Word-count limits force political actors to make decisions on how to
budget their public discourse online, such that prioritizing certain strategies would require
sacrificing allotted words for others. With this assumption, I can more safely make inferences of
how certain patterns of emotional appeals can speak to a given actor’s constraints in relation to
their peers. In the present study, the unconstrained or ‘expressive’ approach is marked by a

193
political actor’s willingness to appeal to negativity, which means said rational actor has weighed
potential risks to be less than someone avoidant of negativity. Conversely, the risk-averse or
‘constrained’ approach, avoids the riskier negativity in favor of the safer alternative of positivity.
Thus, rhetorical constraints that are inferred to be in place as a result of heightened electoral risk
can be quantified by relative use of positivity in addition to an avoidance of negativity with the
baseline being the dominant group, white men.
While the utility of negative campaigning has been long established in political science
literature for its attention-capturing capacity (Rydell and Mackie 2008, Valentino et al. 2008,
Brader 2005), my study extends the framework by illustrating the nuanced use of discrete
negative emotions, chiefly anger, as an increasingly standard rhetorical tool among political
elites. With anger as the current norm, as the results of my analysis suggest, political elites may
try to further differentiate themselves and reap the benefits of ‘going negative’ beyond anger by
possibly going more negative with other discrete negative emotions. Variations in the use of
overall negative sentiment appeals and appeals to specific negative valence emotions—disgust
and fear—were evident along racial lines such that white MCs, in general, appealed to negative
emotions more than BIPOC MCs. BIPOC political actors, bound by social constraints informed
by social dominance theory and their demographically unconventional and subordinate status,
may have some awareness of the corresponding differences in how certain emotional appeals
they make may not be received as they would have intended by predominantly white audiences
(Piston 2010; Reeves 1997; Terkildsen 1993; Phoenix 2020). Therefore, because of the risks of
triggering detrimental outgroup biases and electoral punishment, BIPOC MCs may appraise a
given discursive situation to have greater risks than a white MC. This in turn may incentivize
BIPOC MCs to be more sparing with their negative emotional appeals and make the rational
choice to avoid this rhetorical strategy in comparison to their white counterparts.

194
Given that my analysis reveals disgust as a differentiating factor between expressive and
constrained approaches, this negative valence emotion presents itself as an area of future
research. Specifically, investigating the public's perception of elected officials who employ riskier
strategies like disgust-associated rhetoric is crucial. This becomes particularly salient in the
context of recent findings linking emotions like disgust to the viral dissemination of fake news on
platforms like Twitter (Vosoughi, Roy, and Aral 2018).
I also investigated the variance of negativity appeals among RGIDs during specific
campaign periods marked by elevated electoral risk and the incumbent advantage. Contrary to
the prevailing monolithic perspective that challengers uniformly adopt a negative tone, the data
indicates a more nuanced approach. Specifically, the propensity to 'go negative' as tied to the
Prospect theory (Druckman, Kifer, and Parkin 2009Druckman et al. 2009, McDermott, Fowler,
and Smirnov 2008), was prevalent among candidates belonging to the dominant race and gender
demographic—white males. white women and BIPOC challengers used less negative emotive
appeals than white male challengers while also relying more on positive sentiment appeals,
which is suggestive of race/gender minority candidates adopting a more constrained approach
in response to the amplified perceived electoral risk derived from their disadvantaged social and
institutional positioning. This is consistent with my framework in which social identity acts as a
baseline rhetorical constraint whose appraisal of going negative supersedes the reward incentives
of going negative as a challenger.
Additionally, my campaign data underpins the well-documented advantage incumbents
possess (Alford and Hibbing 1981; Ansolabehere, Brady, and Fiorina 1988; Cox and Katz 1996;
Erikson 1971; Ferejohn 1977; Fiorina 1977; Jacobson 1987; King and Gelman 1991; Krehbiel and
Wright 1983; and Mayhew 1974), evidencing a near-uniform aversion to deploying negative
sentiment appeals in favor of adopting a more positive tone to their communications across

195
different race and gender RGIDs. Such reticence aligns coherently with the conventional wisdom
that incumbents would rather not attract undue attention to the electoral race, thus strategically
avoiding the high-stakes arena of negative campaigning while relying on making less polarizing
positive sentiment appeals that dispose individuals to maintain the status quo (Brader 2005;
Brader and Valentino 2007; Marcus and MacKuen 1993; Ridout and Searles 2011). My incumbent
data highlights the robust effect of having institutional advantage, a dynamic that pivots the
incentives that inform whether to go positive or negative from drivers tied to race/gender social
identities to respective partisan identities.
As the campaign period demonstrated a seemingly unified approach amongst
incumbents in relation to challengers, my analysis of the 113th to 116th Congresses finds notable
divergences in the approach to political communication during the governance phase among
RGIDs. Indeed, the marked variation observed in communication strategies post-election
underscores the significant and continuing influence of disparate social identities on MCs’
rhetorical strategies as a race/gender minority. Distinct patterns of appeals to negative and
positive sentiments and valence emotions correlate in a way that supports my model relating
social positioning and associated constraints all vary amongst different RGIDs that seems to
correlate with degree of intersectional disadvantage (i.e., disadvantaged in race only, vs. double
disadvantage, vs. gender only disadvantage). Importantly, the rhetorical patterns that manifest
from the constraining expectations of the public in general coincide with party preference and
the liberal vs. conservative core ideologies.
As elections approach, there is a noticeable reticence to cast votes on potentially divisive
bills, which is a strategic move enacted by the parties to prime party loyalty amongst legislators
(Lindstädt and Vander Wielen 2012). Similarly, previous ads research by Arbour (2014) finds
party’s role in moderating partisan ideological appeals in ads that is contingent on target

196
audiences’ degree of ideological extremity (Arbour 2014). My data corroborates the findings of
these studies that have implicated a significant influence of political parties on elite rhetoric and
behavior. With my analysis, I find a significant role of political party affiliations as a moderator
for emotional rhetorical constraints that is best demonstrated by the variance seen in white
women MCs. Democratic white women, situated within a party framework that promotes
expressive emotional rhetoric across genders, engaged more extensively in negative sentiment
appeals, exhibiting levels of disgust comparable to their white male party peers and higher levels
of fear based rhetoric. Conversely, within the Republican Party, the prevailing traditional gender
norms, which may frown upon negative emotional appeals from women, disincentivizes such
rhetorical choices among white female elites. This results in a more circumspect and constrained
utilization of negative emotional rhetoric compared to their male counterparts in the same party,
as well as their peers in the Democratic Party.
As it pertains to positive emotional appeals, my study finds similar female gender
associations as extant literature (e.g., Krupnikov and Bauer 2014; Osborn 2012; Thomsen 2015;
Russell et al. 2023), while also helping contextualize this “feminizing” approach these authors
have alluded to by parsing out party-associated rhetorical divergences. Democratic white women
appealed to positivity less by virtue of their expressive approach, while Republican white women
leverage positivity in their constrained approach, which can now be surmised to equate to the
“feminizing” approach described in literature.
Positive appeals in the Democratic Party were dominated by BIPOC women MCs, lending
credence to the prolific effects of social positioning on elite discourse even when actors are already
within the institution and bound to the party whose ideologies favor more equitability amongst
disparate RGIDs. The unique positioning of BIPOC women as disadvantaged in both the race and
gender axes suggests that they perceive highest constraints within the Democratic Party. This also

197
is suggestive of the continued effects of social identities and perhaps contextualizes the combined
effects of intersecting race and gender social identities that are both disadvantaged.
Within their increased use of positive sentiment appeals, BIPOC women were found to
particularly rely on appeals to ‘anticipation’. This future focused rhetoric encompasses positive
emotion previously described as ‘enthusiasm’ or ‘hope’ and is a safer, although potentially less
potent (Young 2023; Costa et al. 2022), route to stimulate political engagement and votes than
traditional negative emotional appeals. Furthermore, ‘anticipation’ associated words had a
relatively high co-occurrence with ‘anger’ associated words when used by BIPOC women
distinctly. This may reflect a learned tactic for BIPOC women to present their stances on issues
and signal their partisanship to their constituents. Furthermore, varying co-presentations of
negative valence and positive valence associated words is also suggestive of a more complex
approach to emotional appeals than one that is reduced to a dichotomy, a question potentially
answered by future research.
Significance of Study
My study contributes to the relative dearth of literature examining intersectionality within
political communications and illuminates important associations with varying RGID identities
and how different RGIDs make appeals to both negative and positive emotions. I additionally
provide further perspective on how positive sentiment appeals and positive valence emotional
appeals are differentially used by different social identities and find associations that may launch
future investigations to better understand positive emotions.
From a methods perspective, my analysis utilizes one of the most robust datasets of
congressional Tweets that spans one term each of a Democratic and Republican president and
examines trends in differing periods of campaign and governance. With the various regression

198
models applied to this data set, I was able to reveal data that can bolster the theoretical
foundations of American political communication, which I will discuss in further detail in the
following subsections.
The Social Underpinnings of Congressional Communication
My research enriches the field of political communications by adopting a social
constructionist lens to examine variations in elite communication, thereby identifying the social
underpinnings of political actors’ approaches to navigating political spaces. In doing so, this
study conceptualizes the ‘public outreach’ or discursive element of a representative's 'homestyle'
as fundamentally rooted in a relational communicative process.
Schattschneider (1975) describes politics as the extent to which the public is involved in
the scope of conflict, which in turn serves as the foundation for the most fundamental strategy in
politics—whoever controls the message wins the game. The crucial component here lies in first
evaluating "the processes by which the unstable relation of the public to the conflict is controlled"
(Schattschneider 1975, 3), and the extent to which social processes in particular dictate the
parameters of political conflict.
The context of Schattschneider's assertion was one in which political discourse operated
predominantly as a top-down dynamic, guided chiefly by the most powerful politicians within
an institutional hierarchy. This historical trajectory has, understandably, narrowed the focus of
extant research on elite rhetoric. However, the advent of social media has democratized political
engagement, reinvigorating the social-relational dimensions of politics. This shift not only
broadens the scope of political influence but also highlights the gap of research of political
processes in tandem with social processes.

199
Recent scholarship in political communication has predominantly focused on political
processes, often attributing observations to an individual's position within political institutions
or party affiliations. This focus frequently overlooks the representative-constituent relationship
as deeply entrenched in broader social contexts such as prevailing social hierarchies.
Political outreach remains a relational phenomenon rooted in these broader social
dynamics. Sender-receiver elements of messaging, collective emotional states, and prevailing
social norms substantially influence public expectations for political actors, thereby delineating
the boundaries between acceptable and punishable discourse. As my research reveals, social
identities, including race and gender, play a pivotal role in shaping the strategic communication
approaches of political elites.
The advent of social media platforms like Twitter has expanded Schattschneider’s "scope
of conflict," allowing for a more democratized control of narratives. Figures like Representative
Alexandria Ocasio-Cortez have demonstrated the power of social media clout in influencing
legislative processes, even for a freshman legislator. Furthermore, platforms like Twitter have
democratized the practice of "going public" (Kernell 2006), extending the ability to politically
engage on a national stage—once the exclusive domain of presidential candidates—to
congressional representatives and aspirants alike.
Those who opt for such broad, online engagement become public figures subject to wider
societal norms but are also uniquely positioned to influence those norms. This dynamic
underscores the importance of accounting for social context in political analysis; politics is no
longer confined to Capitol Hill but has infiltrated American news feeds.
The viral potential of social media empowers political actors to control narratives and
incentivizes the use of attention-grabbing strategies, often encompassing negative appeals. My
dissertation thus investigates and reveals that this ability to use such appeals and potentially

200
maximize online engagement may not be equitable across various social identities. Furthermore,
this suggests that this online ecosystem may perpetuate systemic inequities in public discourse
by encouraging greater usage of negativity among those who can 'afford' to be negative.
The Negativity Divide and the Uneven Platform of Digital Media
In the present study, I elucidate a significant divide in emotive strategy, particularly
pertaining to the deployment of negative emotions, in congressional online outreach between
candidates positioned within societal structures of privilege—chiefly white male candidates—
and those who are underrepresented. This divergence underscores the limited potential of social
media to fully equalize the electoral playing field, particularly for candidates operating within
the confines of racialized and gendered expectations from the electorate. While these online
channels ostensibly offer unconventional candidates and those with fewer resources a larger
platform, their communicative behavior remains significantly constrained. These constraints
arise from the unique set of challenges presented by their nonprivileged social positioning,
limiting their expressive agency and thereby compelling them to deviate from established
campaign strategies historically successful for privileged groups.
Given that electoral viability hinges on the ability to forge meaningful connections with
voters, candidates from historically marginalized racial and gender backgrounds must navigate
a labyrinthine path, distinctly different from their privileged counterparts, to achieve electoral
success. I further posit that these candidates from intersectionally marginalized groups are more
inclined to adopt a constrained strategic approach in their public communications, generally
opting for less frequent expressions of negative-valenced emotions, despite the established
efficacy of negative campaigning as a longstanding strategy.

201
While high-profile cases like that of Alexandria Ocasio-Cortez might suggest the
possibility of breaking through these entrenched barriers, it is crucial to recognize such instances
as exceptions rather than norms. This study serves as an essential step in reevaluating traditional
theories of candidate strategic behavior and messaging, which have been historically
conceptualized with a default focus on white male political actors.
The findings of this dissertation underscore the intricate interplay between race and
gender identities, as well as the broader social structures that inform intergroup dynamics, in
shaping the political experiences of candidates pursuing public office. My analysis reveals a
differential utilization of emotional rhetoric among RGID groups aimed at achieving their
electoral objectives. Specifically, candidates from underrepresented and historically marginalized
backgrounds encounter heightened electoral threat in political and public spaces that apply
constraints in their rhetorical choices when electoral stakes escalate, and the risk is too great.
Such a nuanced understanding disrupts the assumption of a homogenous response to
electoral threats among political actors. This is pivotal as not all identity groups stand on an equal
footing nor are perceived uniformly by the electorate. The variances in social positioning demand
different navigational strategies, or constraints, by subordinate RGIDs across a racialized and
gendered socio-political landscape. These constraints underscore greater uncertainty among
candidates with nonprivileged social positioning about how their rhetorical signals may resonate
with voters who harbor ambiguous expectations that are reflective of the candidates’
unconventional status.
In the realm of strategic communication, candidates with underrepresented RGIDs have
limited latitude, particularly in the face of amplified electoral risk. This study contends that the
latitude to 'go negative'—a strategy well-established for its efficacy in capturing attention (Kern
1989) and engaging partisan bases (Iyengar 1995)—is a privilege chiefly accessible to dominant

202
social groups. Further, within the context of governance and public outreach, there are nuanced
differences among underrepresented RGIDs in their employment of negative valence emotive
appeals. Despite being normatively unsavory, their proven effectiveness in expanding reach and
resonance (Vosoughi et al. 2018; Huddy, Feldman, and Cassese 2007; Tiedens and Linton 2001)
lends them considerable allure, both for engaging potential voters and reinforcing partisan
heuristics. This renders negativity a seemingly assured path to electoral success, thereby
motivating its adoption by both major parties. continued congestion of political Twitter feeds
with negative rhetoric that is not accessible to all RGIDs ultimately serves to maintain and
perpetuate existing systems of oppression.
However, the broader implication is worrisome: if the luxury to employ this effective yet
risky strategy is available only to the privileged, its continued encouragement in digital political
dialogue would logically serve to perpetuate existing systems of oppression. This raises a salient
question: to what extent does social media, often celebrated for democratizing discourse, merely
serve as a conduit for the persistent racial and gender inequities that plague our political arena?
Given the normative importance in democratic theory of having an inclusive elected body
that reflects the diversity of the constituents it represents (Mansbridge 1999; Phillips 1995; Pitkin
1967; Williams 1998; Young 2000; Alexander 2012), how social media may disproportionately
constrain candidates from underrepresented groups is an increasingly relevant and important
question that future research can address further.
Tweeted Boundaries: Constrained through the Prism of Race and Gender
While all political actors seeking public office experience constraints associated with
having to adhere to the set of norms that correspond with the messaging platform (in this case
Twitter) their public outreach takes place on, as well as the norms and expectations of their

203
specific partisan ingroup of voters they seek to form a connection with, the degree to which they
are constrained is shaped by the relative group positioning of their intersectional race and gender
identity in the broader social context.
Building on the notion that the degree to which political actors are constrained is shaped
by their intersectional race and gender identity within a broader social context, my study
delineates the unique challenges faced by candidates with marginalized social identities in
navigating online communication platforms. Specifically, the mechanics and norms of platforms
like Twitter can introduce disproportionate constraints on these political actors.
While the adage "any press is better than none" may ring true for some, it fails to
universally apply, particularly when it comes to platforms like Twitter. These platforms
inherently offer less control over messaging, and the constraints of character limitations can
facilitate misinterpretations (Hua and Chapp 2020). The trade-offs between risk and reward are
significantly skewed for candidates from historically underrepresented groups. For instance,
white male candidates are more inclined to deploy negative rhetoric on Twitter, as they
traditionally face lesser electoral repercussions compared to their BIPOC and female
counterparts. This disparity is especially evident among non-incumbent challenger candidates,
who must navigate the compounding effects of negative public perceptions related to their
gender and race.
Further complicating the landscape is the reality that when communicating on a national
level, candidates from marginalized groups often face audiences whose majority may not share
their race or gender. This situation creates a unique set of ambiguous audience expectations and
biases, thereby elevating the risks associated with their messaging. These are challenges that
dominant groups, beneficiaries of positive societal stereotyping, seldom encounter.

204
To fully apprehend the complex interplay of race and gender on congressional
communication, it is indispensable to consider the moderating role of partisan affiliation in
shaping these dynamics. This is particularly crucial given that party affiliation has been
increasingly integrated into the American social fabric, forming a significant psychological
attachment for many (Green 2004). Consequently, party ideology continues to shape voter
expectations vis-à-vis race and gender, complicating the communicative strategies for Democratic
and Republican candidates differently. My data further highlights partisan disparities in the
rhetorical strategies employed by white women members of Congress (MCs). Such discrepancies
in partisan expectations are particularly salient when the political actors are institutionally bound,
whether as incumbents or members of Congress. This beckons the question: when these members
of Congress become incumbents during campaign years, is there a change in their communication
strategy?
In contrast to Blumenthal (1980), who contends that the ever-present threat of elections
creates a constant state of campaigning marked by unvarying negativity, my research aligns more
closely with recent scholarship advocating for a nuanced understanding of temporal variation
within the electoral cycle (MacDonald 2020; Kreiss et al. 2018). Specifically, strategic goals among
political practitioners are subject to temporal shifts, necessitating scholars to account for timing
when studying congressional communication.
In my analysis, it becomes apparent that campaign periods are imbued with heightened
emotional charge compared to governing periods. This shift in rhetorical strategy during election
years aligns with Druckman (2009), who asserts incumbents have a greater compulsion to
respond to challengers as the election nears. This effect is especially pronounced in the era of
social media. Furthermore, my study uncovers a noteworthy variance in emotional rhetoric
across different platforms of public outreach. Specifically, office accounts, which are theoretically

205
distinct from campaign-specific accounts, also exhibit an uptick in negativity during election
years. While my data does not definitively attribute this pattern to incumbent reactions to
challengers, it does highlight a consistent strategic adjustment correlated with election years—a
phenomenon that merits further scholarly exploration.
As my study demonstrates how social identity-based constraints can manifest under
certain contextual factors, my results and analyses contribute meaningful application of race and
gender intersectionality to political communication theory.
Signaling Identity: Advancing a Dynamic Framework of Strategic
Communication
My research offers a substantive contribution to the field of political communication by
introducing an intersectional lens to the study of race and gender, thereby formulating a
theoretical framework that elucidates how social identities shape rhetorical strategies and
electoral risk landscapes. Utilizing principles from sociology and behavioral psychology, this
framework provides a robust explanatory model for addressing inequities manifest in online
political discourse.
Historically in political communication studies, race has been relegated to a control
variable, receiving insufficient attention in the realm of strategic communication theories. In
contrast, my work elevates an intersectional understanding of race and gender, incorporating
theoretical advancements from race and ethnic politics to better elucidate the variances in
rhetorical appeals across disparate race-gender groups. This intersectional lens allows for a
critical reassessment of long-standing assumptions in the discipline, particularly those
formulated on datasets with limited racial diversity. For instance, the preconceived notion that
all congressional members of the same political party face a uniform set of constraints is

206
debunked by discernible patterns in emotional appeals among subordinate RGIDs within the
same party. Both Republican and Democratic MCs showed variances between dominant and
marginalized social identity groups and their white male co-partisan elites, all typically having a
more constrained and risk averse approach in comparison to the dominant group.
Intersectionality is a comprehensive approach that considers the spectrum of contexts and
identities that can identify when individual identity constraints or combined identity constraints
become significant. For example, challenger candidates who have disadvantaged race or gender
identities represent an instance where disadvantage in either dimension results in constraints.
However, when a challenger wins a seat for the Democratic Party, the racial identity becomes the
primary constraining factor against negative sentiment. If the Democratic MC is an
underrepresented nonwhite racial minority, then gender may apply further constraints for the
MC—a discrepancy not evident in white Democratic MCs.
As Smooth posited in 2011, the application of intersectionality is challenging and intricate
but vital for capturing the nuanced experiences of underrepresented RGIDs. Traditional
assumptions, such as political actors operating from a unified strategic framework simply due to
their shared ambition for office, are problematized by my findings. These reveal discernible intraparty differences, emphasizing the profound impact of race and gender intersectionality that
cannot be adequately captured by standard analytical techniques focused solely on institutional
or partisan dynamics.
Importantly, taking an intersectional approach that exposes differences in deployment of
negative and positive valence emotions subsequently reveals possible variations in strategy
preferentially taken by different race gender groups, such as fear rhetoric most frequently used
by white women. Positive valence emotional analysis may also alternative strategies that

207
constrained intersectional identities may undertake. Anticipation appeals made by Democratic
BIPOC women represent such a strategy that is also a prime area for future research.
My study exposes the added complexities unconventional candidates face post-election,
as they navigate the latent biases and stereotypes perpetuated by dominant structures of
privilege. Although my research may not encompass all intersecting identities, it underscores the
urgency for continued inquiry into the complex interplay of social identities like race and gender
in shaping political discourse and strategy. This is particularly pertinent as efforts continue to
craft a governmental landscape that authentically reflects its diverse constituency—a
transformation that has been lamentably slow in materializing.
Limitations and Future Directions
While my robust dataset and analytical framework provide substantial contributions to
the understanding of political communication, it is imperative to recognize the study's
limitations. One noteworthy constraint pertains to the scope of the investigation, particularly
concerning how underrepresented candidates interpret their risk environment. Although some
overlap exists in how nonwhite racial minorities and white women perceive risk within specific
political contexts, the evaluation of risk for various stimuli varies substantially across these
diverse race-gender groups. Such heterogeneous effects (or varying baselines) necessitate further
exploration in order to offer a more nuanced understanding of candidate behavior. Future studies
should aim to identify particular factors that modulate risk perception across different racial and
gender groups. This would serve to contextualize the diverse risk assessments that candidates
make in varying political settings.
Additionally, the level of political office introduces another layer of complexity to the
calculus of risk-taking in public messaging. Previous research indicates that women running for

208
presidential office are subjected to greater scrutiny compared to those vying for congressional
seats (Conroy et al. 2015; Falk 2010; Lawrence and Rose 2010). This suggests that the stakes
associated with adopting riskier communication strategies may differ markedly depending on
the political office in question. Such nuanced distinctions between various levels of office further
underscore the need for tailored approaches in future research.
Another limitation is the study's limited capacity to assess the constituent make-up of the
candidates' electoral districts. My analysis operates under the assumption that candidates
primarily address constituents who share their partisan affiliations. However, this assumption
may not hold consistently in terms of racial composition. For example, MCs of color, like Lauren
Underwood (D-IL 14th District), do not necessarily represent minority-majority districts, with a
similar corollary for white MCs. Consequently, a dissonance between a candidate's racial identity
and the racial majority of their district could introduce variability in risk perceptions and
constraints. Future analyses should consider whether an MC is representing a white-majority or
minority-majority district, to explore if racial outgroup status significantly alters the perceived
risk landscape, sufficiently to offset the benefits of partisan ingroup signaling.
My study pays considerable attention to evaluating what elites say on Twitter and draws
inferences from tendencies in their communication behavior, however there is no direct attention
in assessing the impacts of such emotive rhetoric. Specifically, my study does not delve into the
audience's responses to these appeals. Prior research suggests that public reactions to negative
sentiment vary significantly when the message is relayed by a female versus a male candidate
(Macdonald et al. 2022). Further, the reception of the message can be influenced by racial priming,
where public opinion and policy support are influenced by the racial and ethnic cues contained
within the news content (Hua and Jamieson 2021). To assess the efficacy of these communicative
strategies, future research could analyze macro-level electoral outcomes or employ experimental

209
survey methodologies. Such studies could focus on literature that anticipates members' strategic
incentives to formulate rhetoric that “runs against Congress,” thereby exploring the direct
impacts of these messaging strategies on election outcomes.
While the analytical focus of this dissertation encompasses a diverse set of negative and
positive discrete emotions in representatives’ strategic rhetoric, it does not fully address how and
when messages may hit upon more than one emotional note, as they often do in political ads and
campaigns. Additionally, the study does not delve into the nuanced intentions behind appeals to
specific valence emotions as they may be used to scaffold broader rhetorical ploys such as how a
reference to an oppositional political figure can appeal to constituent anger and increase
engagement to subsequently raise issue awareness and set agendas (van Kessel et al. 2020). This
also leads to the idea that emotions expressed to an outgroup audience may not produce a
response congruent with how in group members may respond (Phoenix 2020), which beckons
the question—are there variances in how underrepresented candidates specifically deploy
different valence emotions.
The avenue for future research is extensive. There is a need to explore the interaction
between emotive and issue rhetoric, including how references to current political figures, like the
President, may impact the tone and reception of the message. Given that my study identifies
membership to the opposition party as a significant predictor of anger, incorporating this
relationship into methodologies to detect appeals to anger could be highly insightful.
Beyond this, there's ample room for examining intersectional variations in risk-taking
strategies. Incumbents may find issue priming, especially on high-salience issues, particularly
risky, given their natural inclination to minimize campaign visibility (Druckman, Kifer, and
Parkin 2009). Moreover, the scholarly tradition exploring the role of moral foundations in shaping
political ideology (Graham, Haidt, and Nosek 2009; Ekins and Haidt 2016; Clifford and Jerit 2013)

210
highlights the need for a deeper investigation into how violations of these foundations impact
emotional sensitivities (Wagemans, Brandt, and Zeelenberg 2018; Molho et al. 2017; Landy and
Piazza 2019). Understanding the interplay between RGID factors, like race and gender, and public
reception of these emotive strategies could illuminate how systems of privilege potentially skew
the accountability structure in democratic representation.
Future research should aim to integrate a more holistic textual analysis framework that
encompasses sentiment, morality, issues, and intentions. Given the role of oppositional party
membership as a strong predictor of anger in the current study, forthcoming inquiries could
evaluate its potential utility in sentiment analysis models for tweet content. Additionally,
explorations into intersectionality and risk-taking behavior merit attention, specifically regarding
high-salience issues that incumbents typically steer clear of (Druckman, Kifer, and Parkin 2009).
Moral foundations theory has generated a rich scholarly tradition (Graham et al. 2009;
Gray and Keeney 2015; Graham 2015), which offers frameworks for understanding the
ideological underpinnings of political support (Ekins and Haidt 2016; MacWilliams 2016) and the
emotional responses it elicits (Wagemans et al. 2018; Mohlo et al. 2017; Landy and Piazza 2019).
Expanding on this, future research could delve into the intricacies of morality and ideology as
they intersect with emotional valence, particularly disgust, which emerged as a significant
variable in the current study.
Among the sampling limitations of this study was an underrepresentation of particular
candidates and incumbents, especially with regard to Republican women and candidates of color
(BIPOC). Furthermore, the study could benefit from a more nuanced exploration of intersectional
identities, particularly as they intersect with risk-taking strategies such as the employment of
disgust-associated rhetoric. Emerging research links the emotion of disgust, alongside fear and
surprise, to the dissemination of false stories on social media platforms (Vosoughi, Roy, and Aral

211
2018). Understanding how public perceptions differ when elected officials employ such tactics
could further elucidate the ways in which privilege shapes democratic accountability.
On the methodological front, the data collection was constrained by its focus on Twitter,
albeit the platform's significance in the realm of political communication. Yet, the ever-evolving
landscape of digital platforms necessitates an expansion of this research focus to include new
platforms and a global comparative perspective. During this dissertation, notable shifts in the
Twitter platform, now rebranded as 'X,' introduced technical changes that may influence the
reach and impact of political messages. Continued research is needed to assess the enduring
relevance of existing strategic approaches.
Complementary methodologies, such as qualitative interviews, could offer a more
nuanced understanding of electoral threat perceptions and strategic responses among
minoritized candidates. This could be particularly informative when analyzed in tandem with
existing quantitative models, offering a fuller and more textured comprehension of the strategic
considerations that underlie public messaging and appeals. Qualitative insights can not only
enrich the existing framework but also offer candidates' unique perspectives on the 'risk
environment,' potentially reframing the way we understand these strategic choices.
Despite these limitations, the research contributes a nuanced understanding of how social
identities intersect with political communication, thereby enriching the existing corpus of
political communication studies.
Concluding Thoughts
My findings espouse a significant divide for appealing to negative emotions by
historically marginalized groups whose perpetuation through persistent and prolific negative
rhetoric also serves to sustain representational inequities. The prospect of mitigating these

212
disparities may seem bleak through this lens. However, the dynamism of social media platforms
serves as a testament to the rapid malleability of societal norms, contributing to tangible shifts
like the increased racial and gender diversity in Congress and the historic election of a Black
President of the United States.
This dissertation has laid bare systemic inequities that adversely affect historically
underrepresented political elites, thereby curtailing their ability to drive meaningful change.
Given that societal transformations are ultimately rooted in individual shifts in perception and
behavior, it is my aspiration that this work enlightens readers about the outsized challenges that
unconventional candidates encounter. I hope it urges them to critically examine their own biases
when evaluating this emerging cohort of political leaders.
Additionally, this dissertation aims to serve as a foundational text for scholars in
elucidating how tenets from race and ethnic politics can be applied constructively within the
domain of political communication. I extend an invitation to scholars within the field to be
audacious in incorporating meaningful applications of intersectionality, encompassing race,
gender, and additional social identities. While applying intersectionality and other novel
concepts might get messy, continued progress within our discipline hinges on the will and
determination to engage in this substantive work.

213
References
Aarøe, Lene, Michael Bang Petersen, and Kevin Arceneaux. 2020. “The Behavioral Immune
System Shapes Partisan Preferences in Modern Democracies: Disgust Sensitivity Predicts
Voting for Socially Conservative Parties.” Political Psychology 41(6): 1073–91.
Abelson, Robert P., Donald R. Kinder, Mark D. Peters, and Susan T. Fiske. 1982. “Affective and
Semantic Components in Political Person Perception.” Journal of Personality and Social
Psychology 42(4): 619–30.
Abramowitz, Alan I. 2018. The Great Alignment: Race, Party Transformation, and the Rise of Donald
Trump. New Haven: Yale University Press.
Abrams, Dominic, and Michael A. Hogg. 1988. “Comments on the Motivational Status of SelfEsteem in Social Identity and Intergroup Discrimination.” European Journal of Social
Psychology 18(4): 317–34.
Alamillo, Rudy, and Loren Collingwood. 2017. “Chameleon Politics: Social Identity and Racial
Cross-over Appeals.” Politics, Groups, and Identities 5(4): 533–60.
Albertson, Bethany, Lindsay Dun, and Shana Kushner Gadarian. 2020. “The Emotional Aspects
of Political Persuasion.” In The Oxford Handbook of Electoral Persuasion, eds. Elizabeth Suhay,
Bernard Grofman, and Alexander H. Trechsel. Oxford University Press, 168–83.
https://academic.oup.com/edited-volume/28311/chapter/215011751 (May 3, 2023).
Alexander, Amy C. 2012. “Change in Women’s Descriptive Representation and the Belief in
Women’s Ability to Govern: A Virtuous Cycle.” Politics & Gender 8(4): 437–64.
Alford, John R., and John R. Hibbing. 1981. “Increased Incumbency Advantage in the House.”
The Journal of Politics 43(4): 1042–61.
Alvarez, Linda. 2020. “No Safe Space: Neoliberalism and the Production of Violence in the Lives
of Central American Migrants.” Journal of Race, Ethnicity, and Politics 5(1): 4–36.
Amengay, Abdelkarim, and Daniel Stockemer. 2019. “The Radical Right in Western Europe: A
Meta-Analysis of Structural Factors.” Political Studies Review 17(1): 30–40.
American Psychological Association. 2021. “Inclusive Language Guidelines.”
https://www.apa.org/about/apa/equitydiversity-inclusion/language-guidelines.pdf.

214
Andersen, Margaret L. 2003. “Whitewashing Race: A Critical Perspective on Whiteness.” In White
out: The Continuing Significance of Racism, eds. Ashley W. Doane and Eduardo Bonilla-Silva.
New York, NY: Routledge, 21–34. http://journals.sagepub. com/doi/10.1177/0891243211-
407491 (October 16, 2023).
Ansolabehere, Stephen, David Brady, and Morris Fiorina. 1992. “The Vanishing Marginals and
Electoral Responsiveness.” British Journal of Political Science 22(1): 21–38.
Ansolabehere, Stephen, and Shanto Iyengar. 1995. Going Negative: How Attack Ads Shrink and
Polarize the Electorate. New York, NY: Free Press.
Ansolabehere, Stephen, and Shiro Kuriwaki. 2022. “Congressional Representation:
Accountability from the Constituent’s Perspective.” American Journal of Political Science
66(1): 123–39.
Ansolabehere, Stephen, James M. Snyder, and Charles Stewart. 2001. “Candidate Positioning in
U.S. House Elections.” American Journal of Political Science 45(1): 136–59.
Arbour, Brian. 2014. “Issue Frame Ownership: The Partisan Roots of Campaign Rhetoric.”
Political Communication 31(4): 604–27.
Arceneaux, Kevin. 2017. “Anxiety Reduces Empathy Toward Outgroup Members But Not
Ingroup Members.” Journal of Experimental Political Science 4(1): 68–80.
Auter, Zachary J., and Jeffrey A. Fine. 2016. “Negative Campaigning in the Social Media Age:
Attack Advertising on Facebook.” Political Behavior 38(4): 999–1020.
———. 2018. “Social Media Campaigning: Mobilization and Fundraising on Facebook.” Social
Science Quarterly 99(1): 185–200.
Averill, James R. 1982. Anger and Aggression. New York, NY: Springer New York.
http://link.springer.com/10.1007/978-1-4612-5743-1 (October 14, 2023).
Bäck, Hanna, Royce Carroll, Emma Renström, and Alexander Ryan. 2023. “Elite Communication
and Affective Polarization among Voters.” Electoral Studies 84: 102639.
Banda, Kevin K. 2016. “Issue Ownership, Issue Positions, and Candidate Assessment.” Political
Communication 33(4): 651–66.
Banda, Kevin K., and Erin C. Cassese. 2022. “Hostile Sexism, Racial Resentment, and Political
Mobilization.” Political Behavior 44(3): 1317–35.
Banks, Antoine J. 2014. “The Public’s Anger: White Racial Attitudes and Opinions Toward Health
Care Reform.” Political Behavior 36(3): 493–514.
Barberá, Pablo et al. 2019. “Who Leads? Who Follows? Measuring Issue Attention and Agenda
Setting by Legislators and the Mass Public Using Social Media Data.” American Political
Science Review 113(4): 883–901.

215
Bauer, Nichole M. 2013. “Rethinking Stereotype Reliance: Understanding the Connection
between Female Candidates and Gender Stereotypes.” Politics and the Life Sciences 32(1): 22–
42.
———. 2015. “Who Stereotypes Female Candidates? Identifying Individual Differences in
Feminine Stereotype Reliance.” Politics, Groups, and Identities 3(1): 94–110.
———. 2017. “The Effects of Counterstereotypic Gender Strategies on Candidate Evaluations.”
Political Psychology 38(2): 279–95.
———. 2019. “Gender Stereotyping in Political Decision Making.” In Oxford Research Encyclopedia
of Politics, https://oxfordre.com/politics/display/10.1093/acrefore/9780190228637. -
001.0001/acrefore-9780190228637-e-772 (October 6, 2023).
———. 2020. The Qualifications Gap: Why Women Must Be Better than Men to Win Political Office.
Cambridge: Cambridge University Press. https://www.cambridge.org/core/books-
/qualifications-gap/2F2629A699402762D0E66E3F02C7394C (October 14, 2023).
Bawden, David, and Lyn Robinson. 2020. “Information Overload: An Introduction.” In Oxford
Research Encyclopedia of Politics, Oxford University Press. https:// -
oxfordre.com/politics/view/10.1093/acrefore/9780190228637.001.0001/acrefore9780190228637-e-1360 (July 20, 2023).
Bennett, W. Lance, and Alexandra Segerberg. 2013. The Logic of Connective Action: Digital Media
and the Personalization of Contentious Politics. Cambridge: Cambridge University Press.
http://ebooks.cambridge.org/ref/id/CBO9781139198752 (August 18, 2020).
Ben-Nun Bloom, Pazit. 2014. “Disgust, Harm, and Morality in Politics: Disgust, Harm, and
Morality in Politics.” Political Psychology 35(4): 495–513.
Benoit, William L., and Glenn J. Hansen. 2004. “Presidential Debate Watching, Issue Knowledge,
Character Evaluation, and Vote Choice.” Human Communication Research 30(1): 121–44.
Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1986. Voting: A Study of Opinion
Formation in a Presidential Campaign. Chicago, IL: University of Chicago Press.
https://press.uchicago.edu/ucp/books/book/chicago/V/bo3616092.html (October 13,
2023).
Berinsky, Adam J. et al. 2011. “Sex and Race: Are Black Candidates More Likely to Be
Disadvantaged by Sex Scandals?” Political Behavior 33(2): 179–202.
Blumenthal, Sidney. 1982. The Permanent Campaign. Simon and Schuster.
Bonilla-Silva, Eduardo. 2006. Racism without Racists: Color-Blind Racism and the Persistence of Racial
Inequality in the United States. Rowman & Littlefield Publishers.
Brader, Ted. 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.

216
———. 2006. “Campaigning for Hearts and Minds: How Emotional Appeals in Political Ads
Work.” Bibliovault OAI Repository, the University of Chicago Press 4.
———. 2020. “Campaigning for Hearts and Minds: How Emotional Appeals in Political Ads
Work.” In Campaigning for Hearts and Minds, University of Chicago Press.
https://www.degruyter.com/document/doi/10.7208/9780226788302/html (August 26,
2023).
Brady, William J. et al. 2017. “Emotion Shapes the Diffusion of Moralized Content in Social
Networks.” Proceedings of the National Academy of Sciences 114(28): 7313–18.
Brescoll, Victoria L. 2016. “Leading with Their Hearts? How Gender Stereotypes of Emotion Lead
to Biased Evaluations of Female Leaders.” The Leadership Quarterly 27(3): 415–28.
Brisbane, Laura, Whitney Hua, and Thomas Jamieson. 2023. “Morality and the Glass Ceiling:
How Elite Rhetoric Reflects Gendered Strategies and Perspectives.” Politics & Gender: 1–35.
Broockman, David E., and Daniel M. Butler. 2017. “The Causal Effects of Elite Position-Taking on
Voter Attitudes: Field Experiments with Elite Communication.” American Journal of Political
Science 61(1): 208–21.
Brooks, Deborah Jordan. 2011. “Testing the Double Standard for Candidate Emotionality: Voter
Reactions to the Tears and Anger of Male and Female Politicians.” The Journal of Politics
73(2): 597–615.
Brooks, Deborah Jordan, and John G. Geer. 2007. “Beyond Negativity: The Effects of Incivility on
the Electorate.” American Journal of Political Science 51(1): 1–16.
Brown, Mark B. 2020. “James Baldwin and the Politics of White Identity.” Contemporary Political
Theory. http://link.springer.com/10.1057/s41296-020-00401-9 (September 7, 2020).
Brunel, Frédéric F., and Michelle R. Nelson. 2000. “Explaining Gendered Responses to ‘Help-Self’
and ‘Help-Others’ Charity Ad Appeals: The Mediating Role of World-Views.” Journal of
Advertising 29(3): 15–28.
Bullock, John G. 2011. “Elite Influence on Public Opinion in an Informed Electorate.” American
Political Science Review 105(03): 496–515.
———. 2020. “Party Cues.” In The Oxford Handbook of Electoral Persuasion, eds. Elizabeth Suhay,
Bernard Grofman, and Alexander H. Trechsel. Oxford University Press, 128–50.
https://academic.oup.com/edited-volume/28311/chapter/215010740 (May 29, 2023).
Burrell, Barbara. 1992. “Women Candidates in Open-Seat Primaries for the U. S. House: 1968-
1990.” Legislative Studies Quarterly 17(4): 493–508.
———. 2014. Gender in Campaigns for the U.S. House of Representatives. eds. Susan J. Carroll and
Kira Sanbonmatsu. University of Michigan Press. https://www.jstor.org/stable/10.3998/-
mpub.213944 (October 15, 2023).

217
Burrell, Barbara C. 2005. “Gender, Presidential Elections and Public Policy: Making Women’s
Votes Matter.” Journal of Women, Politics & Policy 27(1–2): 31–50.
Bystrom, Dianne. 2006. “Gender and Campaign Communication: TV Ads, Web Sites, and Media
Coverage.” https://escholarship.org/uc/item/3nh303r0 (October 7, 2023).
Cacioppo, John T., Wendi L. Gardner, and Gary G. Berntson. 1997. “Beyond Bipolar
Conceptualizations and Measures: The Case of Attitudes and Evaluative Space.” Personality
and Social Psychology Review 1(1): 3–25.
———. 1999. “The Affect System Has Parallel and Integrative Processing Components: Form
Follows Function.” Journal of Personality and Social Psychology 76(5): 839–55.
Carver, Charles S., and Michael F. Scheier. 1990. “Origins and Functions of Positive and Negative
Affect: A Control-Process View.” Psychological Review 97(1): 19–35.
Casas, Andreu, and David Morar. 2019. “Different Channel, Same Strategy? Filling Empirical
Gaps in Congress Literature.” https://papers.ssrn.com/abstract=3395307 (October 10,
2023).
Cassese, Erin C., and Mirya R. Holman. 2018. “Party and Gender Stereotypes in Campaign
Attacks.” Political Behavior 40(3): 785–807.
Ceccobelli, Diego. 2018. “Not Every Day Is Election Day: A Comparative Analysis of Eighteen
Election Campaigns on Facebook.” Journal of Information Technology & Politics 15(2): 122–41.
Chadwick, Andrew. 2017. The Hybrid Media System: Politics and Power. Second Edition. New York,
NY: Oxford University Press.
Choi, Virginia K., Snehesh Shrestha, Xinyue Pan, and Michele J. Gelfand. 2022. “When Danger
Strikes: A Linguistic Tool for Tracking America’s Collective Response to Threats.”
Proceedings of the National Academy of Sciences 119(4): e2113891119.
Chong, Dennis, and James N. Druckman. 2007. “Framing Public Opinion in Competitive
Democracies.” The American Political Science Review 101(4): 637–55.
Citrin, Jack, Donald Philip Green, and David O. Sears. 1990. “White Reactions to Black
Candidates: When Does Race Matter?” The Public Opinion Quarterly 54(1): 74–96.
Civettini, Andrew J. W., and David P. Redlawsk. 2009. “Voters, Emotions, and Memory.” Political
Psychology 30(1): 125–51.
Clifford, Scott, and Jennifer Jerit. 2013. “How Words Do the Work of Politics: Moral Foundations
Theory and the Debate over Stem Cell Research.” The Journal of Politics 75(3): 659–71.
———. 2018. “Disgust, Anxiety, and Political Learning in the Face of Threat.” American Journal of
Political Science 62(2): 266–79.
Cole, Elizabeth R. 2009. “Intersectionality and Research in Psychology.” American Psychologist
64(3): 170–80.

218
Collins, Patricia Hill. 2000. Black Feminist Thought | Knowledge, Consciousness, and the Politics of
Empowerment. 2nd Edition. New York: Routledge. https://www-taylorfranciscom.libproxy2.usc.edu/pdfviewer/ (October 13, 2023).
Conroy, Meredith, Sarah Oliver, Ian Breckenridge-Jackson, and Caroline Heldman. 2015. “From
Ferraro to Palin: Sexism in Coverage of Vice Presidential Candidates in Old and New
Media.” Politics, Groups, and Identities 3(4): 573–91.
Cormack, Lindsey. 2016. “Gender and Vote Revelation Strategy in the United States Congress.”
Journal of Gender Studies 25(6): 626–40.
Costa, Mia. 2021. “Ideology, Not Affect: What Americans Want from Political Representation.”
American Journal of Political Science 65(2): 342–58.
Cowie, Roddy, and Randolph R. Cornelius. 2003. “Describing the Emotional States That Are
Expressed in Speech.” Speech Communication 40(1): 5–32.
Cox, Gary W., and Jonathan N. Katz. 1996. “Why Did the Incumbency Advantage in U.S. House
Elections Grow?” American Journal of Political Science 40(2): 478–97.
Crawford, Jarret T. 2017. “Are Conservatives More Sensitive to Threat than Liberals? It Depends
on How We Define Threat and Conservatism.” Social Cognition 35(4): 354–73.
Crenshaw, Kimberle. 1989. “Demarginalizing the Intersection of Race and Sex: A Black Feminist
Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Politics.”
University of Chicago Legal Forum 1989(1): 139–67.
Crespin, Sarina Rhinehart, Michael H. 2020. “Women in Congress.” In New Directions in
Congressional Politics, Routledge.
Cryer, Jennifer. 2019. “Navigating Identity in Campaign Messaging: The Influence of Race &
Gender on Strategy in U.S. Congressional Elections.”
Cryer, Jennifer Elaine. 2020. “Navigating Identity: How Candidate Identities Influence Campaign
Communication, Electoral Outcomes, and Voter Perceptions.” PhD Thesis. Stanford
University.
Cunningham, Emily, Catherine A. Forestell, and Cheryl L. Dickter. 2013. “Induced Disgust
Affects Implicit and Explicit Responses toward Gay Men and Lesbians.” European Journal of
Social Psychology 43(5): 362–69.
Davidson, Richard J. 1993. “The Neuropsychology of Emotion and Affective Style.” In Handbook
of Emotions, New York, NY, US: The Guilford Press, 143–54.
Derryberry, Douglas. 1991. “The Immediate Effects of Positive and Negative Feedback Signals.”
Journal of Personality and Social Psychology 61(2): 267–78.
DeSteno, David, Nilanjana Dasgupta, Monica Y. Bartlett, and Aida Cajdric. 2004. “Prejudice From
Thin Air: The Effect of Emotion on Automatic Intergroup Attitudes.” Psychological Science
15(5): 319–24.

219
Diener, Ed, and Carol Diener. 1996. “Most People Are Happy.” Psychological Science 7(3): 181–85.
van Dijk, Teun A. 2006. “Discourse and Manipulation.” Discourse & Society 17(3): 359–83.
Dillard, James, and Michael Pfau. 2002. The Persuasion Handbook: Developments in Theory and
Practice. 2455 Teller Road, Thousand Oaks California 91320 United States: SAGE
Publications, Inc. https://sk.sagepub.com/reference/hdbk_persuasion (October 15, 2023).
Dittmar, Kelly. 2015. Navigating Gendered Terrain: Stereotypes and Strategy in Political Campaigns.
Philadelphia: Temple University Press.
———. 2020. “Gendered Aspects of Political Persuasion in Campaigns.” In The Oxford Handbook
of Electoral Persuasion, eds. Elizabeth Suhay, Bernard Grofman, and Alexander H. Trechsel.
Oxford University Press, 0. https://doi.org/10.1093/oxfordhb/9780190860806.013.37 (May
31, 2023).
Doan, Alesha E., and Donald P. Haider-Markel. 2010. “The Role of Intersectional Stereotypes on
Evaluations of Gay and Lesbian Political Candidates.” Politics & Gender 6(1): 63–91.
Doane, Ashley W., and Eduardo Bonilla-Silva, eds. 2003. White out: The Continuing Significance of
Racism. New York: Routledge.
Dolan, Kathleen. 2010. “The Impact of Gender Stereotyped Evaluations on Support for Women
Candidates.” Political Behavior 32(1): 69–88.
———. 2014. “Gender Stereotypes, Candidate Evaluations, and Voting for Women Candidates:
What Really Matters?” Political Research Quarterly 67(1): 96–107.
Domke, David et al. 2000. “Elite Messages and Source Cues: Moving Beyond Partisanship.”
Political Communication 17(4): 395–402.
Dovi, Suzanne. 2002. “Preferable Descriptive Representatives: Will Just Any Woman, Black, or
Latino Do?” American Political Science Review 96(4): 729–43.
Downs, Anthony. 1957. “An Economic Theory of Political Action in a Democracy.” Journal of
Political Economy 65(2): 135–50.
Druckman, James N. 2001. “On the Limits of Framing Effects: Who Can Frame?” The Journal of
Politics 63(4): 1041–66.
Druckman, James N., Lawrence R. Jacobs, and Eric Ostermeier. 2004. “Candidate Strategies to
Prime Issues and Image.” The Journal of Politics 66(4): 1180–1202.
Druckman, James N., Martin J. Kifer, and Michael Parkin. 2009. “Campaign Communications in
U.S. Congressional Elections.” American Political Science Review 103(3): 343–66.
Druckman, James N, Martin J Kifer, and Michael Parkin. 2010. “Timeless Strategy Meets New
Medium: Going Negative on Congressional Campaign Web Sites, 2002–2006.” Political
Communication 27(1): 88–103.

220
Druckman, James N., Martin J. Kifer, and Michael Parkin. 2020. “Campaign Rhetoric and the
Incumbency Advantage.” American Politics Research 48(1): 22–43.
Druckman, James N., and Thomas J. Leeper. 2012. “Learning More from Political Communication
Experiments: Pretreatment and Its Effects.” American Journal of Political Science 56(4): 875–
96.
Druckman, James N., and Rose McDermott. 2008. “Emotion and the Framing of Risky Choice.”
Political Behavior 30(3): 297–321.
Druckman, James N., Erik Peterson, and Rune Slothuus. 2013. “How Elite Partisan Polarization
Affects Public Opinion Formation.” American Political Science Review 107(1): 57–79.
Edelman, Murray. 1988. Constructing the Political Spectacle. Chicago, IL: University of Chicago
Press. https://press.uchicago.edu/ucp/books/book/chicago/C/bo5948882.html (July 20,
2023).
Egloff, Boris et al. 2003. “Facets of Dynamic Positive Affect: Differentiating Joy, Interest, and
Activation in the Positive and Negative Affect Schedule (PANAS).” Journal of Personality and
Social Psychology 85(3): 528–40.
Ekman, Klaus R. Scherer, Paul, ed. 2014. Approaches To Emotion. New York: Psychology Press.
Elder, Laurel. 2012. “The Partisan Gap Among Women State Legislators.” Journal of Women,
Politics & Policy 33(1): 65–85.
Ellemers, Naomi. 2017. “Gender Stereotypes.”
Ensley, Michael J., Michael W. Tofias, and Scott De Marchi. 2009. “District Complexity as an
Advantage in Congressional Elections.” American Journal of Political Science 53(4): 990–1005.
Erikson, Robert S. 1971. “The Advantage of Incumbency in Congressional Elections.” Polity 3(3):
395–405.
Erikson, Robert S., Michael MacKuen, and James A. Stimson. 2002. The Macro Polity. New York:
Cambridge University Press.
Evans, Heather K., and Jennifer Hayes Clark. 2016. “‘You Tweet Like a Girl!’: How Female
Candidates Campaign on Twitter.” American Politics Research 44(2): 326–52.
Evans, Heather K., Victoria Cordova, and Savannah Sipole. 2014. “Twitter Style: An Analysis of
How House Candidates Used Twitter in Their 2012 Campaigns.” PS: Political Science &
Politics 47(2): 454–62.
Falk, Erika. 2010. Women for President: Media Bias in Nine Campaigns. University of Illinois Press.
https://www.jstor.org/stable/10.5406/j.ctt4cgg2h (October 14, 2023).
Fenno, Richard F. 1978. Home Style: House Members in Their Districts. Longman.
Ferejohn, John A. 1977. “On the Decline of Competition in Congressional Elections.” The American
Political Science Review 71(1): 166–76.

221
Ferrara, Emilio, and Zeyao Yang. 2015. “Quantifying the Effect of Sentiment on Information
Diffusion in Social Media.” PeerJ Computer Science 1: e26.
Finkel, Eli J. et al. 2020. “Political Sectarianism in America.” Science 370(6516): 533–36.
Fiorina, Morris P. 1977. “The Case of the Vanishing Marginals: The Bureaucracy Did It.” The
American Political Science Review 71(1): 177–81.
———. 1989. Congress, Keystone of the Washington Establishment. 2nd ed. New Haven: Yale
University Press.
Fischer, Agneta H. et al. 2004. “Gender and Culture Differences in Emotion.” Emotion 4(1): 87–94.
Fischer, Agneta, and Marianne LaFrance. 2015. “What Drives the Smile and the Tear: Why
Women Are More Emotionally Expressive Than Men.” Emotion Review 7(1): 22–29.
Flores, René D. 2018. “Can Elites Shape Public Attitudes Toward Immigrants?: Evidence from the
2016 U.S. Presidential Election.” Social Forces 96(4): 1649–90.
Foss, Sonja K., and Kimberly C. Elliott. 2016. “Acting to Alter Privilege:By Maintaining the
Structural-Performative Paradox.” Women and Language 38(2).
Fowler, Anthony. 2016. “What Explains Incumbent Success? Disentangling Selection on Party,
Selection on Candidate Characteristics, and Office-Holding Benefits.” Quarterly Journal of
Political Science 11(3): 313–38.
Fowler, Anthony, and Andrew B. Hall. 2018. “Do Shark Attacks Influence Presidential Elections?
Reassessing a Prominent Finding on Voter Competence.” The Journal of Politics 80(4): 1423–
37.
Fox, Richard. 1997. Gender Dynamics in Congressional Elections. 2455 Teller Road, Thousand
Oaks California 91320 United States: SAGE Publications, Inc. https://sk.sagepub. -
com/books/gender-dynamics-in-congressional-elections (October 13, 2023).
Fredrickson, Barbara L. 2001. “The Role of Positive Emotions in Positive Psychology: The
Broaden-and-Build Theory of Positive Emotions.” American Psychologist 56(3): 218–26.
———. 2003. “The Value of Positive Emotions: The Emerging Science of Positive Psychology Is
Coming to Understand Why It’s Good to Feel Good.” American Scientist 91(4): 330–35.
———. 2004. “The Broaden–and–Build Theory of Positive Emotions” eds. F. A. Huppert, N.
Baylis, and B. Keverne. Philosophical Transactions of the Royal Society of London. Series B:
Biological Sciences 359(1449): 1367–77.
———. 2013. “Updated Thinking on Positivity Ratios.” American Psychologist 68(9): 814–22.
Fredrickson, Barbara L., and Thomas Joiner. 2018. “Reflections on Positive Emotions and Upward
Spirals.” Perspectives on Psychological Science 13(2): 194–99.
Fredrickson, Barbara L., and Robert W. Levenson. 1998. “Positive Emotions Speed Recovery from
the Cardiovascular Sequelae of Negative Emotions.” Cognition and Emotion 12(2): 191–220.

222
Fredrickson, Barbara L., Roberta A. Mancuso, Christine Branigan, and Michele M. Tugade. 2000.
“The Undoing Effect of Positive Emotions.” Motivation and Emotion 24(4): 237–58.
Fridkin, Kim, and Patrick Kenney. 2009. “The Role of Gender Stereotypes in U.S. Senate
Campaigns.” Politics & Gender 5: 301–24.
Fridkin, Kim L., and Patrick J. Kenney. 2014. The Changing Face of Representation: The Gender of U.S.
Senators and Constituent Communications. University of Michigan Press. https:// -
www.jstor.org/stable/10.3998/mpub.6200859 (October 7, 2023).
Frijda, Nico. 2005. “Emotion Experience.” Cognition & Emotion 19(4): 473–97.
Frijda, Nico H. 1986. The Emotions. Cambridge ; New York : Paris: Cambridge University Press ;
Editions de la Maison des sciences de l’homme.
Frijda, Nico H., Peter Kuipers, and Elisabeth ter Schure. 1989. “Relations among Emotion,
Appraisal, and Emotional Action Readiness.” Journal of Personality and Social Psychology
57(2): 212–28.
Frijda, Nico H., and Batja Mesquita. 1994. “The Social Roles and Functions of Emotions.” In
Emotion and Culture: Empirical Studies of Mutual Influence, Washington, DC, US: American
Psychological Association, 51–87.
Fulton, Sarah A., and Kostanca Dhima. 2021. “The Gendered Politics of Congressional Elections.”
Political Behavior 43(4): 1611–37.
Gabel, Matthew, and Kenneth Scheve. 2007. “Estimating the Effect of Elite Communications on
Public Opinion Using Instrumental Variables.” American Journal of Political Science 51(4):
1013–28.
Gainous, Jason, and Kevin M. Wagner. 2014. Tweeting to Power: The Social Media Revolution in
American Politics. New York: Oxford University Press.
Garcia-Prieto, Patricia, Erwan Bellard, and Susan C. Schneider. 2003. “Experiencing Diversity,
Conflict, and Emotions in Teams.” Applied Psychology 52(3): 413–40.
Geer, John G., and Lynn Vavreck. 2014. “Negativity, Information, and Candidate PositionTaking.” Political Communication 31(2): 218–36.
Geer, John Gray. 2006. In Defense of Negativity: Attack Ads in Presidential Campaigns. Nachdr.
Chicago: University of Chicago Press.
Gershon, Sarah. 2012. “When Race, Gender, and the Media Intersect: Campaign News Coverage
of Minority Congresswomen.” Journal of Women, Politics & Policy 33(2): 105–25.
Gershon, Sarah Allen. 2008. “Communicating Female and Minority Interests Online: A Study of
Web Site Issue Discussion among Female, Latino, and African American Members of
Congress.” The International Journal of Press/Politics 13(2): 120–40.

223
Gerstlé, Jacques, and Alessandro Nai. 2019. “Negativity, Emotionality and Populist Rhetoric in
Election Campaigns Worldwide, and Their Effects on Media Attention and Electoral
Success.” European journal of communication (London) 34(4): 410–44.
Gervais, Bryan T., Heather K. Evans, and Annelise Russell. 2020. “Fear and Loathing on Twitter:
Exploring Negative Rhetoric in Tweets During the 2018 Midterm Election.” In The Roads to
Congress 2018, eds. Sean D. Foreman, Marcia L. Godwin, and Walter Clark Wilson. Cham:
Springer International Publishing, 31–52. http://link.springer.com/10.1007/978-3-030-
19819-0_3 (August 18, 2019).
Gilens, Martin. 2001. “Political Ignorance and Collective Policy Preferences.” American Political
Science Review 95(2): 379–96.
Giner-Sorolla, Roger. 2019. “The Past Thirty Years of Emotion Research: Appraisal and Beyond.”
Cognition and Emotion 33(1): 48–54.
Goffman, Erving. 1959. The Presentation of Self in Everyday Life. Repr. New York: Bantam.
Gordon, Ann, David M. Shafie, and Ann N. Crigler. 2003. “Is Negative Advertising Effective for
Female Candidates?: An Experiment in Voters’ Uses of Gender Stereotypes.” Harvard
International Journal of Press/Politics 8(3): 35–53.
Grady, Constance. 2020. “Why the Term ‘BIPOC’ Is so Complicated, Explained by Linguists.”
Vox. https://www.vox.com/2020/6/30/21300294/bipoc-what-does-it-mean-critical-racelinguistics-jonathan-rosa-deandra-miles-hercules (September 10, 2023).
Graham, Jesse, Jonathan Haidt, and Brian A. Nosek. 2009. “Liberals and Conservatives Rely on
Different Sets of Moral Foundations.” Journal of Personality and Social Psychology 96(5): 1029–
46.
Graham, Jesse, Brian A. Nosek, and Jonathan Haidt. 2012. “The Moral Stereotypes of Liberals and
Conservatives: Exaggeration of Differences across the Political Spectrum” ed. Liane Young.
PLoS ONE 7(12): e50092.
Green, Donald P., Susan L. Goldman, and Peter Salovey. 1993. “Measurement Error Masks
Bipolarity in Affect Ratings.” Journal of Personality and Social Psychology 64(6): 1029–41.
Greene, Steven. 2004. “Social Identity Theory and Party Identification*.” Social Science Quarterly
85(1): 136–53.
Grimmer, Justin. 2013. Representational Style in Congress: What Legislators Say and Why It Matters.
Cambridge: Cambridge University Press. https://www.cambridge.org/core/books-
/representational-style-in-congress/AD32D8ABA29E78C048B41077D64BDD9A (July 20,
2023).
Groeling, Tim. 2010. When Politicians Attack: Party Cohesion in the Media. Cambridge University
Press.
Gronke, Paul. 2000. The Electorate, the Campaign, and the Office: A Unified Approach to Senate and
House Elections. Ann Arbor: University of Michigan Press.

224
Grose, Christian R., Neil Malhotra, and Robert Parks Van Houweling. 2015. “Explaining
Explanations: How Legislators Explain Their Policy Positions and How Citizens React.”
American Journal of Political Science 59(3): 724–43.
Groseclose, Tim. 2001. “A Model of Candidate Location When One Candidate Has a Valence
Advantage.” American Journal of Political Science 45(4): 862–86.
Gross, Justin H., and Kaylee T. Johnson. 2016. “Twitter Taunts and Tirades: Negative
Campaigning in the Age of Trump.” PS: Political Science & Politics 49(04): 748–54.
Gross, Kimberly. 2008. “Framing Persuasive Appeals: Episodic and Thematic Framing, Emotional
Response, and Policy Opinion.” Political Psychology 29(2): 169–92.
Grossmann, Matthew, and David A. Hopkins. 2016. Asymmetric Politics: Ideological Republicans and
Group Interest Democrats. New York, NY: Oxford University Press.
Guisinger, Alexandra, and Elizabeth N. Saunders. 2017. “Mapping the Boundaries of Elite Cues:
How Elites Shape Mass Opinion across International Issues.” International Studies Quarterly
61(2): 425–41.
Hansen, Eric R., and Sarah A. Treul. 2021. “Inexperienced or Anti-Establishment? Voter
Preferences for Outsider Congressional Candidates.” Research & Politics 8(3):
205316802110349.
Haselmayer, Martin. 2019. “Negative Campaigning and Its Consequences: A Review and a Look
Ahead.” French Politics 17(3): 355–72.
Hasen, Rick. 2023. “‘Conservatives Should Look MoreClosely at Systemic Election Reforms’
#ELB.” Election Law Blog. https://electionlawblog.org/?p=138961 (October 9, 2023).
Hatemi, Peter K., Rose McDermott, J.Michael Bailey, and Nicholas G. Martin. 2012. “The Different
Effects of Gender and Sex on Vote Choice.” Political Research Quarterly 65(1): 76–92.
Hawkesworth, Mary. 2003. “Congressional Enactments of Race-Gender: Toward a Theory of
Raced-Gendered Institutions.” The American Political Science Review 97(4): 529–50.
Hayes, Danny, and Jennifer L. Lawless. 2016. Women on the Run: Gender, Media, and Political
Campaigns in a Polarized Era. 1st ed. Cambridge University Press. https://www. -
cambridge.org/core/product/identifier/9781316336007/type/book (July 31, 2023).
Hemphill, Libby, and Andrew J. Roback. 2014. “Tweet Acts: How Constituents Lobby Congress
via Twitter.” In Proceedings of the 17th ACM Conference on Computer Supported Cooperative
Work & Social Computing, Baltimore Maryland USA: ACM, 1200–1210.
https://dl.acm.org/doi/10.1145/2531602.2531735 (October 15, 2023).
Hemphill, Libby, Annelise Russell, and Angela M. Schöpke-Gonzalez. 2021. “What Drives U.S.
Congressional Members’ Policy Attention on Twitter?” Policy & Internet 13(2): 233–56.

225
Herndon, Astead W. 2018. “The Districts Are Mostly White. The Candidates Are Not.” The New
York Times. https://www.nytimes.com/2018/07/19/us/politics/minority-candidates-
.html (July 20, 2023).
Herring, Cedric. 2009. “Does Diversity Pay?: Race, Gender, and the Business Case for Diversity.”
American Sociological Review 74(2): 208–24.
Hertel, Guido, Jochen Neuhof, Thomas Theuer, and Norbert L. Kerr. 2000. “Mood Effects on
Cooperation in Small Groups: Does Positive Mood Simply Lead to More Cooperation?”
Cognition & Emotion 14(4): 441–72.
Hess, Ursula, Sylvie Blairy, and Robert E. Kleck. 2000. “The Influence of Facial Emotion Displays,
Gender, and Ethnicity on Judgments of Dominance and Affiliation.” Journal of Nonverbal
Behavior 24(4): 265–83.
Hibbing, John R., Kevin B. Smith, Johnathan C. Peterson, and Balazs Feher. 2014. “The Deeper
Sources of Political Conflict: Evidence from the Psychological, Cognitive, and NeuroSciences.” Trends in Cognitive Sciences 18(3): 111–13.
Highton, Benjamin. 2011. “Prejudice Rivals Partisanship and Ideology When Explaining the 2008
Presidential Vote across the States.” PS: Political Science & Politics 44(3): 530–35.
Hitchon, Jacqueline C., Chingching Chang, and Rhonda Harris. 1997. “Should Women Emote?
Perceptual Bias and Opinion Change in Response to Political Ads for Candidates of
Different Genders.” Political Communication 14(1): 49–69.
Hodson, Gordon, and Kimberly Costello. 2007. “Interpersonal Disgust, Ideological Orientations,
and Dehumanization as Predictors of Intergroup Attitudes.” Psychological Science 18(8): 691–
98.
Hogg, Michael A., and Scott A. Reid. 2006. “Social Identity, Self-Categorization, and the
Communication of Group Norms.” Communication Theory 16(1): 7–30.
Hopkins, Daniel J. 2018. “The Exaggerated Life of Death Panels? The Limited but Real Influence
of Elite Rhetoric in the 2009–2010 Health Care Debate.” Political Behavior 40(3): 681–709.
Hua, Whitney, Laura Brisbane, Ann N. Crigler, and Marion R. Just. 2019. “The Moral and
Emotional Microfoundations of Populist Rhetoric in the 2016 U.S. Presidential Primaries.”
Presented at the 2019 Annual Meeting of the International Society of Political Psychology,
Lisbon, Portugal.
Hua, Whitney, and Christopher Chapp. 2020. “It’s Not What You Say, It’s Where You Say It:
Comparing Congressional Communication across Platforms.” Presented at the 2020 Annual
Meeting of the American Political Science Association. Online.
Hua, Whitney, and Thomas Jamieson. 2022. “Whose Lives Matter? Race, Public Opinion, and
Military Conflict.” Politics, Groups, and Identities: 1–23.
Hua, Whitney, and Jane Junn. 2021. “Amidst Pandemic and Racial Upheaval: Where Asian
Americans Fit.” The Journal of Race, Ethnicity, and Politics 6(1): 16–32.

226
Huddy, Leonie, and Stanley Feldman. 2009. “On Assessing the Political Effects of Racial
Prejudice.” Annual Review of Political Science 12(1): 423–47.
Huddy, Leonie, and Nayda Terkildsen. 1993. “Gender Stereotypes and the Perception of Male
and Female Candidates.” American Journal of Political Science 37(1): 119–47.
Huder, Josh. 2013. “The House’s Competitiveness Problem… or Lack Thereof | The Government
Affairs Institute.” https://gai.georgetown.edu/the-houses-competitiveness-problem-orlack-thereof/ (October 12, 2023).
Hutcherson, Cendri A., and James J. Gross. 2011. “The Moral Emotions: A Social–Functionalist
Account of Anger, Disgust, and Contempt.” Journal of Personality and Social Psychology
100(4): 719–37.
Hutchings, Vincent L., and Nicholas A. Valentino. 2004. “The Centrality of Race in American
Politics.” Annual Review of Political Science 7(1): 383–408.
Hwang, Sungwook. 2013. “The Effect of Twitter Use on Politicians’ Credibility and Attitudes
toward Politicians.” Journal of Public Relations Research 25(3): 246–58.
Inbar, Yoel, David A. Pizarro, and Paul Bloom. 2009. “Conservatives Are More Easily Disgusted
than Liberals.” Cognition and Emotion 23(4): 714–25.
———. 2012. “Disgusting Smells Cause Decreased Liking of Gay Men.” Emotion 12(1): 23–27.
Ishizuka, Mitsuru, Alena Neviarouskaya, and Mostafa Al Masum Shaikh. 2012. “Textual Affect
Sensing and Affective Communication:” International Journal of Cognitive Informatics and
Natural Intelligence 6(4): 81–102.
Ito, Tiffany A., Jeff T. Larsen, N. Kyle Smith, and John T. Cacioppo. 1998. “Negative Information
Weighs More Heavily on the Brain: The Negativity Bias in Evaluative Categorizations.”
Journal of Personality and Social Psychology 75(4): 887–900.
Iyengar, Shanto. 1990. “The Accessibility Bias in Politics: Television News and Public Opinion.”
International Journal of Public Opinion Research 2(1): 1–15.
———. 2019. “The Origins and Consequences of Affective Polarization in the United States.” : 35.
Iyengar, Shanto, and Donald R. Kinder. 2010. News That Matters: Television and American Opinion,
Updated Edition. University of Chicago Press.
Iyengar, Shanto, Gaurav Sood, and Yphtach Lelkes. 2012. “Affect, Not Ideology: A Social Identity
Perspective on Polarization.” The Public Opinion Quarterly 76(3): 405–31.
Iyengar, Shanto, and Sean J. Westwood. 2015. “Fear and Loathing across Party Lines: New
Evidence on Group Polarization.” American Journal of Political Science 59(3): 690–707.
Jacobsmeier, Matthew L. 2014. “Racial Stereotypes and Perceptions of Representatives’ Ideologies
in U.S. House Elections.” Legislative Studies Quarterly 39(2): 261–91.

227
Jacobson, Gary C. 1987. “The Marginals Never Vanished: Incumbency and Competition in
Elections to the U.S. House of Representatives, 1952-82.” American Journal of Political Science
31(1): 126–41.
———. 2004. “Partisan and Ideological Polarization in the California Electorate.” State Politics &
Policy Quarterly 4(2): 113–39.
———. 2013. “Partisan Polarization in American Politics: A Background Paper.” Presidential
Studies Quarterly 43(4): 688–708.
Jarvis, Sharon E. 2021. Conservative Political Communication: How Right-Wing Media and Messaging
(Re)Made American Politics. Routledge.
Jenkins, Jennifer M., and Keith Oatley. 1996. “Chapter 23 - Emotional Episodes and Emotionality
Through the Life Span.” In Handbook of Emotion, Adult Development, and Aging, eds. Carol
Magai and Susan H. McFadden. San Diego: Academic Press, 421–41. https://www. -
sciencedirect.com/science/article/pii/B9780124649958500248 (October 16, 2023).
Johnson, Carol. 2013. “From Obama to Abbott.” Australian Feminist Studies 28(75): 14–29.
Johnson, Kristen, and Dan Goldwasser. 2018. “Classification of Moral Foundations in Microblog
Political Discourse.” In Proceedings of the 56th Annual Meeting of the Association for
Computational Linguistics (Volume 1: Long Papers), Melbourne, Australia: Association for
Computational Linguistics, 720–30. https://www.aclweb.org/anthology/P18-1067
(August 7, 2019).
Johnston, Anne. 2006. “Review of Campaigning for Hearts and Minds: How Emotional Appeals
in Political Ads Work.” The Public Opinion Quarterly 70(4): 625–27.
Jordan-Zachery, Julia. 2018. “Gendered Black Politics in the 21st Century.” National Political
Science Review 19(2): 180.
Jost, John T. 2007. “Coda--After ‘The End of the End of Ideology.’” American Psychologist 62(9):
1077–80.
———. 2017. “Ideological Asymmetries and the Essence of Political Psychology.” Political
Psychology 38(2): 167–208.
———. 2021. Left and Right: The Psychological Significance of a Political Distinction. Oxford
University Press.
Jost, John T., Jack Glaser, Arie W. Kruglanski, and Frank J. Sulloway. 2003. “Political
Conservatism as Motivated Social Cognition.” Psychological Bulletin 129(3): 339–75.
Just, Marion R. et al. 1996. Crosstalk: Citizens, Candidates, and the Media in a Presidential Campaign.
Chicago, IL: University of Chicago Press. https://press.uchicago.edu/ucp/books-
/book/chicago/C/bo3636890.html (July 20, 2023).
Kahn, Kim Fridkin. 1996. The Political Consequences of Being a Woman: How Stereotypes Influence the
Conduct and Consequences of Political Campaigns. New York: Columbia University Press.

228
Kahn, Kim Fridkin, and Patrick J. Kenney. 1999. “Do Negative Campaigns Mobilize or Suppress
Turnout? Clarifying the Relationship between Negativity and Participation.” American
Political Science Review 93(4): 877–89.
Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decision under
Risk.” Econometrica 47(2): 263–91.
Kam, Cindy D., and Beth A. Estes. 2016. “Disgust Sensitivity and Public Demand for Protection.”
The Journal of Politics 78(2): 481–96.
Kanouse, David E., and L. Reid Hanson. 1987. “Negativity in Evaluations.” In
Attribution: Perceiving the Causes of Behavior, Hillsdale, NJ, US: Lawrence Erlbaum
Associates, Inc, 47–62.
Kaslovsky, Jaclyn, and Jon C. Rogowski. 2022. “Under the Microscope: Gender and
Accountability in the U.S. Congress.” American Political Science Review 116(2): 516–32.
Kemp, Elyria, Pamela A. Kennett-Hensel, and Jeremy Kees. 2013. “Pulling on the Heartstrings:
Examining the Effects of Emotions and Gender in Persuasive Appeals.” Journal of
Advertising 42(1): 69–79.
Kern, Montague. 1989. 30-Second Politics: Political Advertising in the Eighties. New York: Praeger.
Kernell, Samuel. 2006. Going Public: New Strategies of Presidential Leadership. CQ Press.
Kinder, Donald R., and Allison Dale-Riddle. 2012. “The End of Race?” In The End of Race?, Yale
University Press. https://www.degruyter.com/document/doi/10.12987/9780300183597-
/html (October 5, 2023).
King, Gary, and Andrew Gelman. 1991. “Systemic Consequences of Incumbency Advantage in
U.S. House Elections.” American Journal of Political Science 35(1): 110–38.
King, David C., and Richard E. Matland. 2003. “Sex and the Grand Old Party: An Experimental
Investigation of the Effect of Candidate Sex on Support for a Republican Candidate.”
American Politics Research 31(6): 595–612.
Kosar, Kevin R. 2023. “Conservatives Should Look More Closely at Systemic Election Reforms.”
https://www.aei.org/research-products/report/conservatives-should-look-more-closelyat-systemic-election-reforms/.
Krehbiel, Keith, and John R. Wright. 1983. “The Incumbency Effect in Congressional Elections: A
Test of Two Explanations.” American Journal of Political Science 27(1): 140–57.
Kreiss, Daniel, Regina G. Lawrence, and Shannon C. McGregor. 2018. “In Their Own Words:
Political Practitioner Accounts of Candidates, Audiences, Affordances, Genres, and Timing
in Strategic Social Media Use.” Political Communication 35(1): 8–31.
Kruks, Sonia. 2005. “Simone de Beauvoir and the Politics of Privilege.” Hypatia 20(1): 178–205.
Krupnikov, Yanna, and Nichole M. Bauer. 2014. “The Relationship Between Campaign
Negativity, Gender and Campaign Context.” Political Behavior 36(1): 167–88.

229
Krupnikov, Yanna, and Spencer Piston. 2015. “Accentuating the Negative: Candidate Race and
Campaign Strategy.” Political Communication 32(1): 152–73.
Ladd, Jonathan M. 2011. “Why Americans Hate the Media and How It Matters.” In Why Americans
Hate the Media and How It Matters, Princeton University Press. https://www.-
degruyter.com/document/doi/10.1515/9781400840359/html (August 26, 2023).
LaFrance, Marianne, and Mahzarin Banaji. 1992. “Toward a Reconsideration of the GenderEmotion Relationship.” Emotion and Social Behavior 14: 178–201.
Lagares Diez, Nieves, Erika Jaráiz Gulías, and Gina Sibaja Quesada. 2022. “Emotions,
Engagement and Social Media.” In Digital Political Communication Strategies:
Multidisciplinary Reflections, The Palgrave Macmillan Series in International Political
Communication, ed. Berta García-Orosa. Cham: Springer International Publishing, 137–54.
https://doi.org/10.1007/978-3-030-81568-4_9 (October 14, 2023).
Landy, Justin F., Jared Piazza, and Geoffrey P. Goodwin. 2016. “When It’s Bad to Be Friendly and
Smart: The Desirability of Sociability and Competence Depends on Morality.” Personality
and Social Psychology Bulletin 42(9): 1272–90.
Lassen, David S., and Adam R. Brown. 2011. “Twitter: The Electoral Connection?” Social Science
Computer Review 29(4): 419–36.
Lau, Richard R., and Gerald M. Pomper. 2002. “Effectiveness of Negative Campaigning in U.S.
Senate Elections.” American Journal of Political Science 46(1): 47–66.
Lau, Richard R., and David P. Redlawsk. 2006. How Voters Decide: Information Processing in Election
Campaigns. Cambridge University Press.
Lau, Richard R., Lee Sigelman, and Ivy Brown Rovner. 2007. “The Effects of Negative Political
Campaigns: A Meta-Analytic Reassessment.” Journal of Politics 69(4): 1176–1209.
Lawless, Jennifer L. 2015. “Female Candidates and Legislators.” Annual Review of Political Science
18(1): 349–66.
Lawless, Jennifer L., and Kathryn Pearson. 2008. “The Primary Reason for Women’s
Underrepresentation? Reevaluating the Conventional Wisdom.” The Journal of Politics 70(1):
67–82.
Lawrence, Regina G., and Melody Rose. 2010. Hillary Clinton’s Race for the White House: Gender
Politics and the Media on the Campaign Trail. Boulder, Colo.: Lynne Rienner.
Lazarus, Jeffrey, and Shauna Reilly. 2010. “The Electoral Benefits of Distributive Spending.”
Political Research Quarterly 63(2): 343–55.
Lazarus, Richard S. 1991. Emotion and Adaptation. Oxford University Press.
Lerman, Amy E., and Meredith L. Sadin. 2016. “Stereotyping or Projection? How White and Black
Voters Estimate Black Candidates’ Ideology.” Political Psychology 37(2): 147–63.

230
Lerner, Jennifer S., and Dacher Keltner. 2001. “Fear, Anger, and Risk.” Journal of Personality and
Social Psychology 81(1): 146–59.
Levenson, Robert W. 1999. “The Intrapersonal Functions of Emotion.” Cognition & Emotion 13(5):
481–504.
Lewis, Kristi M. 2000. “When Leaders Display Emotion: How Followers Respond to Negative
Emotional Expression of Male and Female Leaders.” Journal of Organizational Behavior 21(2):
221–34.
Lewis-Beck, Michael S., Charles Tien, and Richard Nadeau. 2010. “Obama’s Missed Landslide: A
Racial Cost?” PS: Political Science & Politics 43(1): 69–76.
Lindstädt, René, and Ryan J. Vander Wielen. 2012. “Dynamic Partisanship: Party Loyalty and
Agenda Setting in the U.S. House.” SSRN Electronic Journal. http:// -
www.ssrn.com/abstract=1922816 (September 4, 2023).
Lipinski, Daniel. 2004. Congressional Communication: Content & Consequences. Ann Arbor:
University of Michigan Press.
Lodge, Milton, and Charles S. Taber. 2005. “The Automaticity of Affect for Political Leaders,
Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis.” Political
Psychology 26(3): 455–82.
Loseke, Donileen. 2009. “Examining Emotion as Discourse: Emotion Codes and Presidential
Speeches Justifying War.” The Sociological Quarterly 50: 497–524.
Lüders, Adrian, Christina Mühlberger, and Eva Jonas. 2020. “Motivational and Affective Drivers
of Right-Wing Populism Support: Insights From an Austrian Presidential Election.” Social
Psychological Bulletin 15(3): 1–17.
Lupia, Arthur, and Jesse O. Menning. 2009. “When Can Politicians Scare Citizens Into Supporting
Bad Policies?” American Journal of Political Science 53(1): 90–106.
Macdonald, Maggie. 2020. “Changing Audiences, Changing Messages? The Use of Social Media
in Congressional Campaigns.” PhD Thesis. Emory University.
Macdonald, Maggie, Whitney Hua, and Annelise Russell. 2023. “Constrained Communication
and Negativity Bias: Gendered Emotional Appeals on Facebook.” Journal of Women, Politics
& Policy: 1–14.
Mackie, Diane M., Thierry Devos, and Eliot R. Smith. 2000. “Intergroup Emotions: Explaining
Offensive Action Tendencies in an Intergroup Context.” Journal of Personality and Social
Psychology 79(4): 602–16.
Mackie, Diane M., and Eliot R. Smith. 2015. “Intergroup Emotions.” In APA Handbook of
Personality and Social Psychology, Volume 2: Group Processes., eds. Mario Mikulincer, Phillip
R. Shaver, John F. Dovidio, and Jeffry A. Simpson. Washington: American Psychological
Association, 263–93. http://content.apa.org/books/14342-010 (July 20, 2023).

231
———. 2018. “Intergroup Emotions Theory: Production, Regulation, and Modification of GroupBased Emotions.” In Advances in Experimental Social Psychology, Elsevier, 1–69.
https://linkinghub.elsevier.com/retrieve/pii/S0065260118300121 (August 1, 2023).
Mackie, Diane M., Eliot R. Smith, Ishani Banerji, and Anudhi Munasinghe. 2023. “Group-Based
Emotion Processes Generalize across Group Exemplars and Types.” Group Processes &
Intergroup Relations: 13684302231166856.
Mackie, Diane M., Eliot R. Smith, and Devin G. Ray. 2008. “Intergroup Emotions and Intergroup
Relations.” Social and Personality Psychology Compass 2(5): 1866–80.
MacKuen, Michael, Jennifer Wolak, Luke Keele, and George E. Marcus. 2010. “Civic
Engagements: Resolute Partisanship or Reflective Deliberation.” American Journal of Political
Science 54(2): 440–58.
Mandel, Ruth B. 1983. In the Running: The New Woman Candidate. Boston: Beacon Press.
Mann, Thomas E., and Raymond E. Wolfinger. 1980. “Candidates and Parties in Congressional
Elections.” The American Political Science Review 74(3): 617–32.
Mansbridge, Jane. 1999. “Should Blacks Represent Blacks and Women Represent Women? A
Contingent ‘Yes.’” The Journal of Politics 61(3): 628–57.
Marcus, George. 2003. “The Psychology of Emotion.” In , 182–221.
Marcus, George E. 1988. “The Structure of Emotional Response: 1984 Presidential Candidates.”
The American Political Science Review 82(3): 737–61.
———. 2002. The Sentimental Citizen: Emotion in Democratic Politics. University Park, PA:
Pennsylvania State University Press.
Marcus, George E., and Michael B. MacKuen. 1993. “Anxiety, Enthusiasm, and the Vote: The
Emotional Underpinnings of Learning and Involvement During Presidential Campaigns.”
The American Political Science Review 87(3): 672–85.
Marcus, George E., W. Russell Neuman, and Michael MacKuen. 2000. Affective Intelligence and
Political Judgment. Chicago: University of Chicago Press.
Marcus, George E., Nicholas A. Valentino, Pavlos Vasilopoulos, and Martial Foucault. 2019.
“Applying the Theory of Affective Intelligence to Support for Authoritarian Policies and
Parties.” Political Psychology 40(S1): 109–39.
Mark, David. 2006. Going Dirty: The Art of Negative Campaigning. Lanham, Md.: Rowman &
Littlefield.
Marland, Alex, and Brooks DeCillia. 2020. “Reputation and Brand Management by Political
Parties: Party Vetting of Election Candidates in Canada.” Journal of Nonprofit & Public Sector
Marketing 32(4): 342–63.
Martin, Aaron et al. 2020. “‘Challenging Public Rhetoric Justifying Immigrants as “Indecent”’”
ed. Lincoln Geraghty. Cogent Arts & Humanities 7(1): 1740538.

232
Mason, Lilliana. 2015. “‘I Disrespectfully Agree’: The Differential Effects of Partisan Sorting on
Social and Issue Polarization.” American Journal of Political Science 59(1): 128–45.
———. 2018. Uncivil Agreement: How Politics Became Our Identity. Chicago, Illinois ; London: The
University of Chicago Press.
Mason, Lilliana, and Julie Wronski. 2018. “One Tribe to Bind Them All: How Our Social Group
Attachments Strengthen Partisanship.” Political Psychology 39(S1): 257–77.
Mason, Shannon, and Lenandlar Singh. 2022. “Reporting and Discoverability of ‘Tweets’ Quoted
in Published Scholarship: Current Practice and Ethical Implications.” Research Ethics 18(2):
93–113.
Mayhew, David R. 1974. Congress: The Electoral Connection. 2nd ed. New Haven: Yale University
Press.
Mcdermott, Monika L. 1998. “Race and Gender Cues in Low-Information Elections.” Political
Research Quarterly 51(4): 895–918.
McDermott, Rose. 1998. Risk-Taking in International Politics: Prospect Theory in American Foreign
Policy. Paperback ed. Ann Arbor: University of Michigan Press.
McDermott, Rose, James H. Fowler, and Oleg Smirnov. 2008. “On the Evolutionary Origin of
Prospect Theory Preferences.” The Journal of Politics 70(2): 335–50.
McDonald, Maura, Rachel Porter, and Sarah A. Treul. 2020. “Running as a Woman? Candidate
Presentation in the 2018 Midterms.” Political Research Quarterly 73(4): 967–87.
McElroy, Gail, and Michael Marsh. 2010. “Candidate Gender and Voter Choice: Analysis from a
Multimember Preferential Voting System.” Political Research Quarterly 63(4): 822–33.
McIntosh, Peggy. 1988. White Privilege and Male Privilege: A Personal Account of Coming to See
Correspondences through Work in Women’s Studies. Center for Research on Women: Wellesley
College. Working Paper.
———. 1989. “White Privilege: Unpacking the Invisible Knapsack.” In On Privilege, Fraudulence,
and Teaching As Learning, Routledge, 29–34. https://www.taylorfrancis.com/books/-
9781351133784/chapters/10.4324/9781351133791-4 (September 10, 2023).
———. 2019. On Privilege, Fraudulence, and Teaching As Learning: Selected Essays 1981--2019. New
York: Routledge.
Meeks, Lindsey. 2017. “Getting Personal: Effects of Twitter Personalization on Candidate
Evaluations.” Politics & Gender 13(1): 1–25.
Meiselman, L. Herbert, ed. 2021. Emotion Measurement. Second Edition. Duxford: WoodHead
Publishing.
Mohammad, Saif M. 2021. “Chapter 11 - Sentiment Analysis: Automatically Detecting Valence,
Emotions, and Other Affectual States from Text.” In Emotion Measurement (Second Edition),

233
ed. Herbert L. Meiselman. Woodhead Publishing, 323–79.
https://www.sciencedirect.com/science/article/pii/B9780128211243000119 (October 9,
2023).
Mohammad, Saif M., and Peter D. Turney. 2013. “Crowdsourcing a Word-Emotion Association
Lexicon.” http://arxiv.org/abs/1308.6297 (June 25, 2023).
Molho, Catherine et al. 2017. “Disgust and Anger Relate to Different Aggressive Responses to
Moral Violations.” Psychological Science 28(5): 609–19.
Mullen, Brian et al. 1986. “Newcasters’ Facial Expressions and Voting Behavior of Viewers: Can
a Smile Elect a President?” Journal of Personality and Social Psychology 51(2): 291–95.
Mummolo, Jonathan, Erik Peterson, and Sean Westwood. 2021. “The Limits of Partisan Loyalty.”
Political Behavior 43(3): 949–72.
Munro, Geoffrey D., Julia Zirpoli, Adam Schuman, and Jeff Taulbee. 2013. “Third-Party Labels
Bias Evaluations of Political Platforms and Candidates.” Basic and Applied Social Psychology
35(2): 151–63.
Mutz, Diana C. 2007. “Effects of ‘In-Your-Face’ Television Discourse on Perceptions of a
Legitimate Opposition.” American Political Science Review 101(4): 621–35.
Mutz, Diana C., and Byron Reeves. 2005. “The New Videomalaise: Effects of Televised Incivility
on Political Trust.” American Political Science Review 99(1): 1–15.
Nadeau, Richard, Richard G. Niemi, and Timothy Amato. 1995. “Emotions, Issue Importance,
and Political Learning.” American Journal of Political Science 39(3): 558.
Neuman, W. Russell, ed. 2007. The Affect Effect: Dynamics of Emotion in Political Thinking and
Behavior. Chicago: University of Chicago Press.
Neuman, W. Russell, George E. Marcus, and Michael B. MacKuen. 2018. “Hardwired for News:
Affective Intelligence and Political Attention.” Journal of Broadcasting & Electronic Media
62(4): 614–35.
Nieto, Leticia et al. 2010. Beyond Inclusion, beyond Empowerment a Developmental Strategy to Liberate
Everyone. Cuetzpalin.
Nosek, Brian A. 2007. “Implicit–Explicit Relations.” Current Directions in Psychological Science
16(2): 65–69.
———. 2007. “Pervasiveness and Correlates of Implicit Attitudes and Stereotypes.” European
Review of Social Psychology 18(1): 36–88.
O’Doherty, J. et al. 2001. “Abstract Reward and Punishment Representations in the Human
Orbitofrontal Cortex.” Nature Neuroscience 4(1): 95–102.
Orey, Byron D’Andra. 2006. “Deracialization or Racialization: The Making of a Black Mayor in
Jackson, Mississippi.” Politics & Policy 34(4): 814–36.

234
Orey, Byron D’Andra, and Boris E. Ricks. 2007. “A Systematic Analysis of the Deracialization
Concept.” National Political Science Review 11: 325–34.
Osborn, Tracy L. 2012a. “How Parties Affect Gender and Representation.” In How Women
Represent Women: Political Parties, Gender and Representation in the State Legislatures, ed. Tracy
L. Osborn. Oxford University Press, 0. https://doi.org/10.1093/acprof:oso-
/9780199845347.003.0002 (October 16, 2023).
———. 2012b. “Introduction.” In How Women Represent Women: Political Parties, Gender and
Representation in the State Legislatures, ed. Tracy L. Osborn. Oxford University Press.
https://doi.org/10.1093/acprof:oso/9780199845347.003.0001 (October 16, 2023).
Oxley, Douglas R. et al. 2008. “Political Attitudes Vary with Physiological Traits.” Science
321(5896): 1667–70.
Park, Ju Yeon. 2021. “When Do Politicians Grandstand? Measuring Message Politics in
Committee Hearings.” The Journal of Politics 83(1): 214–28.
Perry, Huey L. 1991. “Deracialization as an Analytical Construct in American Urban Politics.”
Urban Affairs Quarterly 27(2): 181–91.
Persons, Georgia A. 2011. The Expanding Boundaries of Black Politics. Transaction Publishers.
Pew Research Center. 2017. Partisan Conflict and Congressional Outreach. Washington, D.C.: Pew
Research Center. https://www.pewresearch.org/politics/2017/02/23/partisan-conflictand-congressional-outreach/.
Pew Research Center. 2020. Congress Soars to New Heights on Social Media. Washington, D.C.: Pew
Research Center. https://www.pewresearch.org/internet/2020/07/16/congress-soars-tonew-heights-on-social-media/.
Pew Research Center. 2023. The Changing Face of Congress in 8 Charts. Washington, D.C.: Pew
Research Center. https://www.pewresearch.org/short-reads/2023/02/07/the-changingface-of-congress/.
Phillips, Anne. 1998. The Politics of Presence. OUP Oxford.
Phillips, Christian Dyogi. 2021. Nowhere to Run: Race, Gender, and Immigration in American Elections.
New York, NY: Oxford University Press.
Phoenix, Davin L. 2019. The Anger Gap: How Race Shapes Emotion in Politics. Cambridge:
Cambridge University Press. https://www.cambridge.org/core/books/angergap/967D469A566F389E0C69099BA5486069 (October 14, 2023).
Picard, Rosalind W. 1997. Affective Computing. Cambridge (Mass.): MIT Press.
Piston, Spencer. 2010. “How Explicit Racial Prejudice Hurt Obama in the 2008 Election.” Political
Behavior 32(4): 431–51.
———. 2018. Class Attitudes in America: Sympathy for the Poor, Resentment of the Rich, and Political
Implications. Cambridge University Press.

235
Piston, Spencer, Yanna Krupnikov, Kerri Milita, and John Barry Ryan. 2018. “Clear as Black and
White: The Effects of Ambiguous Rhetoric Depend on Candidate Race.” The Journal of
Politics 80(2): 662–74.
Plant, E. Ashby, Janet Shibley Hyde, Dacher Keltner, and Patricia G. Devine. 2000. “The Gender
Stereotyping of Emotions.” Psychology of Women Quarterly 24(1): 81–92.
Plutchik, Robert. 1980. “Chapter 1 – A General Psychoevolutionary Theory of Emotion.” In
Theories of Emotion, eds. Robert Plutchik and Henry Kellerman. Academic Press, 3–33.
https://www.sciencedirect.com/science/article/pii/B9780125587013500077 (October 9,
2023).
———. 1982. “A Psychoevolutionary Theory of Emotions.” Social Science Information 21(4–5): 529–
53.
———. 2001. “The Nature of Emotions: Human Emotions Have Deep Evolutionary Roots, a Fact
That May Explain Their Complexity and Provide Tools for Clinical Practice.” American
Scientist 89(4): 344–50.
“Political Party Affiliation and Presidential Campaign Discourse.” 2004. Communication Quarterly
52(2): 81–97.
Pratto, Felicia, and Andrew L. Stewart. 2011. “Social Dominance Theory.” In The Encyclopedia of
Peace Psychology, John Wiley & Sons, Ltd. https://onlinelibrary.wiley.com/-
doi/abs/10.1002/9780470672532.wbepp253 (October 8, 2023).
Reeves, Keith. 1997a. Voting Hopes or Fears? White Voters, Black Candidates & Racial Politics in
America. New York: Oxford Univ. Press.
———. 1997b. Voting Hopes Or Fears?: White Voters, Black Candidates & Racial Politics in America.
Oxford University Press.
Reingold, Beth, ed. 2008. Legislative Women: Getting Elected, Getting Ahead. Boulder, Colo: Lynne
Rienner.
Reingold, Beth, and Jessica Harrell. 2010. “The Impact of Descriptive Representation on Women’s
Political Engagement: Does Party Matter?” Political Research Quarterly 63(2): 280–94.
Reingold, Beth, Kirsten Widner, and Rachel Harmon. 2020. “Legislating at the Intersections: Race,
Gender, and Representation.” Political Research Quarterly 73(4): 819–33.
Rheault, Ludovic, Kaspar Beelen, Christopher Cochrane, and Graeme Hirst. 2016. “Measuring
Emotion in Parliamentary Debates with Automated Textual Analysis” ed. Joseph Najbauer.
PLOS ONE 11(12): e0168843.
Ridout, Travis N., and Kathleen Searles. 2011. “It’s My Campaign I’ll Cry If I Want to: How and
When Campaigns Use Emotional Appeals: How and When Campaigns Use Emotional
Appeals.” Political Psychology 32(3): 439–58.

236
Rilling, James K., and Alan G. Sanfey. 2011. “The Neuroscience of Social Decision-Making.”
Annual Review of Psychology 62(1): 23–48.
Rogowski, Jon C., and Andrew R. Stone. 2020. “Words Speak Louder than Actions: Public
Responsiveness to Elite Communication.” Political Behavior 42(2): 327–54.
Roth, Leopold Helmut Otto, and Anton-Rupert Laireiter. 2021. “Factor Structure of the ‘Top Ten’
Positive Emotions of Barbara Fredrickson.” Frontiers in psychology 12: 641804–641804.
Rozin, Paul, and April E. Fallon. 1987. “A Perspective on Disgust.” Psychological Review 94(1): 23–
41.
Rudolph, Thomas J., Amy Gangl, and Dan Stevens. 2000. “The Effects of Efficacy and Emotions
on Campaign Involvement.” The Journal of Politics 62(4): 1189–97.
Russell, Annelise. 2021a. “Constituent Connections: Senators’ Reputation Building in the Age of
Social Media.” Journal of Information Technology & Politics: 1–17.
———. 2021b. Tweeting Is Leading: How Senators Communicate and Represent in the Age of Twitter.
New York, NY: OXford University Press.
Russell, Annelise, Maggie Macdonald, and Whitney Hua. 2023. “Sit Still, Talk Pretty: Partisan
Differences Among Women Candidates’ Campaign Appeals.” Journal of Women, Politics &
Policy 44(3): 354–70.
Russell, James A., and James M. Carroll. 1999. “On the Bipolarity of Positive and Negative Affect.”
Psychological Bulletin 125(1): 3–30.
Rydell, Robert J. et al. 2008. “Arousal, Processing, and Risk Taking: Consequences of Intergroup
Anger.” Personality and Social Psychology Bulletin 34(8): 1141–52.
Sagi, Eyal, and Morteza Dehghani. 2014. “Measuring Moral Rhetoric in Text.” Social Science
Computer Review 32(2): 132–44.
Sanbonmatsu, Kira. 2002. “Gender Stereotypes and Vote Choice.” American Journal of Political
Science 46(1): 20–34.
Sanbonmatsu, Kira, and Kathleen Dolan. 2009. “Do Gender Stereotypes Transcend Party?”
Political Research Quarterly 62(3): 485–94.
Santoro, Lauren Ratliff et al. 2021. “Exploring the Direct and Indirect Effects of Elite Influence on
Public Opinion.” PLOS ONE 16(11): e0257335.
Sapiro, Virginia, Katherine Cramer Walsh, Patricia Strach, and Valerie Hennings. 2011. “Gender,
Context, and Television Advertising: A Comprehensive Analysis of 2000 and 2002 House
Races.” Political Research Quarterly 64(1): 107–19.
Schaffner, Brian F. 2011. “Racial Salience and the Obama Vote.” Political Psychology 32(6): 963–88.
Schattschneider, Elmer Eric. 1975. The Semisovereign People: A Realist’s View of Democracy in
America. Wadsworth Publishing Company.

237
Scheller, Simon. 2019. “The Strategic Use of Fear Appeals in Political Communication.” Political
Communication 36(4): 586–608.
Scherer, Klaus R. 2005. “What Are Emotions? And How Can They Be Measured?” Social Science
Information 44(4): 695–729.
Scheurich, James Joseph. 1993. “Toward a White Discourse on White Racism.” Educational
Researcher 22(8): 5–10.
Schumaker, Paul, and Nancy Elizabeth Burns. 1988. “Gender Cleavages and the Resolution of
Local Policy Issues.” American Journal of Political Science 32(4): 1070–95.
Seger, Charles R., Eliot R. Smith, and Diane M. Mackie. 2009. “Subtle Activation of a Social
Categorization Triggers Group-Level Emotions.” Journal of Experimental Social Psychology
45(3): 460–67.
Shapiro, Matthew A., and Libby Hemphill. 2017. “Politicians and the Policy Agenda: Does Use of
Twitter by the U.S. Congress Direct New York Times Content?” Policy & Internet 9(1): 109–
32.
Sidanius, Jim, and Felicia Pratto. 2003. “Social Dominance Theory and the Dynamics of
Inequality: A Reply to Schmitt, Branscombe, & Kappen and Wilson & Liu.” The British
journal of social psychology / the British Psychological Society 42: 207–13; discussion 215.
Sigelman, Carol K., Lee Sigelman, Barbara J. Walkosz, and Michael Nitz. 1995. “Black Candidates,
White Voters: Understanding Racial Bias in Political Perceptions.” American Journal of
Political Science 39(1): 243–65.
Simien, Evelyn M. 2007. “Doing Intersectionality Research: From Conceptual Issues to Practical
Examples.” Politics & Gender 3(2): 264–71.
Smith, Eliot R., and Diane M. Mackie. 2015. “Dynamics of Group-Based Emotions: Insights From
Intergroup Emotions Theory.” Emotion Review 7(4): 349–54.
Smith, Nicola J., and Donna Lee. 2015. “What’s Queer about Political Science?” The British Journal
of Politics and International Relations 17(1): 49–63.
Smooth, Wendy. 2011. “Standing for Women? Which Women? The Substantive Representation
of Women’s Interests and the Research Imperative of Intersectionality.” Politics & Gender
7(3): 436–41.
Soroka, Stuart, and Stephen McAdams. 2015. “News, Politics, and Negativity.” Political
Communication 32(1): 1–22.
Soroka, Stuart N. 2014. Negativity in Democratic Politics: Causes and Consequences. Cambridge:
Cambridge University Press. https://www.cambridge.org/core/books/negativity-indemocratic-politics/D88A1796CAD568E302A74C04475710E8 (October 5, 2023).
Stephens-Davidowitz, Seth. 2014. “The Cost of Racial Animus on a Black Candidate: Evidence
Using Google Search Data.” Journal of Public Economics 118: 26–40.

238
Sterling, Joanna, and John T Jost. 2018. “Moral Discourse in the Twitterverse.” Journal of Language
and Politics 17(2): 195–221.
Stevens, Tim M., Noelle Aarts, and Art Dewulf. 2020. “Using Emotions to Frame Issues and
Identities in Conflict: Farmer Movements on Social Media.” Negotiation and Conflict
Management Research https://onlinelibrary.wiley.com/doi/abs/10.1111/ncmr.12177
(October 8, 2023).
Stieglitz, Stefan, and Linh Dang-Xuan. 2013. “Emotions and Information Diffusion in Social
Media—Sentiment of Microblogs and Sharing Behavior.” Journal of Management Information
Systems 29(4): 217–48.
Stier, Sebastian, Arnim Bleier, Haiko Lietz, and Markus Strohmaier. 2018. “Election Campaigning
on Social Media: Politicians, Audiences, and the Mediation of Political Communication on
Facebook and Twitter.” Political Communication 35(1): 50–74.
Suhay, Elizabeth, Bernard Grofman, Alexandre H. Trechsel, and Oxford University Press, eds.
2020. The Oxford Handbook of Electoral Persuasion. New York: Oxford University Press.
Sullivan, Denis G., and Roger D. Masters. 1987. “‘Gut Reactions’ and the Political Effects of the
Media.” PS 20(4): 880–89.
Sydnor, Emily. 2019. “Disrespectful Democracy: The Psychology of Political Incivility.” In
Disrespectful Democracy, Columbia University Press. https://www.degruyter.com-
/document/doi/10.7312/sydn18924/html (October 5, 2023).
Sylwester, Karolina, and Matthew Purver. 2015. “Twitter Language Use Reflects Psychological
Differences between Democrats and Republicans” ed. Christopher M. Danforth. PLOS ONE
10(9): e0137422.
Teele, Dawn Langan, Joshua Kalla, and Frances Rosenbluth. 2018. “The Ties That Double Bind:
Social Roles and Women’s Underrepresentation in Politics.” American Political Science
Review 112(3): 525–41.
Terkildsen, Nayda. 1993. “When White Voters Evaluate Black Candidates: The Processing
Implications of Candidate Skin Color, Prejudice, and Self-Monitoring.” American Journal of
Political Science 37(4): 1032–53.
Tesler, Michael. 2012. “The Spillover of Racialization into Health Care: How President Obama
Polarized Public Opinion by Racial Attitudes and Race: SPILLOVER OF RACIALIZATION
INTO HEALTH CARE.” American Journal of Political Science 56(3): 690–704.
Thomsen, Danielle M. 2015. “Why So Few (Republican) Women? Explaining the Partisan
Imbalance of Women in the U.S. Congress.” Legislative Studies Quarterly 40(2): 295–323.
Tiedens, Larissa Z., Phoebe C. Ellsworth, and Batja Mesquita. 2000. “Sentimental Stereotypes:
Emotional Expectations for High-and Low-Status Group Members.” Personality and Social
Psychology Bulletin 26(5): 560–75.

239
Tokeshi, Matthew. 2023. “Anti-Black Prejudice in Asian American Public Opinion.” Politics,
Groups, and Identities 11(2): 366–89.
Tran, Veronique, Patricia Garcia-Prieto, and Susan C Schneider. 2011. “The Role of Social Identity,
Appraisal, and Emotion in Determining Responses to Diversity Management.” Human
Relations 64(2): 161–76.
Vaish, Amrisha, Tobias Grossmann, and Amanda Woodward. 2008. “Not All Emotions Are
Created Equal: The Negativity Bias in Social-Emotional Development.” Psychological bulletin
134(3): 383–403.
Valentino, Nicholas A. et al. 2011. “Election Night’s Alright for Fighting: The Role of Emotions in
Political Participation.” The Journal of Politics 73(1): 156–70.
Valentino, Nicholas A., Vincent L. Hutchings, Antoine J. Banks, and Anne K. Davis. 2008. “Is a
Worried Citizen a Good Citizen? Emotions, Political Information Seeking, and Learning via
the Internet.” Political Psychology 29(2): 247–73.
Vasilopoulos, Pavlos, George E. Marcus, Nicholas A. Valentino, and Martial Foucault. 2019.
“Fear, Anger, and Voting for the Far Right: Evidence From the November 13, 2015 Paris
Terror Attacks.” Political Psychology 40(4): 679–704.
Vasilopoulou, Sofia, and Markus Wagner. 2017. “Fear, Anger and Enthusiasm about the
European Union: Effects of Emotional Reactions on Public Preferences towards European
Integration.” European Union Politics 18(3): 382–405.
Visalvanich, Neil. 2017. “Asian Candidates in America: The Surprising Effects of Positive Racial
Stereotyping.” Political Research Quarterly 70(1): 68–81.
Wagemans, Fieke M. A., Mark J. Brandt, and Marcel Zeelenberg. 2018. “Disgust Sensitivity Is
Primarily Associated with Purity-Based Moral Judgments.” Emotion 18(2): 277–89.
Wagner, Markus. 2014. “Fear and Anger in Great Britain: Blame Assignment and Emotional
Reactions to the Financial Crisis.” Political Behavior 36(3): 683–703.
Wagner, Markus, and Davide Morisi. 2019. “Anxiety, Fear, and Political Decision Making.” In
Oxford Research Encyclopedia of Politics, Oxford University Press.
https://oxfordre.com/politics/view/10.1093/acrefore/9780190228637.001.0001/acrefore9780190228637-e-915 (July 26, 2023).
Waldman, Paul, and Kathleen Hall Jamieson. 2003. “Rhetorical Convergence and Issue
Knowledge in the 2000 Presidential Election.” Presidential Studies Quarterly 33(1): 145–63.
Ward, Orlanda. 2016. “Seeing Double: Race, Gender, and Coverage of Minority Women’s
Campaigns for the U.S. House of Representatives.” Politics & Gender 12(2): 317–43.
Watson, David, and Auke Tellegen. 1999. “Issues in Dimensional Structure of Affect—Effects of
Descriptors, Measurement Error, and Response Formats: Comment on Russell and Carroll
(1999).” Psychological Bulletin 125(5): 601–10.

240
Weber, Christopher. 2013. “Emotions, Campaigns, and Political Participation.” Political Research
Quarterly 66(2): 414–28.
Webster, Steven W. 2020. American Rage: How Anger Shapes Our Politics. Cambridge: Cambridge
University Press. https://www.cambridge.org/core/books/american-rage/EE2BAF959B773D58529F148782E58CF2 (October 14, 2023).
Webster, Steven W., and Alan I. Abramowitz. 2017. “The Ideological Foundations of Affective
Polarization in the U.S. Electorate.” American Politics Research 45(4): 621–47.
Westen, Drew. 2008. The Political Brain: The Role of Emotion in Deciding the Fate of the Nation. New
York, United States: PublicAffairs.
Widmann, Tobias. 2021. “How Emotional Are Populists Really? Factors Explaining Emotional
Appeals in the Communication of Political Parties.” Political Psychology 42(1): 163–81.
Williams, Melissa S. 1998. Voice, Trust, and Memory: Marginalized Groups and the Failings of Liberal
Representation. Princeton University Press. https://www.jstor.org/stable/j.ctv1fkgdhc
(September 4, 2023).
Windett, Jason. 2014. “Differing Paths to the Top: Gender, Ambition, and Running for Governor.”
Journal of Women, Politics & Policy 35(4): 287–314.
Wineinger, Catherine N. 2022. Gendering the Republican: Intraparty Politics and Republican Women’s
Representation in Congress. 1st ed. Oxford University Press.
https://academic.oup.com/book/38819 (October 16, 2023).
Woessner, Matthew C. 2005. “Scandal, Elites, and Presidential Popularity: Considering the
Importance of Cues in Public Support of the President.” Presidential Studies Quarterly 35(1):
94–115.
Wyer Jr, Robert S., and Thomas K. Srull. 2014. Handbook of Social Cognition: Volume 1: Basic
Processes. Psychology Press.
Yarchi, Moran, Christian Baden, and Neta Kligler-Vilenchik. 2021. “Political Polarization on the
Digital Sphere: A Cross-Platform, Over-Time Analysis of Interactional, Positional, and
Affective Polarization on Social Media.” Political Communication 38(1–2): 98–139.
Young, Lauren E. 2023. “Mobilization Under Threat: Emotional Appeals and Pro-Opposition
Political Participation Online.” Political Behavior 45(2): 445–68.
Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge [England] ; New York, NY,
USA: Cambridge University Press.
Zou, Linda X., and Sapna Cheryan. 2017. “Two Axes of Subordination: A New Model of Racial
Position.” Journal of Personality and Social Psychology 112(5): 696–717.

Signaling Identity: How Race and Gender Shape What
Representatives Say Online
APPENDIX
Contents
A OLS Regressions of MC Use of Emotion-Associated Rhetoric on Twitter 242
A.1 113th to 116th MC official House accounts ........................ 242
A.1.1 raceGen factor models ................................ 242
A.1.2 gender*race interaction term models ........................ 245
A.1.3 separate raceGen group models .......................... 250
A.1.4 Subset by Party .................................... 258
A.1.5 Model estimates of separate race/gender effects ................. 274
A.2 2020 incumbent and challenger candidate accounts on Twitter ............. 279
A.2.1 raceGen factor models ................................ 279
A.2.2 woman*white interaction term models ...................... 289
A.2.3 raceGen group*party interaction term models .................. 291
A.2.4 main effects raceGen group models ........................ 299
A.2.5 raceGen group*party subset by inc status ..................... 313
A.2.6 Subgroup main and *party effects coefs. ...................... 329
B Model estimates of RGIDs and party interaction effects on emotive rhetoric 332
B.1 Incumbent Representatives in the 113th to 116th Congresses (2017 to 2020) ...... 332
B.2 Challenger and Incumbent Candidates in the 2020 Congressional Races ....... 335
C NRC Emotion-Associated Dictionary 339
C.1 Number of k Features in NRC Dictionary ......................... 339
C.2 Correlation of Emotion Categories ............................. 339
C.3 Measuring Emotional Appeals (Operationalizing Risk-Taking) ................ 340
241

A OLS Regressions of MC Use of Emotion-Associated Rhetoric on
Twitter
A.1 113th to 116th MC official House accounts
A.1.1 raceGen factor models
242

Table A.1: OLS regressions of all 113 to 116 MCs’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.007⇤⇤ 0.002 0.006⇤⇤⇤ 0.007⇤⇤
(0.003) (0.026) (0.002) (0.003)
BIPOC Woman (fct) 0.013⇤⇤⇤ 0.003 0.010⇤⇤⇤ 0.010⇤⇤⇤
(0.004) (0.018) (0.003) (0.004)
White Woman (fct) 0.001 0.010 0.0001 0.007⇤⇤
(0.003) (0.027) (0.002) (0.003)
Republican 0.037⇤⇤⇤ 0.048⇤⇤⇤ 0.014⇤⇤⇤ 0.014⇤⇤⇤
(0.003) (0.008) (0.002) (0.003)
idExtreme 0.201⇤⇤⇤ 0.131⇤⇤⇤ 0.091⇤⇤⇤ 0.114⇤⇤⇤
(0.008) (0.006) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009⇤⇤ 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.004) (0.002) (0.003)
Unopposed 0.024⇤⇤⇤ 0.017⇤⇤⇤ 0.014⇤⇤⇤ 0.019⇤⇤⇤
(0.004) (0.007) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.033⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.009) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.007) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013⇤⇤⇤ 0.005⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.005) (0.001) (0.002)
Constant 0.212⇤⇤⇤ 0.145⇤⇤⇤ 0.137⇤⇤⇤ 0.160⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.410 0.361 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
243

Table A.2: OLS regressions of all 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.001 0.001 0.003 0.009⇤⇤⇤
(0.003) (0.025) (0.003) (0.003)
BIPOC Woman (fct) 0.016⇤⇤⇤ 0.030 0.020⇤⇤⇤ 0.014⇤⇤⇤
(0.004) (0.024) (0.004) (0.003)
White Woman (fct) 0.014⇤⇤⇤ 0.021 0.001 0.011⇤⇤⇤
(0.003) (0.020) (0.003) (0.003)
Republican 0.019⇤⇤⇤ 0.016⇤⇤ 0.0003 0.014⇤⇤⇤
(0.003) (0.007) (0.003) (0.003)
idExtreme 0.047⇤⇤⇤ 0.022⇤⇤⇤ 0.017⇤⇤ 0.014⇤⇤
(0.007) (0.006) (0.007) (0.006)
Competitive Seat 0.014⇤⇤⇤ 0.015⇤⇤⇤ 0.003 0.009⇤⇤⇤
(0.003) (0.006) (0.003) (0.003)
Unopposed 0.014⇤⇤⇤ 0.012⇤⇤⇤ 0.005 0.008⇤⇤
(0.004) (0.005) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.007) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.0003
(0.002) (0.006) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.012⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Constant 0.330⇤⇤⇤ 0.153⇤⇤⇤ 0.179⇤⇤⇤ 0.181⇤⇤⇤
(0.014) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.329 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
244

A.1.2 gender*race interaction term models
245

Table A.3: OLS regressions of 113 to 116 MCs’ logged average use of negative emotive appeals on
Twitter (woman*White)
negative anger disgust fear
(1) (2) (3) (4)
Woman 0.006 0.005 0.005⇤ 0.003
(0.004) (0.026) (0.003) (0.004)
White 0.007⇤⇤ 0.002 0.006⇤⇤⇤ 0.007⇤⇤
(0.003) (0.018) (0.002) (0.003)
Republican 0.037⇤⇤⇤ 0.048⇤ 0.014⇤⇤⇤ 0.014⇤⇤⇤
(0.003) (0.027) (0.002) (0.003)
idExtreme 0.201⇤⇤⇤ 0.131⇤⇤⇤ 0.091⇤⇤⇤ 0.114⇤⇤⇤
(0.008) (0.010) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.007) (0.002) (0.003)
Unopposed 0.024⇤⇤⇤ 0.017⇤⇤⇤ 0.014⇤⇤⇤ 0.019⇤⇤⇤
(0.004) (0.005) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.033⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.008) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.008) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013⇤⇤ 0.005⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.006) (0.001) (0.002)
Woman*White 0.007 0.005 0.005 0.010⇤⇤
(0.005) (0.003) (0.003) (0.005)
Constant 0.219⇤⇤⇤ 0.147⇤⇤⇤ 0.142⇤⇤⇤ 0.167⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.410 0.361 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
246

Table A.4: OLS regressions of 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter (woman*White)
positive joy trust anticipation
(1) (2) (3) (4)
Woman 0.015⇤⇤⇤ 0.031 0.017⇤⇤⇤ 0.024⇤⇤⇤
(0.004) (0.024) (0.004) (0.004)
White 0.001 0.001 0.003 0.009⇤⇤⇤
(0.003) (0.025) (0.003) (0.003)
Republican 0.019⇤⇤⇤ 0.016 0.0003 0.014⇤⇤⇤
(0.003) (0.020) (0.003) (0.003)
idExtreme 0.047⇤⇤⇤ 0.022⇤⇤⇤ 0.017⇤⇤ 0.014⇤⇤
(0.007) (0.008) (0.007) (0.006)
Competitive Seat 0.014⇤⇤⇤ 0.015⇤⇤ 0.003 0.009⇤⇤⇤
(0.003) (0.007) (0.003) (0.003)
Unopposed 0.014⇤⇤⇤ 0.012⇤ 0.005 0.008⇤⇤
(0.004) (0.007) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.0003
(0.002) (0.007) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.012⇤⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Woman*White 0.001 0.010⇤⇤⇤ 0.016⇤⇤⇤ 0.013⇤⇤⇤
(0.005) (0.003) (0.005) (0.005)
Constant 0.332⇤⇤⇤ 0.151⇤⇤⇤ 0.176⇤⇤⇤ 0.171⇤⇤⇤
(0.013) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.329 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
247

Table A.5: OLS regressions of 113 to 116 MCs’ logged average use of negative emotive appeals on
Twitter (woman*BIPOC)
negative anger disgust fear
(1) (2) (3) (4)
Woman 0.001 0.010 0.0001 0.007⇤⇤
(0.003) (0.026) (0.002) (0.003)
BIPOC 0.007⇤⇤ 0.002 0.006⇤⇤⇤ 0.007⇤⇤
(0.003) (0.018) (0.002) (0.003)
Republican 0.037⇤⇤⇤ 0.048⇤ 0.014⇤⇤⇤ 0.014⇤⇤⇤
(0.003) (0.027) (0.002) (0.003)
idExtreme 0.201⇤⇤⇤ 0.131⇤⇤⇤ 0.091⇤⇤⇤ 0.114⇤⇤⇤
(0.008) (0.007) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009⇤ 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.005) (0.002) (0.003)
Unopposed 0.024⇤⇤⇤ 0.017⇤⇤⇤ 0.014⇤⇤⇤ 0.019⇤⇤⇤
(0.004) (0.004) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.033⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.006) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.008) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013⇤⇤ 0.005⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.006) (0.001) (0.002)
Woman*BIPOC 0.007 0.005 0.005 0.010⇤⇤
(0.005) (0.003) (0.003) (0.005)
Constant 0.212⇤⇤⇤ 0.145⇤⇤⇤ 0.137⇤⇤⇤ 0.160⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.410 0.361 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
248

Table A.6: OLS regressions of 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter (woman*BIPOC)
positive joy trust anticipation
(1) (2) (3) (4)
Woman 0.014⇤⇤⇤ 0.021 0.001 0.011⇤⇤⇤
(0.003) (0.025) (0.003) (0.003)
BIPOC 0.001 0.001 0.003 0.009⇤⇤⇤
(0.003) (0.024) (0.003) (0.003)
Republican 0.019⇤⇤⇤ 0.016 0.0003 0.014⇤⇤⇤
(0.003) (0.020) (0.003) (0.003)
idExtreme 0.047⇤⇤⇤ 0.022⇤⇤⇤ 0.017⇤⇤ 0.014⇤⇤
(0.007) (0.005) (0.007) (0.006)
Competitive Seat 0.014⇤⇤⇤ 0.015⇤⇤⇤ 0.003 0.009⇤⇤⇤
(0.003) (0.005) (0.003) (0.003)
Unopposed 0.014⇤⇤⇤ 0.012⇤⇤⇤ 0.005 0.008⇤⇤
(0.004) (0.005) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.004) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.0003
(0.002) (0.007) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.012⇤⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Woman*BIPOC 0.001 0.010⇤⇤⇤ 0.016⇤⇤⇤ 0.013⇤⇤⇤
(0.005) (0.003) (0.005) (0.005)
Constant 0.330⇤⇤⇤ 0.153⇤⇤⇤ 0.179⇤⇤⇤ 0.181⇤⇤⇤
(0.014) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.329 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
249

A.1.3 separate raceGen group models
Table A.7: OLS regressions of all 113 to 116 MCs’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.003 0.006 0.004⇤⇤ 0.001
(0.003) (0.026) (0.002) (0.003)
Republican 0.039⇤⇤⇤ 0.049⇤⇤⇤ 0.012⇤⇤⇤ 0.012⇤⇤⇤
(0.005) (0.018) (0.003) (0.004)
idExtreme 0.196⇤⇤⇤ 0.128⇤⇤⇤ 0.089⇤⇤⇤ 0.109⇤⇤⇤
(0.008) (0.027) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.007) (0.002) (0.003)
Unopposed 0.023⇤⇤⇤ 0.016⇤⇤⇤ 0.013⇤⇤⇤ 0.018⇤⇤⇤
(0.004) (0.006) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.032⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.004) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013 0.006⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.011) (0.001) (0.002)
White Man*Republican 0.005 0.005 0.001 0.001
(0.005) (0.003) (0.003) (0.005)
Constant 0.218⇤⇤⇤ 0.146⇤⇤⇤ 0.142⇤⇤⇤ 0.167⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.409 0.361 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
250

Table A.8: OLS regressions of all 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.007⇤⇤ 0.012 0.005⇤ 0.001
(0.003) (0.024) (0.003) (0.002)
Republican 0.012⇤⇤ 0.008 0.001 0.006
(0.005) (0.025) (0.005) (0.004)
idExtreme 0.043⇤⇤⇤ 0.016 0.011 0.009
(0.007) (0.020) (0.007) (0.006)
Competitive Seat 0.013⇤⇤⇤ 0.015⇤⇤⇤ 0.004 0.008⇤⇤⇤
(0.003) (0.005) (0.003) (0.003)
Unopposed 0.014⇤⇤⇤ 0.012⇤⇤ 0.006 0.008⇤⇤
(0.004) (0.005) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.005) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.008⇤⇤⇤ 0.0002
(0.002) (0.004) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.011 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.010) (0.002) (0.002)
White Man*Republican 0.008 0.007⇤⇤ 0.002 0.009⇤
(0.005) (0.003) (0.005) (0.004)
Constant 0.330⇤⇤⇤ 0.150⇤⇤⇤ 0.176⇤⇤⇤ 0.170⇤⇤⇤
(0.014) (0.028) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.327 0.602 0.439
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
251

Table A.9: OLS regressions of all 113 to 116 MCs’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.004 0.004 0.003 0.001
(0.003) (0.026) (0.002) (0.003)
Republican 0.034⇤⇤⇤ 0.049⇤⇤⇤ 0.011⇤⇤⇤ 0.010⇤⇤⇤
(0.003) (0.018) (0.002) (0.003)
idExtreme 0.196⇤⇤⇤ 0.130⇤⇤⇤ 0.088⇤⇤⇤ 0.108⇤⇤⇤
(0.007) (0.027) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009 0.009⇤⇤⇤ 0.007⇤⇤
(0.003) (0.009) (0.002) (0.003)
Unopposed 0.023⇤⇤⇤ 0.016⇤⇤ 0.013⇤⇤⇤ 0.018⇤⇤⇤
(0.004) (0.006) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.033⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.004) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.007) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013 0.005⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.008) (0.001) (0.002)
BIPOC Man*Republican 0.004 0.005 0.002 0.030⇤⇤⇤
(0.009) (0.003) (0.005) (0.008)
Constant 0.213⇤⇤⇤ 0.147⇤⇤⇤ 0.137⇤⇤⇤ 0.164⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.409 0.360 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
252

Table A.10: OLS regressions of all 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.0004 0.011 0.007⇤⇤ 0.013⇤⇤⇤
(0.003) (0.025) (0.003) (0.003)
Republican 0.020⇤⇤⇤ 0.023 0.008⇤⇤ 0.017⇤⇤⇤
(0.003) (0.024) (0.003) (0.003)
idExtreme 0.046⇤⇤⇤ 0.016 0.008 0.012⇤
(0.007) (0.020) (0.007) (0.006)
Competitive Seat 0.014⇤⇤⇤ 0.015⇤⇤ 0.004 0.009⇤⇤⇤
(0.003) (0.007) (0.003) (0.003)
Unopposed 0.014⇤⇤⇤ 0.012⇤ 0.006 0.007⇤⇤
(0.004) (0.006) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.010⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.0004
(0.002) (0.005) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.012⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
BIPOC Man*Republican 0.033⇤⇤⇤ 0.001 0.042⇤⇤⇤ 0.011
(0.008) (0.003) (0.008) (0.007)
Constant 0.325⇤⇤⇤ 0.149⇤⇤⇤ 0.176⇤⇤⇤ 0.178⇤⇤⇤
(0.014) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.327 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
253

Table A.11: OLS regressions of all 113 to 116 MCs’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.011⇤⇤⇤ 0.002 0.009⇤⇤⇤ 0.007⇤⇤
(0.004) (0.026) (0.003) (0.004)
Republican 0.035⇤⇤⇤ 0.047⇤⇤⇤ 0.012⇤⇤⇤ 0.012⇤⇤⇤
(0.003) (0.018) (0.002) (0.003)
idExtreme 0.200⇤⇤⇤ 0.129⇤⇤⇤ 0.091⇤⇤⇤ 0.112⇤⇤⇤
(0.008) (0.027) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.008) (0.002) (0.003)
Unopposed 0.024⇤⇤⇤ 0.016⇤⇤ 0.013⇤⇤⇤ 0.018⇤⇤⇤
(0.004) (0.006) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.032⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.005) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.004 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.007) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013 0.006⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.008) (0.001) (0.002)
BIPOC Woman*Republican 0.0005 0.001 0.007 0.017
(0.013) (0.003) (0.008) (0.012)
Constant 0.217⇤⇤⇤ 0.150⇤⇤⇤ 0.141⇤⇤⇤ 0.167⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.409 0.361 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
254

Table A.12: OLS regressions of all 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.014⇤⇤⇤ 0.026 0.016⇤⇤⇤ 0.015⇤⇤⇤
(0.004) (0.024) (0.004) (0.003)
Republican 0.019⇤⇤⇤ 0.016 0.002 0.011⇤⇤⇤
(0.003) (0.024) (0.003) (0.003)
idExtreme 0.049⇤⇤⇤ 0.025 0.015⇤⇤ 0.016⇤⇤
(0.007) (0.020) (0.007) (0.006)
Competitive Seat 0.014⇤⇤⇤ 0.015⇤⇤ 0.004 0.009⇤⇤⇤
(0.003) (0.007) (0.003) (0.003)
Unopposed 0.015⇤⇤⇤ 0.013⇤⇤ 0.005 0.009⇤⇤
(0.004) (0.006) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.00004
(0.002) (0.005) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.011⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
BIPOC Woman*Republican 0.001 0.009⇤⇤ 0.029⇤⇤ 0.002
(0.013) (0.003) (0.012) (0.011)
Constant 0.327⇤⇤⇤ 0.145⇤⇤⇤ 0.178⇤⇤⇤ 0.171⇤⇤⇤
(0.013) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.666 0.328 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
255

Table A.13: OLS regressions of all 113 to 116 MCs’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.010⇤⇤ 0.013 0.005⇤⇤ 0.006⇤
(0.004) (0.026) (0.002) (0.003)
Republican 0.032⇤⇤⇤ 0.047⇤⇤ 0.010⇤⇤⇤ 0.011⇤⇤⇤
(0.003) (0.018) (0.002) (0.003)
idExtreme 0.196⇤⇤⇤ 0.131⇤⇤⇤ 0.088⇤⇤⇤ 0.112⇤⇤⇤
(0.007) (0.027) (0.005) (0.007)
Competitive Seat 0.014⇤⇤⇤ 0.009 0.009⇤⇤⇤ 0.007⇤⇤⇤
(0.003) (0.009) (0.002) (0.003)
Unopposed 0.024⇤⇤⇤ 0.017⇤⇤ 0.013⇤⇤⇤ 0.018⇤⇤⇤
(0.004) (0.007) (0.003) (0.004)
Oppose President 0.035⇤⇤⇤ 0.032⇤⇤⇤ 0.014⇤⇤⇤ 0.032⇤⇤⇤
(0.002) (0.005) (0.001) (0.002)
Minority Party 0.013⇤⇤⇤ 0.005 0.006⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.008) (0.002) (0.002)
Campaign Period 0.018⇤⇤⇤ 0.013 0.005⇤⇤⇤ 0.007⇤⇤⇤
(0.002) (0.008) (0.001) (0.002)
White Woman*Republican 0.016⇤⇤ 0.008⇤⇤ 0.008⇤⇤ 0.007
(0.006) (0.003) (0.004) (0.006)
Constant 0.214⇤⇤⇤ 0.147⇤⇤⇤ 0.138⇤⇤⇤ 0.163⇤⇤⇤
(0.014) (0.034) (0.009) (0.013)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.536 0.410 0.360 0.465
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
256

Table A.14: OLS regressions of all 113 to 116 MCs’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.0004 0.012 0.004 0.005
(0.004) (0.024) (0.004) (0.003)
Republican 0.025⇤⇤⇤ 0.021 0.003 0.015⇤⇤⇤
(0.003) (0.024) (0.003) (0.003)
idExtreme 0.039⇤⇤⇤ 0.011 0.012⇤ 0.008
(0.007) (0.020) (0.007) (0.006)
Competitive Seat 0.013⇤⇤⇤ 0.015⇤⇤ 0.003 0.008⇤⇤⇤
(0.003) (0.007) (0.003) (0.003)
Unopposed 0.013⇤⇤⇤ 0.011 0.006 0.007⇤
(0.004) (0.006) (0.004) (0.004)
Oppose President 0.027⇤⇤⇤ 0.028⇤⇤⇤ 0.009⇤⇤⇤ 0.009⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
Minority Party 0.008⇤⇤⇤ 0.005 0.008⇤⇤⇤ 0.0004
(0.002) (0.005) (0.002) (0.002)
Campaign Period 0.019⇤⇤⇤ 0.012⇤ 0.010⇤⇤⇤ 0.016⇤⇤⇤
(0.002) (0.006) (0.002) (0.002)
White Woman*Republican 0.034⇤⇤⇤ 0.016⇤⇤⇤ 0.016⇤⇤⇤ 0.016⇤⇤⇤
(0.006) (0.003) (0.006) (0.005)
Constant 0.327⇤⇤⇤ 0.146⇤⇤⇤ 0.180⇤⇤⇤ 0.171⇤⇤⇤
(0.013) (0.027) (0.013) (0.012)
Observations 32,768 32,768 32,768 32,768
Adjusted R2 0.667 0.327 0.602 0.440
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
257

A.1.4 Subset by Party
Table A.15: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.012⇤⇤⇤ 0.005 0.008⇤⇤⇤ 0.009⇤⇤⇤
(0.003) (0.040) (0.002) (0.003)
idExtreme 0.342⇤⇤⇤ 0.290⇤⇤⇤ 0.159⇤⇤⇤ 0.220⇤⇤⇤
(0.014) (0.026) (0.009) (0.012)
Competitive Seat 0.004 0.0005 0.003 0.00001
(0.005) (0.037) (0.003) (0.004)
Unopposed 0.021⇤⇤⇤ 0.015⇤⇤ 0.014⇤⇤⇤ 0.014⇤⇤⇤
(0.006) (0.008) (0.004) (0.005)
Oppose President 0.128⇤⇤⇤ 0.089⇤⇤⇤ 0.049⇤⇤⇤ 0.083⇤⇤⇤
(0.004) (0.006) (0.003) (0.004)
Minority Party 0.015⇤⇤⇤ 0.023⇤⇤⇤ 0.024⇤⇤⇤ 0.008⇤⇤
(0.005) (0.004) (0.003) (0.004)
Campaign Period 0.027⇤⇤⇤ 0.016⇤⇤⇤ 0.003 0.005⇤
(0.003) (0.006) (0.002) (0.002)
Constant 0.231⇤⇤⇤ 0.217⇤⇤⇤ 0.192⇤⇤⇤ 0.272⇤⇤⇤
(0.025) (0.049) (0.016) (0.022)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.628 0.481 0.448 0.581
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
DEMS only
258

Table A.16: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.003 0.009 0.006⇤⇤ 0.0001
(0.003) (0.035) (0.003) (0.003)
idExtreme 0.035⇤⇤⇤ 0.018 0.003 0.002
(0.013) (0.035) (0.013) (0.012)
Competitive Seat 0.010⇤⇤ 0.013 0.009⇤⇤ 0.010⇤⇤⇤
(0.004) (0.029) (0.004) (0.004)
Unopposed 0.014⇤⇤⇤ 0.021⇤⇤⇤ 0.003 0.014⇤⇤⇤
(0.006) (0.006) (0.005) (0.005)
Oppose President 0.019⇤⇤⇤ 0.031⇤⇤⇤ 0.031⇤⇤⇤ 0.010⇤⇤⇤
(0.004) (0.005) (0.004) (0.003)
Minority Party 0.017⇤⇤⇤ 0.034⇤⇤⇤ 0.006 0.031⇤⇤⇤
(0.004) (0.005) (0.004) (0.004)
Campaign Period 0.019⇤⇤⇤ 0.007⇤ 0.006⇤⇤ 0.018⇤⇤⇤
(0.003) (0.004) (0.002) (0.002)
Constant 0.379⇤⇤⇤ 0.200⇤⇤⇤ 0.227⇤⇤⇤ 0.179⇤⇤⇤
(0.023) (0.040) (0.023) (0.021)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.687 0.306 0.643 0.462
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
259

Table A.17: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.013⇤⇤⇤ 0.015 0.010⇤⇤⇤ 0.011⇤⇤⇤
(0.004) (0.040) (0.002) (0.003)
idExtreme 0.335⇤⇤⇤ 0.294⇤⇤⇤ 0.154⇤⇤⇤ 0.215⇤⇤⇤
(0.014) (0.027) (0.008) (0.012)
Competitive Seat 0.003 0.0004 0.003 0.0002
(0.005) (0.037) (0.003) (0.004)
Unopposed 0.022⇤⇤⇤ 0.016 0.015⇤⇤⇤ 0.015⇤⇤⇤
(0.006) (0.010) (0.004) (0.005)
Oppose President 0.129⇤⇤⇤ 0.090⇤⇤⇤ 0.050⇤⇤⇤ 0.084⇤⇤⇤
(0.004) (0.007) (0.003) (0.004)
Minority Party 0.016⇤⇤⇤ 0.023⇤⇤⇤ 0.024⇤⇤⇤ 0.008⇤⇤
(0.005) (0.005) (0.003) (0.004)
Campaign Period 0.027⇤⇤⇤ 0.016⇤⇤ 0.003 0.005⇤
(0.003) (0.007) (0.002) (0.002)
Constant 0.215⇤⇤⇤ 0.202⇤⇤⇤ 0.181⇤⇤⇤ 0.260⇤⇤⇤
(0.026) (0.050) (0.016) (0.023)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.628 0.482 0.448 0.581
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
260

Table A.18: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.006⇤ 0.016 0.008⇤⇤ 0.013⇤⇤⇤
(0.003) (0.035) (0.003) (0.003)
idExtreme 0.025⇤⇤ 0.008 0.017 0.009
(0.013) (0.035) (0.012) (0.011)
Competitive Seat 0.010⇤⇤ 0.013 0.009⇤⇤ 0.010⇤⇤⇤
(0.004) (0.030) (0.004) (0.004)
Unopposed 0.015⇤⇤⇤ 0.021⇤⇤⇤ 0.002 0.014⇤⇤⇤
(0.006) (0.007) (0.005) (0.005)
Oppose President 0.020⇤⇤⇤ 0.029⇤⇤⇤ 0.030⇤⇤⇤ 0.011⇤⇤⇤
(0.004) (0.007) (0.004) (0.003)
Minority Party 0.016⇤⇤⇤ 0.033⇤⇤⇤ 0.006 0.030⇤⇤⇤
(0.004) (0.006) (0.004) (0.004)
Campaign Period 0.019⇤⇤⇤ 0.007 0.006⇤⇤ 0.018⇤⇤⇤
(0.003) (0.005) (0.002) (0.002)
Constant 0.383⇤⇤⇤ 0.212⇤⇤⇤ 0.222⇤⇤⇤ 0.191⇤⇤⇤
(0.024) (0.040) (0.023) (0.021)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.687 0.307 0.643 0.463
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
261

Table A.19: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.026⇤⇤⇤ 0.013 0.014⇤⇤⇤ 0.015⇤⇤⇤
(0.004) (0.040) (0.003) (0.004)
idExtreme 0.352⇤⇤⇤ 0.297⇤⇤⇤ 0.161⇤⇤⇤ 0.223⇤⇤⇤
(0.014) (0.026) (0.009) (0.012)
Competitive Seat 0.002 0.0001 0.002 0.001
(0.005) (0.037) (0.003) (0.004)
Unopposed 0.022⇤⇤⇤ 0.016⇤ 0.015⇤⇤⇤ 0.015⇤⇤⇤
(0.006) (0.009) (0.004) (0.005)
Oppose President 0.128⇤⇤⇤ 0.089⇤⇤⇤ 0.049⇤⇤⇤ 0.083⇤⇤⇤
(0.004) (0.007) (0.003) (0.004)
Minority Party 0.016⇤⇤⇤ 0.023⇤⇤⇤ 0.024⇤⇤⇤ 0.009⇤⇤
(0.005) (0.005) (0.003) (0.004)
Campaign Period 0.027⇤⇤⇤ 0.016⇤⇤ 0.003 0.005⇤
(0.003) (0.007) (0.002) (0.002)
Constant 0.217⇤⇤⇤ 0.210⇤⇤⇤ 0.184⇤⇤⇤ 0.263⇤⇤⇤
(0.025) (0.049) (0.016) (0.022)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.629 0.482 0.448 0.582
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
262

Table A.20: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.013⇤⇤⇤ 0.028 0.014⇤⇤⇤ 0.017⇤⇤⇤
(0.004) (0.034) (0.004) (0.003)
idExtreme 0.044⇤⇤⇤ 0.034 0.004 0.019
(0.013) (0.035) (0.013) (0.012)
Competitive Seat 0.010⇤⇤ 0.012 0.010⇤⇤ 0.010⇤⇤
(0.004) (0.029) (0.004) (0.004)
Unopposed 0.014⇤⇤⇤ 0.021⇤⇤⇤ 0.003 0.014⇤⇤⇤
(0.006) (0.008) (0.005) (0.005)
Oppose President 0.019⇤⇤⇤ 0.031⇤⇤⇤ 0.031⇤⇤⇤ 0.009⇤⇤⇤
(0.004) (0.007) (0.004) (0.003)
Minority Party 0.017⇤⇤⇤ 0.035⇤⇤⇤ 0.006 0.031⇤⇤⇤
(0.004) (0.007) (0.004) (0.004)
Campaign Period 0.019⇤⇤⇤ 0.006 0.006⇤⇤ 0.018⇤⇤⇤
(0.003) (0.006) (0.002) (0.002)
Constant 0.373⇤⇤⇤ 0.187⇤⇤⇤ 0.234⇤⇤⇤ 0.173⇤⇤⇤
(0.024) (0.039) (0.023) (0.021)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.688 0.308 0.644 0.463
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
263

Table A.21: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.019⇤⇤⇤ 0.021 0.011⇤⇤⇤ 0.011⇤⇤⇤
(0.004) (0.040) (0.002) (0.003)
idExtreme 0.332⇤⇤⇤ 0.291⇤⇤⇤ 0.151⇤⇤⇤ 0.211⇤⇤⇤
(0.014) (0.027) (0.008) (0.012)
Competitive Seat 0.002 0.001 0.002 0.001
(0.005) (0.037) (0.003) (0.004)
Unopposed 0.023⇤⇤⇤ 0.017⇤ 0.015⇤⇤⇤ 0.016⇤⇤⇤
(0.006) (0.009) (0.004) (0.005)
Oppose President 0.129⇤⇤⇤ 0.090⇤⇤⇤ 0.050⇤⇤⇤ 0.083⇤⇤⇤
(0.004) (0.007) (0.003) (0.004)
Minority Party 0.016⇤⇤⇤ 0.022⇤⇤⇤ 0.024⇤⇤⇤ 0.009⇤⇤
(0.005) (0.005) (0.003) (0.004)
Campaign Period 0.027⇤⇤⇤ 0.016⇤⇤ 0.003 0.005⇤
(0.003) (0.008) (0.002) (0.002)
Constant 0.216⇤⇤⇤ 0.203⇤⇤⇤ 0.184⇤⇤⇤ 0.263⇤⇤⇤
(0.026) (0.049) (0.016) (0.022)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.628 0.483 0.448 0.581
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
264

Table A.22: OLS regressions of all 113 to 116 Democratic MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.001 0.009 0.005 0.001
(0.004) (0.035) (0.003) (0.003)
idExtreme 0.030⇤⇤ 0.001 0.013 0.001
(0.012) (0.036) (0.012) (0.011)
Competitive Seat 0.011⇤⇤ 0.013 0.009⇤⇤ 0.010⇤⇤⇤
(0.004) (0.030) (0.004) (0.004)
Unopposed 0.015⇤⇤⇤ 0.021⇤⇤⇤ 0.003 0.014⇤⇤⇤
(0.006) (0.007) (0.005) (0.005)
Oppose President 0.019⇤⇤⇤ 0.030⇤⇤⇤ 0.030⇤⇤⇤ 0.010⇤⇤⇤
(0.004) (0.006) (0.004) (0.003)
Minority Party 0.016⇤⇤⇤ 0.033⇤⇤⇤ 0.007 0.030⇤⇤⇤
(0.004) (0.006) (0.004) (0.004)
Campaign Period 0.019⇤⇤⇤ 0.007 0.006⇤⇤ 0.018⇤⇤⇤
(0.003) (0.005) (0.002) (0.002)
Constant 0.379⇤⇤⇤ 0.203⇤⇤⇤ 0.226⇤⇤⇤ 0.180⇤⇤⇤
(0.024) (0.040) (0.023) (0.021)
Observations 14,985 14,985 14,985 14,985
Adjusted R2 0.687 0.306 0.643 0.462
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
265

Table A.23: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.013⇤⇤⇤ 0.004 0.002 0.003
(0.005) (0.035) (0.003) (0.004)
idExtreme 0.146⇤⇤⇤ 0.060⇤⇤ 0.077⇤⇤⇤ 0.065⇤⇤⇤
(0.011) (0.024) (0.007) (0.010)
Competitive Seat 0.012⇤⇤⇤ 0.005 0.008⇤⇤⇤ 0.005
(0.004) (0.037) (0.003) (0.004)
Unopposed 0.003 0.002 0.004 0.013⇤⇤
(0.007) (0.009) (0.004) (0.006)
Oppose President 0.018⇤⇤⇤ 0.006 0.004 0.002
(0.004) (0.007) (0.002) (0.003)
Minority Party 0.044⇤⇤⇤ 0.029⇤⇤⇤ 0.030⇤⇤⇤ 0.009⇤⇤
(0.005) (0.005) (0.003) (0.004)
Campaign Period 0.012⇤⇤⇤ 0.011 0.008⇤⇤⇤ 0.009⇤⇤⇤
(0.003) (0.008) (0.002) (0.002)
Constant 0.134⇤⇤⇤ 0.072 0.093⇤⇤⇤ 0.050⇤⇤⇤
(0.020) (0.045) (0.012) (0.018)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.437 0.290 0.280 0.350
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
GOP only
266

Table A.24: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.014⇤⇤⇤ 0.016 0.003 0.005
(0.004) (0.034) (0.004) (0.004)
idExtreme 0.058⇤⇤⇤ 0.017 0.032⇤⇤⇤ 0.030⇤⇤⇤
(0.010) (0.035) (0.010) (0.009)
Competitive Seat 0.014⇤⇤⇤ 0.016 0.004 0.007⇤
(0.004) (0.028) (0.004) (0.004)
Unopposed 0.015⇤⇤ 0.003 0.009 0.002
(0.007) (0.008) (0.007) (0.006)
Oppose President 0.102⇤⇤⇤ 0.052⇤⇤⇤ 0.053⇤⇤⇤ 0.034⇤⇤⇤
(0.004) (0.007) (0.004) (0.003)
Minority Party 0.066⇤⇤⇤ 0.061⇤⇤⇤ 0.036⇤⇤⇤ 0.050⇤⇤⇤
(0.005) (0.007) (0.005) (0.004)
Campaign Period 0.020⇤⇤⇤ 0.016⇤⇤⇤ 0.014⇤⇤⇤ 0.015⇤⇤⇤
(0.003) (0.006) (0.003) (0.002)
Constant 0.359⇤⇤⇤ 0.121⇤⇤⇤ 0.172⇤⇤⇤ 0.191⇤⇤⇤
(0.020) (0.038) (0.019) (0.017)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.654 0.352 0.574 0.424
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
267

Table A.25: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.014⇤ 0.016 0.006 0.037⇤⇤⇤
(0.009) (0.034) (0.005) (0.008)
idExtreme 0.150⇤⇤⇤ 0.059⇤⇤ 0.077⇤⇤⇤ 0.059⇤⇤⇤
(0.011) (0.024) (0.007) (0.009)
Competitive Seat 0.013⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.005
(0.004) (0.036) (0.003) (0.004)
Unopposed 0.003 0.002 0.004 0.014⇤⇤
(0.007) (0.017) (0.004) (0.006)
Oppose President 0.018⇤⇤⇤ 0.006 0.004 0.002
(0.004) (0.013) (0.002) (0.003)
Minority Party 0.045⇤⇤⇤ 0.029⇤⇤⇤ 0.031⇤⇤⇤ 0.009⇤⇤
(0.005) (0.010) (0.003) (0.004)
Campaign Period 0.012⇤⇤⇤ 0.011 0.008⇤⇤⇤ 0.010⇤⇤⇤
(0.003) (0.015) (0.002) (0.002)
Constant 0.121⇤⇤⇤ 0.066 0.090⇤⇤⇤ 0.040⇤⇤
(0.020) (0.044) (0.012) (0.018)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.437 0.290 0.280 0.351
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
268

Table A.26: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.032⇤⇤⇤ 0.009 0.028⇤⇤⇤ 0.029⇤⇤⇤
(0.008) (0.034) (0.008) (0.007)
idExtreme 0.071⇤⇤⇤ 0.027 0.028⇤⇤⇤ 0.038⇤⇤⇤
(0.010) (0.034) (0.010) (0.009)
Competitive Seat 0.015⇤⇤⇤ 0.017 0.004 0.008⇤⇤
(0.004) (0.028) (0.004) (0.004)
Unopposed 0.014⇤⇤ 0.003 0.009 0.002
(0.007) (0.016) (0.007) (0.006)
Oppose President 0.102⇤⇤⇤ 0.052⇤⇤⇤ 0.054⇤⇤⇤ 0.035⇤⇤⇤
(0.004) (0.014) (0.004) (0.003)
Minority Party 0.067⇤⇤⇤ 0.062⇤⇤⇤ 0.036⇤⇤⇤ 0.050⇤⇤⇤
(0.005) (0.015) (0.005) (0.004)
Campaign Period 0.020⇤⇤⇤ 0.016 0.014⇤⇤⇤ 0.016⇤⇤⇤
(0.003) (0.012) (0.003) (0.002)
Constant 0.355⇤⇤⇤ 0.110⇤⇤⇤ 0.180⇤⇤⇤ 0.193⇤⇤⇤
(0.019) (0.037) (0.019) (0.017)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.654 0.352 0.574 0.424
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
269

Table A.27: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.016 0.002 0.002 0.027⇤⇤
(0.014) (0.034) (0.008) (0.012)
idExtreme 0.152⇤⇤⇤ 0.062⇤⇤⇤ 0.078⇤⇤⇤ 0.065⇤⇤⇤
(0.011) (0.024) (0.006) (0.009)
Competitive Seat 0.013⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.005
(0.004) (0.037) (0.003) (0.004)
Unopposed 0.004 0.002 0.004 0.014⇤⇤
(0.007) (0.032) (0.004) (0.006)
Oppose President 0.018⇤⇤⇤ 0.006 0.004 0.002
(0.004) (0.020) (0.002) (0.003)
Minority Party 0.045⇤⇤⇤ 0.029⇤⇤ 0.031⇤⇤⇤ 0.009⇤⇤
(0.005) (0.013) (0.003) (0.004)
Campaign Period 0.012⇤⇤⇤ 0.011 0.008⇤⇤⇤ 0.009⇤⇤⇤
(0.003) (0.023) (0.002) (0.002)
Constant 0.123⇤⇤⇤ 0.069 0.091⇤⇤⇤ 0.046⇤⇤⇤
(0.020) (0.044) (0.012) (0.018)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.437 0.290 0.280 0.350
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
270

Table A.28: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.001 0.017 0.048⇤⇤⇤ 0.004
(0.013) (0.034) (0.013) (0.012)
idExtreme 0.065⇤⇤⇤ 0.024 0.031⇤⇤⇤ 0.033⇤⇤⇤
(0.010) (0.034) (0.010) (0.009)
Competitive Seat 0.015⇤⇤⇤ 0.017 0.003 0.008⇤⇤
(0.004) (0.028) (0.004) (0.004)
Unopposed 0.015⇤⇤ 0.004 0.008 0.002
(0.007) (0.031) (0.007) (0.006)
Oppose President 0.102⇤⇤⇤ 0.052⇤⇤⇤ 0.053⇤⇤⇤ 0.034⇤⇤⇤
(0.004) (0.019) (0.004) (0.003)
Minority Party 0.067⇤⇤⇤ 0.061⇤⇤⇤ 0.036⇤⇤⇤ 0.050⇤⇤⇤
(0.005) (0.024) (0.005) (0.004)
Campaign Period 0.020⇤⇤⇤ 0.016 0.014⇤⇤⇤ 0.015⇤⇤⇤
(0.003) (0.020) (0.003) (0.002)
Constant 0.348⇤⇤⇤ 0.107⇤⇤⇤ 0.176⇤⇤⇤ 0.187⇤⇤⇤
(0.019) (0.038) (0.019) (0.017)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.654 0.352 0.574 0.424
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
271

Table A.29: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.010⇤ 0.002 0.001 0.015⇤⇤⇤
(0.005) (0.034) (0.003) (0.005)
idExtreme 0.150⇤⇤⇤ 0.063⇤⇤⇤ 0.078⇤⇤⇤ 0.070⇤⇤⇤
(0.011) (0.024) (0.007) (0.009)
Competitive Seat 0.013⇤⇤⇤ 0.005 0.009⇤⇤⇤ 0.006
(0.004) (0.036) (0.003) (0.004)
Unopposed 0.002 0.002 0.004 0.014⇤⇤
(0.007) (0.011) (0.004) (0.006)
Oppose President 0.018⇤⇤⇤ 0.006 0.004 0.002
(0.004) (0.008) (0.002) (0.003)
Minority Party 0.045⇤⇤⇤ 0.029⇤⇤⇤ 0.031⇤⇤⇤ 0.010⇤⇤
(0.005) (0.005) (0.003) (0.004)
Campaign Period 0.012⇤⇤⇤ 0.011 0.008⇤⇤⇤ 0.010⇤⇤⇤
(0.003) (0.009) (0.002) (0.002)
Constant 0.124⇤⇤⇤ 0.069 0.091⇤⇤⇤ 0.046⇤⇤⇤
(0.020) (0.044) (0.012) (0.018)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.437 0.290 0.280 0.350
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
272

Table A.30: OLS regressions of all 113 to 116 Republican MCs’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.032⇤⇤⇤ 0.023 0.015⇤⇤⇤ 0.019⇤⇤⇤
(0.005) (0.034) (0.005) (0.005)
idExtreme 0.056⇤⇤⇤ 0.019 0.037⇤⇤⇤ 0.028⇤⇤⇤
(0.010) (0.034) (0.010) (0.009)
Competitive Seat 0.014⇤⇤⇤ 0.016 0.005 0.007⇤
(0.004) (0.028) (0.004) (0.004)
Unopposed 0.013⇤ 0.001 0.008 0.001
(0.007) (0.008) (0.007) (0.006)
Oppose President 0.102⇤⇤⇤ 0.052⇤⇤⇤ 0.053⇤⇤⇤ 0.034⇤⇤⇤
(0.004) (0.009) (0.004) (0.003)
Minority Party 0.066⇤⇤⇤ 0.061⇤⇤⇤ 0.035⇤⇤⇤ 0.050⇤⇤⇤
(0.005) (0.007) (0.005) (0.004)
Campaign Period 0.020⇤⇤⇤ 0.016⇤⇤ 0.014⇤⇤⇤ 0.016⇤⇤⇤
(0.003) (0.007) (0.003) (0.002)
Constant 0.350⇤⇤⇤ 0.110⇤⇤⇤ 0.175⇤⇤⇤ 0.188⇤⇤⇤
(0.019) (0.037) (0.019) (0.017)
Observations 17,783 17,783 17,783 17,783
Adjusted R2 0.655 0.352 0.574 0.424
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Congress, month, and state fixed effects, leadership,
age, monthly avg. tweet WC, and logged total n
tweets omitted from table. Standard errors are clustered by MC and congress.
273

A.1.5 Model estimates of separate race/gender effects
Table A.31: Subgroup main effects (Dems only) from separate OLS regressions of 113 to 116 MCs’
monthly logged average use of negative emotive appeals on Twitter
n adj. R2
negative
(1) White Man 0.012*** 14,985 0.628
(0.003)
(2) BIPOC Man -0.013*** 14,985 0.628
(0.004)
(3) BIPOC Woman -0.026*** 14,985 0.629
(0.004)
(4) White Woman 0.019*** 14,985 0.628
(0.004)
anger
(1) White Man 0.005 14,985 0.481
(0.040)
(2) BIPOC Man -0.015 14,985 0.482
(0.040)
(3) BIPOC Woman -0.013 14,985 0.482
(0.040)
(4) White Woman 0.021 14,985 0.483
(0.040)
disgust
(1) White Man 0.008*** 14,985 0.448
(0.002)
(2) BIPOC Man -0.010*** 14,985 0.448
(0.002)
(3) BIPOC Woman -0.014*** 14,985 0.448
(0.003)
(4) White Woman 0.011*** 14,985 0.448
(0.002)
fear
(1) White Man 0.009*** 14,985 0.581
(0.003)
(2) BIPOC Man -0.011*** 14,985 0.581
(0.003)
(3) BIPOC Woman -0.015*** 14,985 0.582
(0.004)
(4) White Woman 0.011*** 14,985 0.581
(0.003)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include MC/district vars (idExtreme, competitive seat, unopposed, oppose president, minority party, period, leadership, age), congress, month, and state fixed effects, monthly avg. tweet WC, and logged total n tweets
omitted from table. Standard errors are clustered by MC
and congress.
274

Table A.32: Subgroup main effects (GOP only) from separate OLS regressions of 113 to 116 MCs’
monthly logged average use of negative emotive appeals on Twitter
n adj. R2
negative
(1) White Man 0.013*** 17,783 0.437
(0.005)
(2) BIPOC Man -0.014* 17,783 0.437
(0.009)
(3) BIPOC Woman -0.016 17,783 0.437
(0.014)
(4) White Woman -0.010* 17,783 0.437
(0.005)
anger
(1) White Man 0.004 17,783 0.290
(0.035)
(2) BIPOC Man -0.016 17,783 0.290
(0.034)
(3) BIPOC Woman -0.002 17,783 0.290
(0.034)
(4) White Woman 0.002 17,783 0.290
(0.034)
disgust
(1) White Man 0.002 17,783 0.280
(0.003)
(2) BIPOC Man -0.006 17,783 0.280
(0.005)
(3) BIPOC Woman -0.002 17,783 0.280
(0.008)
(4) White Woman -0.001 17,783 0.280
(0.003)
fear
(1) White Man 0.003 17,783 0.350
(0.004)
(2) BIPOC Man -0.037*** 17,783 0.351
(0.008)
(3) BIPOC Woman -0.027** 17,783 0.350
(0.012)
(4) White Woman 0.015*** 17,783 0.350
(0.005)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include MC/district vars (idExtreme, competitive seat, unopposed, oppose president, minority party, period, leadership, age), congress, month, and state fixed effects, monthly avg. tweet WC, and logged total n tweets
omitted from table. Standard errors are clustered by MC
and congress.
275

negative appeals
276

Table A.33: Subgroup main effects (GOP only) from separate OLS regressions of 113 to 116 MCs’
monthly logged average use of positive emotive appeals on Twitter
n adj. R2
positive
(1) White Man -0.014*** 17,783 0.654 (0.004)
(2) BIPOC Man -0.032*** 17,783 0.654 (0.008)
(3) BIPOC Woman -0.001 17,783 0.654 (0.013)
(4) White Woman 0.032*** 17,783 0.655 (0.005)
joy
(1) White Man -0.016 17,783 0.352 (0.034)
(2) BIPOC Man -0.009 17,783 0.352 (0.034)
(3) BIPOC Woman 0.017 17,783 0.352 (0.034)
(4) White Woman 0.023 17,783 0.352 (0.034)
trust
(1) White Man 0.003 17,783 0.574 (0.004)
(2) BIPOC Man -0.028*** 17,783 0.574 (0.008)
(3) BIPOC Woman -0.048*** 17,783 0.574 (0.013)
(4) White Woman 0.015*** 17,783 0.574 (0.005)
anticipation
(1) White Man -0.005 17,783 0.424 (0.004)
(2) BIPOC Man -0.029*** 17,783 0.424 (0.007)
(3) BIPOC Woman -0.004 17,783 0.424 (0.012)
(4) White Woman 0.019*** 17,783 0.424 (0.005)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include MC/district vars (idExtreme, competitive seat, unopposed, oppose president, minority party, period, leadership, age), congress, month, and state fixed effects, monthly avg. tweet WC, and logged total n tweets
omitted from table. Standard errors are clustered by MC
and congress.
positive appeals
277

Table A.34: Subgroup main effects (Dems only) from separate OLS regressions of 113 to 116 MCs’
monthly logged average use of positive emotive appeals on Twitter
n adj. R2
positive
(1) White Man -0.003 14,985 0.687 (0.003)
(2) BIPOC Man -0.006* 14,985 0.687 (0.003)
(3) BIPOC Woman 0.013*** 14,985 0.688 (0.004)
(4) White Woman 0.001 14,985 0.687 (0.004)
joy
(1) White Man -0.009 14,985 0.306 (0.035)
(2) BIPOC Man -0.016 14,985 0.307 (0.035)
(3) BIPOC Woman 0.028 14,985 0.308 (0.034)
(4) White Woman 0.009 14,985 0.306 (0.035)
trust
(1) White Man 0.006** 14,985 0.643 (0.003)
(2) BIPOC Man 0.008** 14,985 0.643 (0.003)
(3) BIPOC Woman -0.014*** 14,985 0.644 (0.004)
(4) White Woman -0.005 14,985 0.643 (0.003)
anticipation
(1) White Man 0.0001 14,985 0.462 (0.003)
(2) BIPOC Man -0.013*** 14,985 0.463 (0.003)
(3) BIPOC Woman 0.017*** 14,985 0.463 (0.003)
(4) White Woman 0.001 14,985 0.462 (0.003)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include MC/district vars (idExtreme, competitive seat, unopposed, oppose president, minority party, period, leadership, age), congress, month, and state fixed effects, monthly avg. tweet WC, and logged total n tweets
omitted from table. Standard errors are clustered by MC
and congress.
278

A.2 2020 incumbent and challenger candidate accounts on Twitter
A.2.1 raceGen factor models
Table A.35: OLS regressions of all 2020 House candidate accounts’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.025⇤⇤⇤ 0.011 0.016⇤⇤⇤ 0.007
(0.005) (0.032) (0.004) (0.005)
BIPOC Woman (fct) 0.023⇤⇤⇤ 0.007 0.012⇤⇤⇤ 0.012⇤⇤
(0.006) (0.024) (0.004) (0.005)
White Woman (fct) 0.032⇤⇤⇤ 0.019 0.024⇤⇤⇤ 0.018⇤⇤⇤
(0.005) (0.032) (0.003) (0.005)
Republican 0.017⇤⇤⇤ 0.030⇤⇤⇤ 0.016⇤⇤⇤ 0.004
(0.004) (0.010) (0.003) (0.004)
Incumbent 0.027⇤⇤⇤ 0.025⇤⇤⇤ 0.050⇤⇤⇤ 0.049⇤⇤⇤
(0.006) (0.006) (0.004) (0.006)
Competitiveness Scale 0.003 0.007 0.003⇤⇤ 0.005⇤⇤
(0.002) (0.009) (0.001) (0.002)
District Party Fit 0.001 0.007 0.002 0.014⇤⇤⇤
(0.005) (0.010) (0.003) (0.005)
Constant 0.033 0.023 0.007 0.072⇤⇤⇤
(0.021) (0.035) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.319 0.221 0.161 0.302
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
279

Table A.36: OLS regressions of all 2020 House candidate accounts’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.017⇤⇤⇤ 0.019 0.002 0.007
(0.005) (0.027) (0.005) (0.005)
BIPOC Woman (fct) 0.016⇤⇤⇤ 0.030 0.007 0.018⇤⇤⇤
(0.005) (0.034) (0.005) (0.005)
White Woman (fct) 0.019⇤⇤⇤ 0.024 0.004 0.013⇤⇤⇤
(0.004) (0.024) (0.004) (0.004)
Republican 0.008⇤⇤ 0.024⇤⇤⇤ 0.041⇤⇤⇤ 0.012⇤⇤⇤
(0.004) (0.009) (0.004) (0.003)
Incumbent 0.087⇤⇤⇤ 0.053⇤⇤⇤ 0.076⇤⇤⇤ 0.048⇤⇤⇤
(0.006) (0.008) (0.006) (0.005)
Competitiveness Scale 0.002 0.001 0.003⇤ 0.0002
(0.002) (0.008) (0.002) (0.002)
District Party Fit 0.016⇤⇤⇤ 0.007 0.008⇤ 0.001
(0.004) (0.008) (0.004) (0.004)
Constant 0.307⇤⇤⇤ 0.160⇤⇤⇤ 0.224⇤⇤⇤ 0.150⇤⇤⇤
(0.019) (0.028) (0.019) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.230 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
280

Table A.37: OLS regressions of 2020 House challenger accounts’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.031⇤⇤⇤ 0.007 0.013⇤⇤⇤ 0.002
(0.007) (0.028) (0.005) (0.006)
BIPOC Woman (fct) 0.049⇤⇤⇤ 0.024 0.021⇤⇤⇤ 0.030⇤⇤⇤
(0.007) (0.022) (0.005) (0.007)
White Woman (fct) 0.043⇤⇤⇤ 0.023 0.030⇤⇤⇤ 0.026⇤⇤⇤
(0.006) (0.032) (0.004) (0.005)
Republican 0.008 0.009 0.029⇤⇤⇤ 0.025⇤⇤⇤
(0.005) (0.013) (0.003) (0.005)
Competitiveness Scale 0.001 0.006 0.004⇤⇤ 0.004⇤
(0.003) (0.010) (0.002) (0.002)
District Party Fit 0.011⇤⇤ 0.001 0.003 0.022⇤⇤⇤
(0.005) (0.007) (0.004) (0.005)
Constant 0.066⇤⇤ 0.015 0.015 0.113⇤⇤⇤
(0.026) (0.038) (0.017) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.330 0.229 0.179 0.296
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
281

Table A.38: OLS regressions of 2020 House challenger accounts’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.027⇤⇤⇤ 0.037 0.006 0.015⇤⇤
(0.006) (0.032) (0.006) (0.006)
BIPOC Woman (fct) 0.029⇤⇤⇤ 0.045 0.008 0.024⇤⇤⇤
(0.007) (0.033) (0.007) (0.006)
White Woman (fct) 0.021⇤⇤⇤ 0.027 0.009 0.013⇤⇤
(0.006) (0.031) (0.006) (0.005)
Republican 0.015⇤⇤⇤ 0.025⇤⇤ 0.050⇤⇤⇤ 0.010⇤⇤
(0.005) (0.012) (0.005) (0.004)
Competitiveness Scale 0.002 0.002 0.001 0.001
(0.002) (0.011) (0.002) (0.002)
District Party Fit 0.013⇤⇤⇤ 0.003 0.009⇤ 0.005
(0.005) (0.012) (0.005) (0.005)
Constant 0.265⇤⇤⇤ 0.122⇤⇤⇤ 0.179⇤⇤⇤ 0.148⇤⇤⇤
(0.024) (0.037) (0.024) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.553 0.238 0.465 0.356
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
282

Table A.39: OLS regressions of 2020 House incumbent accounts’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.029⇤⇤⇤ 0.027 0.025⇤⇤⇤ 0.028⇤⇤⇤
(0.009) (0.077) (0.006) (0.009)
BIPOC Woman (fct) 0.005 0.002 0.008 0.004
(0.011) (0.054) (0.007) (0.010)
White Woman (fct) 0.021⇤⇤ 0.021 0.019⇤⇤⇤ 0.007
(0.009) (0.074) (0.006) (0.008)
Republican 0.049⇤⇤⇤ 0.059⇤⇤⇤ 0.006 0.023⇤⇤⇤
(0.008) (0.016) (0.005) (0.007)
Competitiveness Scale 0.0003 0.004 0.001 0.001
(0.004) (0.009) (0.002) (0.003)
District Party Fit 0.012 0.018 0.028⇤ 0.018
(0.024) (0.013) (0.016) (0.022)
Constant 0.215⇤⇤⇤ 0.115 0.138⇤⇤⇤ 0.251⇤⇤⇤
(0.047) (0.083) (0.031) (0.043)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.307 0.251 0.185 0.325
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
283

Table A.40: OLS regressions of 2020 House incumbent accounts’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.001 0.011 0.015⇤ 0.010
(0.008) (0.065) (0.008) (0.008)
BIPOC Woman (fct) 0.003 0.014 0.026⇤⇤⇤ 0.011
(0.009) (0.073) (0.009) (0.009)
White Woman (fct) 0.019⇤⇤ 0.021 0.001 0.016⇤⇤
(0.008) (0.065) (0.008) (0.008)
Republican 0.005 0.008 0.031⇤⇤⇤ 0.024⇤⇤⇤
(0.007) (0.012) (0.007) (0.007)
Competitiveness Scale 0.008⇤⇤ 0.002 0.012⇤⇤⇤ 0.001
(0.003) (0.010) (0.003) (0.003)
District Party Fit 0.088⇤⇤⇤ 0.058⇤⇤⇤ 0.039⇤ 0.063⇤⇤⇤
(0.021) (0.009) (0.021) (0.021)
Constant 0.508⇤⇤⇤ 0.291⇤⇤⇤ 0.382⇤⇤⇤ 0.177⇤⇤⇤
(0.041) (0.054) (0.041) (0.040)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.466 0.189 0.385 0.272
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
284

Table A.41: OLS regressions of 2020 House Democratic candidate accounts’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.043⇤⇤⇤ 0.022 0.030⇤⇤⇤ 0.030⇤⇤⇤
(0.007) (0.031) (0.004) (0.006)
BIPOC Woman (fct) 0.034⇤⇤⇤ 0.012 0.017⇤⇤⇤ 0.025⇤⇤⇤
(0.007) (0.024) (0.005) (0.007)
White Woman (fct) 0.037⇤⇤⇤ 0.023 0.029⇤⇤⇤ 0.025⇤⇤⇤
(0.006) (0.040) (0.004) (0.006)
Incumbent 0.029⇤⇤⇤ 0.023⇤⇤ 0.060⇤⇤⇤ 0.063⇤⇤⇤
(0.008) (0.010) (0.005) (0.007)
Competitiveness Scale 0.008⇤⇤⇤ 0.009 0.005⇤⇤⇤ 0.009⇤⇤⇤
(0.003) (0.008) (0.002) (0.003)
District Party Fit 0.035⇤⇤⇤ 0.021⇤ 0.014⇤⇤⇤ 0.045⇤⇤⇤
(0.007) (0.011) (0.005) (0.007)
Constant 0.062⇤⇤ 0.005 0.020 0.060⇤⇤
(0.029) (0.041) (0.019) (0.027)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.349 0.238 0.187 0.337
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
285

Table A.42: OLS regressions of 2020 House Democratic candidate accounts’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.025⇤⇤⇤ 0.014 0.004 0.007
(0.006) (0.035) (0.006) (0.006)
BIPOC Woman (fct) 0.014⇤⇤ 0.024 0.005 0.015⇤⇤
(0.006) (0.032) (0.006) (0.006)
White Woman (fct) 0.017⇤⇤⇤ 0.016 0.002 0.012⇤⇤
(0.005) (0.027) (0.005) (0.005)
Incumbent 0.086⇤⇤⇤ 0.053⇤⇤⇤ 0.082⇤⇤⇤ 0.054⇤⇤⇤
(0.007) (0.008) (0.007) (0.007)
Competitiveness Scale 0.001 0.007 0.002 0.006⇤⇤
(0.002) (0.009) (0.002) (0.002)
District Party Fit 0.017⇤⇤⇤ 0.013⇤ 0.001 0.009
(0.006) (0.007) (0.006) (0.006)
Constant 0.325⇤⇤⇤ 0.220⇤⇤⇤ 0.247⇤⇤⇤ 0.152⇤⇤⇤
(0.026) (0.037) (0.026) (0.025)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.581 0.242 0.482 0.362
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
286

Table A.43: OLS regressions of 2020 House Republican candidate accounts’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (fct) 0.025⇤⇤ 0.017 0.005 0.002
(0.011) (0.049) (0.007) (0.009)
BIPOC Woman (fct) 0.043⇤⇤⇤ 0.034 0.023⇤⇤⇤ 0.029⇤⇤⇤
(0.013) (0.036) (0.008) (0.011)
White Woman (fct) 0.046⇤⇤⇤ 0.028 0.033⇤⇤⇤ 0.029⇤⇤⇤
(0.009) (0.045) (0.006) (0.008)
Incumbent 0.037⇤⇤⇤ 0.037⇤⇤ 0.041⇤⇤⇤ 0.046⇤⇤⇤
(0.011) (0.015) (0.007) (0.009)
Competitiveness Scale 0.002 0.004 0.001 0.00002
(0.003) (0.010) (0.002) (0.003)
District Party Fit 0.018⇤⇤ 0.024 0.011⇤ 0.009
(0.009) (0.015) (0.006) (0.008)
Constant 0.006 0.020 0.051⇤⇤⇤ 0.050⇤
(0.029) (0.050) (0.020) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.291 0.194 0.166 0.284
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
287

Table A.44: OLS regressions of 2020 House Republican candidate accounts’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (fct) 0.0002 0.026 0.006 0.010
(0.009) (0.038) (0.010) (0.009)
BIPOC Woman (fct) 0.019⇤ 0.039 0.011 0.025⇤⇤
(0.011) (0.051) (0.011) (0.011)
White Woman (fct) 0.034⇤⇤⇤ 0.041 0.015⇤ 0.016⇤⇤
(0.008) (0.036) (0.008) (0.008)
Incumbent 0.087⇤⇤⇤ 0.047⇤⇤ 0.073⇤⇤⇤ 0.039⇤⇤⇤
(0.009) (0.019) (0.010) (0.009)
Competitiveness Scale 0.002 0.007 0.005⇤ 0.005⇤
(0.003) (0.018) (0.003) (0.003)
District Party Fit 0.004 0.004 0.011 0.004
(0.008) (0.017) (0.008) (0.008)
Constant 0.311⇤⇤⇤ 0.130⇤⇤⇤ 0.252⇤⇤⇤ 0.126⇤⇤⇤
(0.026) (0.039) (0.027) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.559 0.227 0.475 0.322
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type, avg.
followers, monthly avg. tweet WC, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
288

A.2.2 woman*white interaction term models
Table A.45: OLS regressions of 2020 House candidate accounts’ monthly logged average use of
negative emotive appeals on Twitter (woman*White)
negative anger disgust fear
(1) (2) (3) (4)
Woman 0.003 0.003 0.004 0.005
(0.007) (0.033) (0.004) (0.006)
White 0.025⇤⇤⇤ 0.011 0.016⇤⇤⇤ 0.007
(0.005) (0.024) (0.004) (0.005)
Republican 0.017⇤⇤⇤ 0.030 0.016⇤⇤⇤ 0.004
(0.004) (0.032) (0.003) (0.004)
Incumbent 0.027⇤⇤⇤ 0.025⇤⇤ 0.050⇤⇤⇤ 0.049⇤⇤⇤
(0.006) (0.012) (0.004) (0.006)
Competitive Scale 0.003 0.007 0.003⇤⇤ 0.005⇤⇤
(0.002) (0.009) (0.001) (0.002)
District Party Fit 0.001 0.007 0.002 0.014⇤⇤⇤
(0.005) (0.007) (0.003) (0.005)
Woman*White 0.034⇤⇤⇤ 0.022⇤⇤⇤ 0.028⇤⇤⇤ 0.013⇤
(0.008) (0.005) (0.006) (0.008)
Constant 0.059⇤⇤⇤ 0.012 0.009 0.079⇤⇤⇤
(0.021) (0.035) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.319 0.221 0.161 0.302
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
289

Table A.46: OLS regressions of 2020 House candidate accounts’ monthly logged average use of
positive emotive appeals on Twitter (woman*White)
positive joy trust anticipation
(1) (2) (3) (4)
Woman 0.001 0.011 0.005 0.011⇤
(0.006) (0.027) (0.006) (0.006)
White 0.017⇤⇤⇤ 0.019 0.002 0.007
(0.005) (0.033) (0.005) (0.005)
Republican 0.008⇤⇤ 0.024 0.041⇤⇤⇤ 0.012⇤⇤⇤
(0.004) (0.024) (0.004) (0.003)
Incumbent 0.087⇤⇤⇤ 0.053⇤⇤⇤ 0.076⇤⇤⇤ 0.048⇤⇤⇤
(0.006) (0.010) (0.006) (0.005)
Competitive Scale 0.002 0.001 0.003⇤ 0.0002
(0.002) (0.009) (0.002) (0.002)
District Party Fit 0.016⇤⇤⇤ 0.007 0.008⇤ 0.001
(0.004) (0.009) (0.004) (0.004)
Woman*White 0.020⇤⇤⇤ 0.012⇤ 0.009 0.002
(0.007) (0.007) (0.007) (0.007)
Constant 0.324⇤⇤⇤ 0.179⇤⇤⇤ 0.222⇤⇤⇤ 0.157⇤⇤⇤
(0.019) (0.028) (0.019) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.230 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
290

A.2.3 raceGen group*party interaction term models
Table A.47: OLS regressions of all 2020 candidates’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.036⇤⇤⇤ 0.017 0.024⇤⇤⇤ 0.023⇤⇤⇤
(0.005) (0.032) (0.003) (0.005)
Republican 0.007 0.025 0.023⇤⇤⇤ 0.016⇤⇤⇤
(0.006) (0.024) (0.004) (0.005)
Incumbent 0.028⇤⇤⇤ 0.026 0.051⇤⇤⇤ 0.049⇤⇤⇤
(0.006) (0.032) (0.004) (0.006)
Competitive Scale 0.003⇤ 0.007 0.003⇤⇤ 0.005⇤⇤⇤
(0.002) (0.009) (0.001) (0.002)
District Party Fit 0.005 0.005 0.001 0.017⇤⇤⇤
(0.005) (0.007) (0.003) (0.005)
White Man*Republican 0.019⇤⇤ 0.009 0.013⇤⇤ 0.023⇤⇤⇤
(0.008) (0.006) (0.005) (0.007)
Constant 0.062⇤⇤⇤ 0.010 0.011 0.086⇤⇤⇤
(0.021) (0.035) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.319 0.221 0.161 0.302
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
291

Table A.48: OLS regressions of all 2020 candidates’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.014⇤⇤⇤ 0.014 0.006 0.010⇤⇤
(0.005) (0.027) (0.005) (0.005)
Republican 0.013⇤⇤ 0.035 0.049⇤⇤⇤ 0.009⇤
(0.005) (0.033) (0.005) (0.005)
Incumbent 0.087⇤⇤⇤ 0.051⇤⇤ 0.075⇤⇤⇤ 0.048⇤⇤⇤
(0.006) (0.024) (0.006) (0.005)
Competitive Scale 0.002 0.0004 0.003⇤ 0.0003
(0.002) (0.007) (0.002) (0.002)
District Party Fit 0.017⇤⇤⇤ 0.010⇤ 0.010⇤⇤ 0.002
(0.005) (0.006) (0.005) (0.004)
White Man*Republican 0.008 0.022⇤⇤⇤ 0.014⇤ 0.006
(0.007) (0.007) (0.007) (0.007)
Constant 0.323⇤⇤⇤ 0.180⇤⇤⇤ 0.220⇤⇤⇤ 0.161⇤⇤⇤
(0.019) (0.028) (0.019) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.230 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
292

Table A.49: OLS regressions of all 2020 candidates’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.019⇤⇤⇤ 0.009 0.013⇤⇤⇤ 0.010⇤
(0.006) (0.032) (0.004) (0.005)
Republican 0.012⇤⇤⇤ 0.028 0.018⇤⇤⇤ 0.004
(0.004) (0.024) (0.003) (0.004)
Incumbent 0.022⇤⇤⇤ 0.023 0.047⇤⇤⇤ 0.046⇤⇤⇤
(0.006) (0.032) (0.004) (0.006)
Competitive Scale 0.004⇤ 0.007 0.003⇤⇤ 0.005⇤⇤⇤
(0.002) (0.011) (0.001) (0.002)
District Party Fit 0.005 0.004 0.001 0.017⇤⇤⇤
(0.005) (0.009) (0.003) (0.005)
BIPOC Man*Republican 0.021⇤ 0.014⇤⇤ 0.021⇤⇤⇤ 0.033⇤⇤⇤
(0.011) (0.006) (0.007) (0.010)
Constant 0.041⇤⇤ 0.020 0.003 0.074⇤⇤⇤
(0.021) (0.035) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.317 0.220 0.159 0.302
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
293

Table A.50: OLS regressions of all 2020 candidates’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.013⇤⇤ 0.001 0.007 0.002
(0.005) (0.027) (0.005) (0.005)
Republican 0.005 0.014 0.039⇤⇤⇤ 0.017⇤⇤⇤
(0.004) (0.033) (0.004) (0.004)
Incumbent 0.084⇤⇤⇤ 0.048⇤⇤ 0.076⇤⇤⇤ 0.046⇤⇤⇤
(0.006) (0.024) (0.006) (0.005)
Competitive Scale 0.002 0.0005 0.003⇤ 0.0002
(0.002) (0.008) (0.002) (0.002)
District Party Fit 0.013⇤⇤⇤ 0.006 0.009⇤ 0.001
(0.005) (0.007) (0.005) (0.004)
BIPOC Man*Republican 0.015 0.017⇤⇤ 0.015 0.007
(0.010) (0.007) (0.010) (0.009)
Constant 0.312⇤⇤⇤ 0.171⇤⇤⇤ 0.224⇤⇤⇤ 0.156⇤⇤⇤
(0.018) (0.028) (0.019) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.228 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
294

Table A.51: OLS regressions of all 2020 candidates’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.006 0.003 0.001 0.005
(0.007) (0.032) (0.004) (0.006)
Republican 0.008⇤⇤ 0.024 0.023⇤⇤⇤ 0.008⇤⇤
(0.004) (0.024) (0.003) (0.004)
Incumbent 0.023⇤⇤⇤ 0.023 0.047⇤⇤⇤ 0.046⇤⇤⇤
(0.006) (0.032) (0.004) (0.006)
Competitive Scale 0.003⇤ 0.007 0.003⇤⇤ 0.005⇤⇤⇤
(0.002) (0.011) (0.001) (0.002)
District Party Fit 0.003 0.006 0.001 0.015⇤⇤⇤
(0.005) (0.008) (0.003) (0.005)
BIPOC Woman*Republican 0.004 0.010⇤ 0.007 0.0004
(0.013) (0.006) (0.008) (0.011)
Constant 0.045⇤⇤ 0.016 0.001 0.075⇤⇤⇤
(0.021) (0.035) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.316 0.220 0.158 0.301
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
295

Table A.52: OLS regressions of all 2020 candidates’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.001 0.012 0.006 0.008
(0.006) (0.027) (0.006) (0.006)
Republican 0.001 0.015 0.042⇤⇤⇤ 0.016⇤⇤⇤
(0.004) (0.033) (0.004) (0.003)
Incumbent 0.084⇤⇤⇤ 0.049⇤⇤ 0.076⇤⇤⇤ 0.046⇤⇤⇤
(0.006) (0.024) (0.006) (0.005)
Competitive Scale 0.002 0.0004 0.003⇤ 0.0003
(0.002) (0.008) (0.002) (0.002)
District Party Fit 0.016⇤⇤⇤ 0.006 0.007 0.001
(0.005) (0.007) (0.005) (0.004)
BIPOC Woman*Republican 0.018 0.023⇤⇤⇤ 0.007 0.016
(0.011) (0.007) (0.011) (0.011)
Constant 0.317⇤⇤⇤ 0.171⇤⇤⇤ 0.223⇤⇤⇤ 0.155⇤⇤⇤
(0.018) (0.028) (0.019) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.228 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
296

Table A.53: OLS regressions of all 2020 candidates’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.020⇤⇤⇤ 0.015 0.017⇤⇤⇤ 0.013⇤⇤⇤
(0.006) (0.032) (0.004) (0.005)
Republican 0.010⇤⇤ 0.027 0.020⇤⇤⇤ 0.007⇤
(0.004) (0.024) (0.003) (0.004)
Incumbent 0.024⇤⇤⇤ 0.024 0.048⇤⇤⇤ 0.047⇤⇤⇤
(0.006) (0.031) (0.004) (0.006)
Competitive Scale 0.003 0.007 0.002⇤ 0.004⇤⇤
(0.002) (0.010) (0.001) (0.002)
District Party Fit 0.002 0.006 0.002 0.014⇤⇤⇤
(0.005) (0.007) (0.003) (0.005)
White Woman*Republican 0.004 0.001 0.005 0.001
(0.010) (0.006) (0.006) (0.009)
Constant 0.046⇤⇤ 0.018 0.0003 0.076⇤⇤⇤
(0.021) (0.034) (0.014) (0.019)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.317 0.220 0.160 0.302
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
297

Table A.54: OLS regressions of all 2020 candidates’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.005 0.007 0.003 0.007
(0.005) (0.027) (0.005) (0.005)
Republican 0.0003 0.013 0.041⇤⇤⇤ 0.015⇤⇤⇤
(0.004) (0.033) (0.004) (0.004)
Incumbent 0.085⇤⇤⇤ 0.049⇤⇤ 0.077⇤⇤⇤ 0.046⇤⇤⇤
(0.006) (0.024) (0.006) (0.005)
Competitive Scale 0.002 0.001 0.003⇤ 0.00001
(0.002) (0.007) (0.002) (0.002)
District Party Fit 0.016⇤⇤⇤ 0.007 0.008⇤ 0.001
(0.004) (0.006) (0.004) (0.004)
White Woman*Republican 0.023⇤⇤⇤ 0.021⇤⇤⇤ 0.010 0.002
(0.009) (0.007) (0.009) (0.008)
Constant 0.317⇤⇤⇤ 0.173⇤⇤⇤ 0.223⇤⇤⇤ 0.156⇤⇤⇤
(0.018) (0.028) (0.018) (0.018)
Observations 13,874 13,874 13,874 13,874
Adjusted R2 0.567 0.228 0.474 0.342
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
298

A.2.4 main effects raceGen group models
Table A.55: OLS regressions of all 2020 White Men Democratic candidates’ monthly logged average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.038⇤⇤⇤ 0.020 0.026⇤⇤⇤ 0.026⇤⇤⇤
(0.005) (0.031) (0.003) (0.005)
Incumbent 0.030⇤⇤⇤ 0.024 0.061⇤⇤⇤ 0.064⇤⇤⇤
(0.008) (0.025) (0.005) (0.007)
Competitive Scale 0.008⇤⇤⇤ 0.009 0.005⇤⇤⇤ 0.009⇤⇤⇤
(0.003) (0.039) (0.002) (0.003)
District Party Fit 0.035⇤⇤⇤ 0.022⇤⇤ 0.015⇤⇤⇤ 0.045⇤⇤⇤
(0.007) (0.010) (0.005) (0.007)
Constant 0.100⇤⇤⇤ 0.020 0.042⇤⇤ 0.087⇤⇤⇤
(0.029) (0.041) (0.019) (0.027)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.349 0.238 0.186 0.337
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
DEMS only
299

Table A.56: OLS regressions of all 2020 White Men Democratic candidates’ monthly logged average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.018⇤⇤⇤ 0.017 0.001 0.011⇤⇤
(0.004) (0.036) (0.004) (0.004)
Incumbent 0.087⇤⇤⇤ 0.053⇤ 0.082⇤⇤⇤ 0.054⇤⇤⇤
(0.007) (0.031) (0.007) (0.007)
Competitive Scale 0.001 0.007 0.001 0.006⇤⇤
(0.002) (0.027) (0.002) (0.002)
District Party Fit 0.017⇤⇤⇤ 0.014⇤⇤ 0.00002 0.009
(0.006) (0.007) (0.006) (0.006)
Constant 0.344⇤⇤⇤ 0.240⇤⇤⇤ 0.242⇤⇤⇤ 0.164⇤⇤⇤
(0.026) (0.036) (0.026) (0.025)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.581 0.242 0.482 0.362
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
300

Table A.57: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.010 0.007 0.004 0.011
(0.010) (0.049) (0.007) (0.009)
Incumbent 0.031⇤⇤⇤ 0.033 0.037⇤⇤⇤ 0.042⇤⇤⇤
(0.011) (0.036) (0.007) (0.009)
Competitive Scale 0.002 0.004 0.001 0.00005
(0.003) (0.045) (0.002) (0.003)
District Party Fit 0.008 0.017 0.005 0.003
(0.009) (0.018) (0.006) (0.008)
Constant 0.012 0.016 0.046⇤⇤ 0.053⇤⇤
(0.030) (0.051) (0.020) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.288 0.192 0.162 0.283
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
GOP only
301

Table A.58: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.009 0.013 0.005 0.003
(0.009) (0.038) (0.009) (0.009)
Incumbent 0.083⇤⇤⇤ 0.042 0.072⇤⇤⇤ 0.036⇤⇤⇤
(0.009) (0.051) (0.010) (0.009)
Competitive Scale 0.002 0.007 0.005⇤ 0.005⇤
(0.003) (0.037) (0.003) (0.003)
District Party Fit 0.002 0.005 0.011 0.001
(0.008) (0.018) (0.008) (0.008)
Constant 0.315⇤⇤⇤ 0.135⇤⇤⇤ 0.254⇤⇤⇤ 0.128⇤⇤⇤
(0.026) (0.039) (0.027) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.557 0.224 0.475 0.322
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
302

Table A.59: OLS regressions of all 2020 BIPOC Women Democratic candidates’ monthly logged
average use of negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.009 0.002 0.002 0.008
(0.006) (0.030) (0.004) (0.006)
Incumbent 0.020⇤⇤⇤ 0.018 0.054⇤⇤⇤ 0.057⇤⇤⇤
(0.008) (0.023) (0.005) (0.007)
Competitive Scale 0.009⇤⇤⇤ 0.009 0.006⇤⇤⇤ 0.010⇤⇤⇤
(0.003) (0.038) (0.002) (0.003)
District Party Fit 0.027⇤⇤⇤ 0.016 0.008⇤ 0.040⇤⇤⇤
(0.007) (0.011) (0.005) (0.006)
Constant 0.079⇤⇤⇤ 0.013 0.031 0.072⇤⇤⇤
(0.029) (0.040) (0.019) (0.027)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.344 0.236 0.179 0.335
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
303

Table A.60: OLS regressions of all 2020 BIPOC Women Democratic candidates’ monthly logged
average use of positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.001 0.014 0.005 0.009
(0.005) (0.035) (0.005) (0.005)
Incumbent 0.082⇤⇤⇤ 0.050 0.082⇤⇤⇤ 0.051⇤⇤⇤
(0.007) (0.032) (0.007) (0.007)
Competitive Scale 0.001 0.007 0.002 0.006⇤⇤⇤
(0.002) (0.028) (0.002) (0.002)
District Party Fit 0.022⇤⇤⇤ 0.016⇤ 0.0004 0.010⇤
(0.006) (0.008) (0.006) (0.006)
Constant 0.335⇤⇤⇤ 0.225⇤⇤⇤ 0.245⇤⇤⇤ 0.154⇤⇤⇤
(0.026) (0.038) (0.026) (0.025)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.580 0.241 0.483 0.361
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
304

Table A.61: OLS regressions of all 2020 Democratic candidates’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.018⇤⇤⇤ 0.015 0.018⇤⇤⇤ 0.012⇤⇤
(0.005) (0.030) (0.004) (0.005)
Incumbent 0.022⇤⇤⇤ 0.020 0.056⇤⇤⇤ 0.058⇤⇤⇤
(0.008) (0.024) (0.005) (0.007)
Competitive Scale 0.008⇤⇤⇤ 0.009 0.005⇤⇤ 0.009⇤⇤⇤
(0.003) (0.038) (0.002) (0.003)
District Party Fit 0.024⇤⇤⇤ 0.016 0.007⇤ 0.038⇤⇤⇤
(0.007) (0.010) (0.005) (0.006)
Constant 0.089⇤⇤⇤ 0.016 0.036⇤ 0.079⇤⇤⇤
(0.029) (0.040) (0.019) (0.027)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.345 0.237 0.182 0.335
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
305

Table A.62: OLS regressions of all 2020 Democratic candidates’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.007 0.007 0.005 0.007
(0.005) (0.036) (0.005) (0.005)
Incumbent 0.083⇤⇤⇤ 0.049 0.083⇤⇤⇤ 0.051⇤⇤⇤
(0.007) (0.031) (0.007) (0.007)
Competitive Scale 0.001 0.007 0.002 0.006⇤⇤
(0.003) (0.027) (0.002) (0.002)
District Party Fit 0.022⇤⇤⇤ 0.018⇤⇤⇤ 0.00000 0.012⇤⇤
(0.006) (0.007) (0.006) (0.006)
Constant 0.338⇤⇤⇤ 0.235⇤⇤⇤ 0.244⇤⇤⇤ 0.161⇤⇤⇤
(0.026) (0.037) (0.025) (0.025)
Observations 7,489 7,489 7,489 7,489
Adjusted R2 0.580 0.241 0.483 0.361
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
306

Table A.63: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.039⇤⇤⇤ 0.026 0.022⇤⇤⇤ 0.019⇤⇤⇤
(0.007) (0.048) (0.005) (0.006)
Incumbent 0.036⇤⇤⇤ 0.036 0.040⇤⇤⇤ 0.045⇤⇤⇤
(0.011) (0.035) (0.007) (0.009)
Competitive Scale 0.002 0.004 0.0005 0.0002
(0.003) (0.045) (0.002) (0.003)
District Party Fit 0.019⇤⇤ 0.024⇤⇤ 0.013⇤⇤ 0.010
(0.009) (0.011) (0.006) (0.008)
Constant 0.042 0.004 0.031 0.065⇤⇤
(0.030) (0.050) (0.020) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.291 0.194 0.164 0.284
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
307

Table A.64: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.020⇤⇤⇤ 0.035 0.007 0.016⇤⇤
(0.006) (0.037) (0.006) (0.006)
Incumbent 0.086⇤⇤⇤ 0.046 0.073⇤⇤⇤ 0.038⇤⇤⇤
(0.009) (0.050) (0.010) (0.009)
Competitive Scale 0.002 0.007 0.005⇤ 0.005⇤
(0.003) (0.035) (0.003) (0.003)
District Party Fit 0.006 0.005 0.013 0.004
(0.008) (0.010) (0.008) (0.008)
Constant 0.328⇤⇤⇤ 0.164⇤⇤⇤ 0.261⇤⇤⇤ 0.140⇤⇤⇤
(0.027) (0.037) (0.027) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.558 0.227 0.475 0.322
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
308

Table A.65: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.027⇤⇤ 0.024 0.015⇤ 0.023⇤⇤
(0.012) (0.049) (0.008) (0.011)
Incumbent 0.032⇤⇤⇤ 0.034 0.038⇤⇤⇤ 0.043⇤⇤⇤
(0.011) (0.036) (0.007) (0.009)
Competitive Scale 0.002 0.004 0.001 0.0002
(0.003) (0.045) (0.002) (0.003)
District Party Fit 0.010 0.019 0.008 0.007
(0.009) (0.019) (0.006) (0.008)
Constant 0.015 0.014 0.047⇤⇤ 0.052⇤⇤
(0.029) (0.051) (0.020) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.288 0.192 0.162 0.283
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
309

Table A.66: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.012 0.024 0.016 0.019⇤
(0.011) (0.038) (0.011) (0.011)
Incumbent 0.084⇤⇤⇤ 0.043 0.072⇤⇤⇤ 0.037⇤⇤⇤
(0.009) (0.050) (0.010) (0.009)
Competitive Scale 0.002 0.007 0.005⇤ 0.005⇤
(0.003) (0.036) (0.003) (0.003)
District Party Fit 0.001 0.003 0.008 0.001
(0.008) (0.016) (0.008) (0.008)
Constant 0.313⇤⇤⇤ 0.138⇤⇤⇤ 0.255⇤⇤⇤ 0.129⇤⇤⇤
(0.026) (0.038) (0.027) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.557 0.224 0.475 0.322
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
310

Table A.67: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
negative emotive appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.038⇤⇤⇤ 0.022 0.030⇤⇤⇤ 0.026⇤⇤⇤
(0.009) (0.048) (0.006) (0.008)
Incumbent 0.034⇤⇤⇤ 0.035 0.040⇤⇤⇤ 0.044⇤⇤⇤
(0.011) (0.036) (0.007) (0.009)
Competitive Scale 0.002 0.004 0.001 0.0001
(0.003) (0.045) (0.002) (0.003)
District Party Fit 0.009 0.017 0.008 0.006
(0.009) (0.013) (0.006) (0.008)
Constant 0.011 0.016 0.050⇤⇤ 0.049⇤
(0.029) (0.050) (0.019) (0.025)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.290 0.193 0.165 0.284
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
311

Table A.68: OLS regressions of all 2020 Republican candidates’ monthly logged average use of
positive emotive appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.032⇤⇤⇤ 0.033 0.016⇤⇤ 0.012
(0.008) (0.037) (0.008) (0.008)
Incumbent 0.086⇤⇤⇤ 0.044 0.073⇤⇤⇤ 0.037⇤⇤⇤
(0.009) (0.050) (0.010) (0.009)
Competitive Scale 0.003 0.007 0.005⇤ 0.005⇤
(0.003) (0.036) (0.003) (0.003)
District Party Fit 0.002 0.004 0.011 0.0003
(0.008) (0.012) (0.008) (0.008)
Constant 0.310⇤⇤⇤ 0.135⇤⇤⇤ 0.254⇤⇤⇤ 0.127⇤⇤⇤
(0.026) (0.038) (0.027) (0.026)
Observations 6,385 6,385 6,385 6,385
Adjusted R2 0.559 0.226 0.475 0.322
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
312

A.2.5 raceGen group*party subset by inc status
Table A.69: OLS regressions of 2020 challengers’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.051⇤⇤⇤ 0.018 0.029⇤⇤⇤ 0.034⇤⇤⇤
(0.007) (0.028) (0.005) (0.006)
Republican 0.018⇤⇤ 0.008 0.035⇤⇤⇤ 0.038⇤⇤⇤
(0.007) (0.023) (0.005) (0.006)
Competitive Scale 0.001 0.007 0.004⇤⇤ 0.004⇤
(0.003) (0.034) (0.002) (0.002)
District Party Fit 0.014⇤⇤ 0.001 0.006 0.026⇤⇤⇤
(0.005) (0.012) (0.004) (0.005)
White Man*Republican 0.020⇤⇤ 0.0002 0.013⇤ 0.028⇤⇤⇤
(0.010) (0.010) (0.007) (0.009)
Constant 0.109⇤⇤⇤ 0.029 0.033⇤ 0.136⇤⇤⇤
(0.026) (0.038) (0.018) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.330 0.228 0.179 0.295
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
challengers only
313

Table A.70: OLS regressions of 2020 challengers’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.034⇤⇤⇤ 0.031 0.010 0.020⇤⇤⇤
(0.007) (0.033) (0.007) (0.006)
Republican 0.007 0.028 0.048⇤⇤⇤ 0.013⇤⇤
(0.006) (0.032) (0.006) (0.006)
Competitive Scale 0.001 0.002 0.001 0.001
(0.002) (0.030) (0.002) (0.002)
District Party Fit 0.011⇤⇤ 0.005 0.009⇤ 0.006
(0.005) (0.009) (0.005) (0.005)
White Man*Republican 0.019⇤⇤ 0.005 0.005 0.008
(0.009) (0.008) (0.009) (0.009)
Constant 0.298⇤⇤⇤ 0.164⇤⇤⇤ 0.189⇤⇤⇤ 0.171⇤⇤⇤
(0.024) (0.037) (0.024) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.553 0.237 0.465 0.356
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
314

Table A.71: OLS regressions of 2020 challengers’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.005 0.012 0.003 0.007
(0.008) (0.028) (0.006) (0.007)
Republican 0.019⇤⇤⇤ 0.001 0.034⇤⇤⇤ 0.029⇤⇤⇤
(0.005) (0.022) (0.004) (0.005)
Competitive Scale 0.002 0.007 0.004⇤⇤ 0.004⇤
(0.003) (0.032) (0.002) (0.002)
District Party Fit 0.013⇤⇤ 0.001 0.006 0.024⇤⇤⇤
(0.005) (0.015) (0.004) (0.005)
BIPOC Man*Republican 0.006 0.016⇤ 0.005 0.010
(0.013) (0.010) (0.008) (0.011)
Constant 0.082⇤⇤⇤ 0.023 0.018 0.123⇤⇤⇤
(0.026) (0.037) (0.018) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.324 0.227 0.173 0.293
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
315

Table A.72: OLS regressions of 2020 challengers’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.020⇤⇤⇤ 0.017 0.010 0.005
(0.008) (0.034) (0.008) (0.007)
Republican 0.012⇤⇤ 0.018 0.044⇤⇤⇤ 0.014⇤⇤⇤
(0.005) (0.032) (0.005) (0.005)
Competitive Scale 0.001 0.002 0.001 0.001
(0.002) (0.032) (0.002) (0.002)
District Party Fit 0.011⇤⇤ 0.003 0.011⇤⇤ 0.005
(0.005) (0.012) (0.005) (0.005)
BIPOC Man*Republican 0.014 0.004 0.025⇤⇤ 0.003
(0.012) (0.008) (0.012) (0.011)
Constant 0.276⇤⇤⇤ 0.141⇤⇤⇤ 0.186⇤⇤⇤ 0.160⇤⇤⇤
(0.024) (0.038) (0.024) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.552 0.233 0.466 0.355
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
316

Table A.73: OLS regressions of 2020 challengers’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.031⇤⇤⇤ 0.016 0.009 0.030⇤⇤⇤
(0.009) (0.028) (0.006) (0.008)
Republican 0.015⇤⇤⇤ 0.005 0.034⇤⇤⇤ 0.026⇤⇤⇤
(0.005) (0.022) (0.004) (0.005)
Competitive Scale 0.002 0.007 0.004⇤⇤ 0.004⇤
(0.003) (0.033) (0.002) (0.002)
District Party Fit 0.014⇤⇤⇤ 0.003 0.005 0.025⇤⇤⇤
(0.005) (0.012) (0.004) (0.005)
BIPOC Woman*Republican 0.009 0.002 0.0004 0.023⇤
(0.014) (0.010) (0.009) (0.013)
Constant 0.077⇤⇤⇤ 0.016 0.017 0.112⇤⇤⇤
(0.027) (0.037) (0.018) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.325 0.227 0.174 0.294
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
317

Table A.74: OLS regressions of 2020 challengers’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.016⇤⇤ 0.023 0.012 0.013⇤
(0.008) (0.033) (0.008) (0.008)
Republican 0.011⇤⇤ 0.018 0.051⇤⇤⇤ 0.013⇤⇤⇤
(0.005) (0.033) (0.005) (0.005)
Competitive Scale 0.001 0.002 0.001 0.001
(0.002) (0.032) (0.002) (0.002)
District Party Fit 0.012⇤⇤ 0.003 0.008 0.005
(0.005) (0.011) (0.005) (0.005)
BIPOC Woman*Republican 0.001 0.012 0.022⇤ 0.008
(0.013) (0.008) (0.013) (0.012)
Constant 0.278⇤⇤⇤ 0.144⇤⇤⇤ 0.179⇤⇤⇤ 0.160⇤⇤⇤
(0.024) (0.038) (0.024) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.552 0.234 0.465 0.356
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
318

Table A.75: OLS regressions of 2020 challengers’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.025⇤⇤⇤ 0.016 0.021⇤⇤⇤ 0.019⇤⇤⇤
(0.007) (0.027) (0.005) (0.006)
Republican 0.015⇤⇤ 0.006 0.033⇤⇤⇤ 0.028⇤⇤⇤
(0.006) (0.023) (0.004) (0.005)
Competitive Scale 0.001 0.006 0.003⇤ 0.004⇤
(0.003) (0.032) (0.002) (0.002)
District Party Fit 0.011⇤⇤ 0.001 0.003 0.022⇤⇤⇤
(0.005) (0.012) (0.004) (0.005)
White Woman*Republican 0.001 0.001 0.003 0.003
(0.011) (0.010) (0.008) (0.010)
Constant 0.084⇤⇤⇤ 0.020 0.019 0.121⇤⇤⇤
(0.026) (0.037) (0.017) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.326 0.228 0.177 0.294
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
319

Table A.76: OLS regressions of 2020 challengers’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.008 0.004 0.009 0.008
(0.006) (0.033) (0.006) (0.006)
Republican 0.010⇤⇤ 0.016 0.050⇤⇤⇤ 0.011⇤⇤
(0.005) (0.032) (0.005) (0.005)
Competitive Scale 0.001 0.002 0.001 0.001
(0.002) (0.031) (0.002) (0.002)
District Party Fit 0.013⇤⇤⇤ 0.004 0.010⇤⇤ 0.005
(0.005) (0.009) (0.005) (0.005)
White Woman*Republican 0.003 0.016⇤ 0.007 0.008
(0.010) (0.008) (0.010) (0.010)
Constant 0.282⇤⇤⇤ 0.147⇤⇤⇤ 0.184⇤⇤⇤ 0.161⇤⇤⇤
(0.024) (0.038) (0.024) (0.023)
Observations 7,993 7,993 7,993 7,993
Adjusted R2 0.552 0.232 0.465 0.355
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
320

Table A.77: OLS regressions of 2020 incumbents’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Man (0/1) 0.013 0.012 0.015⇤⇤⇤ 0.005
(0.008) (0.075) (0.005) (0.007)
Republican 0.071⇤⇤⇤ 0.077 0.004 0.047⇤⇤⇤
(0.015) (0.053) (0.010) (0.013)
Competitive Scale 0.0004 0.004 0.002 0.0005
(0.004) (0.074) (0.002) (0.003)
District Party Fit 0.017 0.022 0.031⇤ 0.011
(0.024) (0.014) (0.016) (0.022)
White Man*Republican 0.030⇤ 0.024⇤⇤ 0.014 0.036⇤⇤
(0.017) (0.010) (0.011) (0.015)
Constant 0.239⇤⇤⇤ 0.137⇤ 0.160⇤⇤⇤ 0.271⇤⇤⇤
(0.046) (0.083) (0.030) (0.042)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.307 0.250 0.185 0.325
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
incumbents only
321

Table A.78: OLS regressions of 2020 incumbents’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Man (0/1) 0.005 0.002 0.019⇤⇤⇤ 0.002
(0.007) (0.065) (0.007) (0.007)
Republican 0.037⇤⇤⇤ 0.045 0.059⇤⇤⇤ 0.002
(0.013) (0.071) (0.013) (0.013)
Competitive Scale 0.009⇤⇤⇤ 0.002 0.013⇤⇤⇤ 0.001
(0.003) (0.063) (0.003) (0.003)
District Party Fit 0.096⇤⇤⇤ 0.064⇤⇤⇤ 0.044⇤⇤ 0.067⇤⇤⇤
(0.021) (0.010) (0.021) (0.021)
White Man*Republican 0.055⇤⇤⇤ 0.046⇤⇤⇤ 0.034⇤⇤ 0.033⇤⇤
(0.015) (0.010) (0.014) (0.014)
Constant 0.511⇤⇤⇤ 0.290⇤⇤⇤ 0.369⇤⇤⇤ 0.175⇤⇤⇤
(0.041) (0.053) (0.040) (0.040)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.467 0.189 0.385 0.271
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
322

Table A.79: OLS regressions of 2020 incumbents’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Man (0/1) 0.024⇤⇤⇤ 0.025 0.020⇤⇤⇤ 0.023⇤⇤⇤
(0.009) (0.076) (0.006) (0.008)
Republican 0.046⇤⇤⇤ 0.057 0.010⇤⇤ 0.020⇤⇤⇤
(0.008) (0.054) (0.005) (0.007)
Competitive Scale 0.001 0.005 0.001 0.001
(0.004) (0.075) (0.002) (0.003)
District Party Fit 0.008 0.014 0.024 0.018
(0.024) (0.015) (0.016) (0.022)
BIPOC Man*Republican 0.018 0.017 0.010 0.014
(0.029) (0.010) (0.019) (0.026)
Constant 0.217⇤⇤⇤ 0.117 0.142⇤⇤⇤ 0.254⇤⇤⇤
(0.047) (0.083) (0.031) (0.043)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.306 0.250 0.184 0.325
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
323

Table A.80: OLS regressions of 2020 incumbents’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Man (0/1) 0.003 0.021 0.008 0.015⇤
(0.008) (0.066) (0.008) (0.008)
Republican 0.007 0.0002 0.038⇤⇤⇤ 0.028⇤⇤⇤
(0.007) (0.072) (0.007) (0.007)
Competitive Scale 0.007⇤⇤ 0.002 0.012⇤⇤⇤ 0.0003
(0.003) (0.064) (0.003) (0.003)
District Party Fit 0.083⇤⇤⇤ 0.055⇤⇤⇤ 0.040⇤ 0.059⇤⇤⇤
(0.021) (0.011) (0.021) (0.021)
BIPOC Man*Republican 0.036 0.021⇤⇤ 0.007 0.010
(0.025) (0.010) (0.025) (0.025)
Constant 0.508⇤⇤⇤ 0.300⇤⇤⇤ 0.376⇤⇤⇤ 0.181⇤⇤⇤
(0.041) (0.055) (0.041) (0.040)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.465 0.188 0.384 0.271
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
324

Table A.81: OLS regressions of 2020 incumbents’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.006 0.018 0.004 0.010
(0.010) (0.077) (0.007) (0.009)
Republican 0.038⇤⇤⇤ 0.046 0.017⇤⇤⇤ 0.011⇤
(0.007) (0.055) (0.005) (0.007)
Competitive Scale 0.001 0.004 0.001 0.0003
(0.004) (0.074) (0.002) (0.003)
District Party Fit 0.003 0.016 0.024 0.015
(0.025) (0.016) (0.016) (0.022)
BIPOC Woman*Republican 0.049 0.033⇤⇤⇤ 0.001 0.065⇤
(0.043) (0.010) (0.028) (0.039)
Constant 0.236⇤⇤⇤ 0.130 0.154⇤⇤⇤ 0.263⇤⇤⇤
(0.047) (0.084) (0.030) (0.042)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.306 0.249 0.183 0.324
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
325

Table A.82: OLS regressions of 2020 incumbents’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
BIPOC Woman (0/1) 0.003 0.015 0.019⇤⇤ 0.008
(0.009) (0.065) (0.009) (0.009)
Republican 0.008 0.009 0.037⇤⇤⇤ 0.024⇤⇤⇤
(0.006) (0.072) (0.006) (0.006)
Competitive Scale 0.008⇤⇤ 0.003 0.012⇤⇤⇤ 0.001
(0.003) (0.064) (0.003) (0.003)
District Party Fit 0.087⇤⇤⇤ 0.053⇤⇤⇤ 0.035⇤ 0.065⇤⇤⇤
(0.021) (0.012) (0.021) (0.021)
BIPOC Woman*Republican 0.021 0.030⇤⇤⇤ 0.055 0.050
(0.038) (0.010) (0.037) (0.037)
Constant 0.507⇤⇤⇤ 0.288⇤⇤⇤ 0.377⇤⇤⇤ 0.169⇤⇤⇤
(0.041) (0.054) (0.040) (0.040)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.465 0.188 0.385 0.271
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
326

Table A.83: OLS regressions of 2020 incumbents’ monthly logged average use of negative emotive
appeals on Twitter
negative anger disgust fear
(1) (2) (3) (4)
White Woman (0/1) 0.003 0.005 0.004 0.011
(0.010) (0.076) (0.006) (0.009)
Republican 0.032⇤⇤⇤ 0.047 0.019⇤⇤⇤ 0.009
(0.007) (0.054) (0.005) (0.007)
Competitive Scale 0.0001 0.004 0.001 0.0004
(0.004) (0.073) (0.002) (0.003)
District Party Fit 0.012 0.018 0.027⇤ 0.018
(0.024) (0.017) (0.016) (0.022)
White Woman*Republican 0.070⇤⇤⇤ 0.040⇤⇤⇤ 0.034⇤⇤ 0.044⇤⇤
(0.020) (0.010) (0.013) (0.018)
Constant 0.239⇤⇤⇤ 0.135 0.157⇤⇤⇤ 0.271⇤⇤⇤
(0.046) (0.082) (0.030) (0.042)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.307 0.250 0.184 0.324
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
327

Table A.84: OLS regressions of 2020 incumbents’ monthly logged average use of positive emotive
appeals on Twitter
positive joy trust anticipation
(1) (2) (3) (4)
White Woman (0/1) 0.002 0.007 0.002 0.008
(0.009) (0.065) (0.009) (0.008)
Republican 0.015⇤⇤ 0.002 0.036⇤⇤⇤ 0.027⇤⇤⇤
(0.006) (0.071) (0.006) (0.006)
Competitive Scale 0.007⇤⇤ 0.002 0.012⇤⇤⇤ 0.0004
(0.003) (0.063) (0.003) (0.003)
District Party Fit 0.091⇤⇤⇤ 0.060⇤⇤⇤ 0.044⇤⇤ 0.064⇤⇤⇤
(0.021) (0.010) (0.021) (0.021)
White Woman*Republican 0.085⇤⇤⇤ 0.058⇤⇤⇤ 0.042⇤⇤ 0.037⇤⇤
(0.017) (0.010) (0.017) (0.017)
Constant 0.516⇤⇤⇤ 0.292⇤⇤⇤ 0.376⇤⇤⇤ 0.176⇤⇤⇤
(0.041) (0.054) (0.040) (0.040)
Observations 5,881 5,881 5,881 5,881
Adjusted R2 0.468 0.190 0.385 0.272
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
Month and state fixed effects, account type,
monthly avg. tweet WC, logged avg. followers, and logged total n tweets omitted from table.
Standard errors are clustered by candidate.
328

A.2.6 Subgroup main and *party effects coefs.
329

Table A.85: Model estimates of race/gender*GOP effects from OLS regressions of 2020 incumbents’ monthly logged average use of negative valenced rhetoric in office tweets.
grp*GOP n adj. Rˆ2
negative
(1) White Man 0.013 0.030* 5,881 0.307 (0.008) (0.017)
(2) BIPOC Man -0.024*** 0.018 5,881 0.306 (0.009) (0.029)
(3) BIPOC Woman 0.006 0.049 5,881 0.306 (0.010) (0.043)
(4) White Woman 0.003 -0.070*** 5,881 0.307 (0.010) (0.020)
anger
(1) White Man 0.012 0.024* 5,881 0.250 (0.075) (0.010)
(2) BIPOC Man -0.025 0.017 5,881 0.250 (0.076) (0.010)
(3) BIPOC Woman 0.018 -0.033*** 5,881 0.249 (0.077) (0.010)
(4) White Woman -0.005 -0.040*** 5,881 0.250 (0.076) (0.010)
disgust
(1) White Man 0.015*** 0.014 5,881 0.185 (0.005) (0.011)
(2) BIPOC Man -0.020*** 0.010 5,881 0.184 (0.06) (0.019)
(3) BIPOC Woman 0.004 -0.001 5,881 0.183 (0.007) (0.028)
(4) White Woman -0.004 -0.034*** 5,881 0.184 (0.006) (0.013)
fear
(1) White Man 0.05 0.036** 5,881 0.325 (0.007) (0.015)
(2) BIPOC Man -0.023*** -0.014 5,881 0.325 (0.008) (0.026)
(3) BIPOC Woman 0.010 -0.65* 5,881 0.324 (0.009) (0.039)
(4) White Woman 0.011 -0.044** 5,881 0.324 (0.009) (0.018)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
330

Table A.86: Subgroup main and raceGen*party interaction effects (incumbents only) from separate
OLS regressions of 2020 incumbents’ monthly logged average use of positive emotive appeals on
Twitter
grp*GOP n adj. Rˆ2
positive
(1) White Man 0.005 -0.055*** 5,881 0.467 (0.007) (0,015)
(2) BIPOC Man -0.003 -0.036 5,881 0.465 (0.008) (0.025)
(3) BIPOC Woman -0.003 0.021 5,881 0.465 (0.009) (0.038)
(4) White Woman -0.002 0.085*** 5,881 0.468 (0.009) (0.017)
joy
(1) White Man 0.002 -0.046*** 5,881 0.189 (0.065) (0.010)
(2) BIPOC Man -0.021 0.021** 5,881 0.188 (0.066) (0.010)
(3) BIPOC Woman 0.015 -0.030*** 5,881 0.188 (0.065) (0.010)
(4) White Woman 0.007 0.058*** 5,881 0.190 (0.065) (0.010)
trust
(1) White Man 0.019*** -0.034** 5,881 0.385 (0.007) (0.014)
(2) BIPOC Man -0.008 -0.007 5,881 0.384 (0.008) (0.025)
(3) BIPOC Woman -0.019** -0.055 5,881 0.385 (0.009) (0.037)
(4) White Woman -0.002 0.042** 5,881 0.385 (0.009) (0.017)
anticipation
(1) White Man 0.002 -0.033** 5,881 0.271 (0.007) (0.014)
(2) BIPOC Man -0.015* -0.010 5,881 0.271 (0.008) (0.025)
(3) BIPOC Woman 0.008 0.050 5,881 0.271 (0.009) (0.037)
(4) White Woman 0.008 0.037** 5,881 0.272 (0.008) (0.017)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
331

B Model estimates of RGIDs and party interaction effects on emotive
rhetoric
B.1 Incumbent Representatives in the 113th to 116th Congresses (2017 to 2020)
332

Table B.1: Model estimates of race/gender*GOP effects from OLS regressions of 113 to 116 MC
monthly logged average use of negative valenced rhetoric in office account tweets.
grp*GOP n adj. R2
negative
(1) White Man 0.003 0.005 32,768 0.536 (0.003) (0.005)
(2) BIPOC Man -0.004 -0.004 32,768 0.536 (0.003) (0.009)
(3) BIPOC Woman -0.011*** -0.0005 32,768 0.536 (0.004) (0.013)
(4) White Woman 0.010** -0.016** 32,768 0.536 (0.004) (0.006)
anger
(1) White Man -0.006 0.005 32,768 0.409 (0.026) (0.003)
(2) BIPOC Man -0.004 -0.005 32,768 0.409 (0.026) (0.003)
(3) BIPOC Woman 0.002 0.001 32,768 0.409 (0.026) (0.003)
(4) White Woman 0.013 -0.008** 32,768 0.409 (0.026) (0.004)
disgust
(1) White Man 0.004** -0.001 32,768 0.361 (0.002) (0.003)
(2) BIPOC Man -0.003 -0.002 32,768 0.361 (0.002) (0.005)
(3) BIPOC Woman -0.009*** 0.007 32,768 0.361 (0.003) (0.008)
(4) White Woman 0.005** -0.008** 32,768 0.361 (0.002) (0.004)
fear
(1) White Man 0.001 0.001 32,768 0.465 (0.003) (0.005)
(2) BIPOC Man -0.001 -0.030*** 32,768 0.465 (0.003) (0.008)
(3) BIPOC Woman -0.007*** -0.017 32,768 0.465 (0.004) (0.012)
(4) White Woman 0.006* 0.007 32,768 0.465 (0.003) (0.006)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
333

Table B.2: Model estimates of race/gender*party effects from OLS regressions of 113 to 116 MCs’
monthly logged average use of positive valenced rhetoric in office tweets
grp*GOP n adj. Rˆ2
positive
(1) White Man -0.007** -0.008 32,768 0.666 (0.003) (0.005)
(2) BIPOC Man 0.0004 -0.033*** 32,768 0.666 (0.003) (0.008)
(3) BIPOC Woman 0.014*** -0.001 32,768 0.666 (0.003) (0.013)
(4) White Woman -0.0004 0.034*** 32,768 0.666 (0.004) (0.006)
joy
(1) White Man -0.012 -0.007** 32,768 0.327 (0.024) (0.003)
(2) BIPOC Man -0.011 -0.001 32,768 0.327 (0.025) (0.003)
(3) BIPOC Woman 0.026 0.009** 32,768 0.327 (0.024) (0.003)
(4) White Woman 0.012 0.016*** 32,768 0.327 (0.024) (0.003)
trust
(1) White Man .005* 0.002 32,768 0.602 (0.003) 0.005
(2) BIPOC Man 0.007** -0.042*** 32,768 0.602 (0.003) (0.008)
(3) BIPOC Woman -0.016*** -0.029** 32,768 0.602 (0.004) (0.012)
(4) White Woman -0.004 0.016*** 32,768 0.602 (0.004) (0.006)
anticipation
(1) White Man -0.001 -0.009* 32,768 0.439 (0.002) (0.004)
(2) BIPOC Man -0.013*** -0.011 32,768 0.439 (0.003) (0.007)
(3) BIPOC Woman 0.015*** -0.002 32,768 0.439 (0.003) (0.011)
(4) White Woman 0.005 0.016*** 32,768 0.439 (0.003) (0.005)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
334

B.2 Challenger and Incumbent Candidates in the 2020 Congressional Races
335

Table B.3: Model estimates of race/gender*GOP effects from OLS regressions of 2020 challengers’
monthly logged average use of positive valenced rhetoric in campaign tweets
grp*GOP n adj. R2
positive
(1) White Man -0.034*** 0.019** 7993 0.553 (0.007) (0.009)
(2) BIPOC Man 0.020*** -0.014 7993 0.553 (0.008) (0.012)
(3) BIPOC Woman 0.016*** -0.001 7993 0.553 (0.008) (0.013)
(4) White Woman 0.008 0.003 7993 0.553 (0.006) (0.010)
joy
(1) White Man -0.031 -0.005 7993 0.237 (0.033) (0.008)
(2) BIPOC Man 0.017 0.004 7993 0.237 (0.004) (0.008)
(3) BIPOC Woman 0.023 0.012 7993 0.237 (0.033) (0.008)
(4) White Woman 0.004 0.016* 7993 0.237 (0.033) (0.008)
trust
(1) White Man -0.010 0.005 7993 0.465 (0.007) (0.009)
(2) BIPOC Man -0.010 0.025** 7993 0.465 (0.008) (0.012)
(3) BIPOC Woman 0.012 -0.022* 7993 0.465 (0.008) (0.013)
(4) White Woman 0.009 -0.007 7993 0.465 (0.006) (0.010)
anticipation
(1) White Man -0.020*** 0.008 7993 0.356 (0.006) (0.009)
(2) BIPOC Man 0.005 0.003 7993 0.356 (0.007) (0.011)
(3) BIPOC Woman 0.013* 0.008 7993 0.356 (0.008) (0.012)
(4) White Woman 0.008 -0.008 7993 0.356 (0.006) (0.010)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
336

Table B.4: Subgroup main effects from separate OLS regressions of 2020 challenger candidates
monthly logged average use of negative emotive appeals on Twitter
grp*GOP n adj. R2
negative
(1) White Man 0.051*** -0.020** 7,993 0.330 (0.007) (0.010)
(2) BIPOC Man -0.005 -0.006 7,993 0.330 (0.008) (0.013)
(3) BIPOC Woman -0.031*** 0.009 7,993 0.330 (0.009) (0.014)
(4) White Woman -0.025*** -0.001 7,993 0.330 (0.007) (0.011)
anger
(1) White Man 0.018 -0.0002 7,993 0.228 (0.028) (0.010)
(2) BIPOC Man 0.012 -0.016* 7,993 0.228 (0.028) (0.010)
(3) BIPOC Woman -0.016 0.002 7,993 0.228 (0.028) (0.010)
(4) White Woman -0.016 -0.001 7,993 0.228 (0.027) (0.010)
disgust
(1) White Man 0.029*** -0.013* 7,993 0.179 (0.005) (0.007)
(2) BIPOC Man -0.003 0.005 7,993 0.179 (0.006) (0.008)
(3) BIPOC Woman -0.009 -0.0004 7,993 0.179 (0.006) (0.009)
(4) White Woman -0.021*** -0.003 7,993 0.179 (0.005) (0.008)
fear
(1) White Man 0.034*** -0.028*** 7,993 0.295 (0.006) (0.009)
(2) BIPOC Man 0.007 0.010 7,993 0.295 (0.007) (0.011)
(3) BIPOC Woman -0.030*** 0.023* 7,993 0.295 (0.008) (0.013)
(4) White Woman -0.019*** -0.003 7,993 0.295 (0.006) (0.010)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars (competitive scale, district
party fit), month and state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged total n tweets omitted
from table. Standard errors are clustered by candidate.
337

Table B.5: Differences between 2020 incumbents’ and challengers’ main effects from separate OLS
regressions of monthly logged average use of negative emotive appeals on Twitter
p-value
White Man
Negative -0.05 0.09
(0.03)
Anger -0.02 0.42
(0.02)
Disgust -0.07*** 0.00
(0.02)
Fear -0.06** 0.03
(0.03)
BIPOC Man
Negative -0.06** 0.05
(0.03)
Anger -0.06** 0.02
(0.03)
Disgust -0.10*** 0.00
(0.02)
Fear -0.13*** 0.00
(0.03)
BIPOC Woman
Negative 0.01 0.62
(0.03)
Anger 0.01 0.71
(0.02)
Disgust -0.05** 0.04
(0.02)
Fear -0.03 0.27
(0.03)
White Woman
Negative 0.01 0.81
(0.03)
Anger 0.01 0.68
(0.02)
Disgust -0.05*** 0.00
(0.02)
Fear -0.02 0.52
(0.03)
Note: ⇤p<0.1; ⇤⇤p<0.05; ⇤⇤⇤p<0.01
All models include candidate/district vars
(competitive scale, district party fit), month and
state fixed effects, account type, monthly avg.
tweet WC, logged avg. followers, and logged
total n tweets omitted from table. Standard errors are clustered by candidate.
338

C NRC Emotion-Associated Dictionary
C.1 Number of k Features in NRC Dictionary
The following table displays the number of k features in each referenced version of the NRC
Emotion-Associated Dictionary. Starting from the left following the emotion category label: 1) the
column displays the number of features from the combined dictionary used in the manuscript,
which contains the revised negative emotion keWys from Hua and Macdonald (2020) and the
positive emotion keys from 2) the NRC 2022 dictionary. This updated dictionary was accessed in
2022 using tidytext and textdata. When compared to 3) the 2020 version of the NRC dictionary, the
differences are overall small with the 2022 version paring down slightly on the number of words.
The 2020 NRC version is included here due to the fact that it was used as the basis for Hua and
Macdonald’s revised negative emotion dictionary keys relied upon in generating the estimates
used for the analysis.
category kFeats NRC 2022 NRC 2020
1 anger 1012 1247 1246
2 anticipation 839 839 837
3 disgust 841 1058 1056
4 fear 1014 1476 1474
5 joy 689 689 687
6 negative 1892 3324 3318
7 positive 2312 2312 2308
8 sadness 921 1191 1187
9 surprise 534 534 532
10 trust 1231 1231 1230
C.2 Correlation of Emotion Categories
Sadness seems to be the least correlated negative discrete emotion with anger. There is 83.4%
agreement between number of negative and anger words per tweet; 78.5% agreement between
number of anger and sadness words per tweet. Overall, disgust and fear are the least correlated
negative emotions.
For positive categories: The frequency of joy and trust words per tweet has correlation coefficient of about 0.6. Joyous rhetoric interestingly is not highly correlated with positive rhetoric, the
339

frequency scores for joy and positive words per tweet has a correlation coefficient of about 0.64.
There is a low correlation between negative and positive emotive categories overall. Joy, in
particularly, is very weakly correlated with all negative emotions and is least correlated with disgust, followed by sadness, fear, and anger.
negative anger disgust sadness fear positive joy trust
negative 1.00 0.83 0.68 0.80 0.87 0.11 0.03 0.13
anger 0.83 1.00 0.66 0.61 0.75 0.10 0.03 0.12
disgust 0.68 0.66 1.00 0.59 0.58 0.05 0.01 0.06
sadness 0.80 0.61 0.59 1.00 0.74 0.08 0.02 0.08
fear 0.87 0.75 0.58 0.74 1.00 0.10 0.03 0.11
positive 0.11 0.10 0.05 0.08 0.10 1.00 0.64 0.72
joy 0.03 0.03 0.01 0.02 0.03 0.64 1.00 0.59
trust 0.13 0.12 0.06 0.08 0.11 0.72 0.59 1.00
Table C.1: Correlation Matrix of Emotion Dictionary Categories
C.3 Measuring Emotional Appeals (Operationalizing Risk-Taking)
For my key outcome variables, computational text analysis was performed using the NRC dictionary to score for negative and positive sentiment and discrete emotional categories.
“The values show that for six of the eight emotions the Turkers have fair agreement, and
for anticipation and trust there is only slight agreement. The values for anger and sadness are
the highest. The average value for the eight emotions is 0.29, and it implies fair agreement”
(Mohammed and Turney 2013, 16)
Table: Agreement at two intensity levels of emotion (emotive and non-emotive): Fleiss’s , and
its interpretation (Mohammed and Turney 2013, 16)
When looking at the tweets with the highest scores in negative language measured by the
original NRC dictionary scores, I see a large discrepancy with what the revised negative sentiment
dictionary picked up:
“Vote! Vote! Vote Vote Vote Vote! Vote Vote Vote Vote Vote Vote Vote Vote Vote! Vote Vote Vote
Vote Vote Vote Vote Vote! Vote Vote! Vote Vote Vote Vote Vote Vote Vote Vote!Vote Vote! Vote Vote
Vote Vote Vote Vote Vote Vote Vote! Vote Vote Vote Vote Vote! Vote Vote!Vote Vote Vote!” - NRC
negative: 50; Revised: 0
“wear a mask wear a mask wear a mask wear a mask wear a mask wear a mask wear a mask
340

wear a mask wear a mask wear a mask wear a mask wear a mask wear a mask wear a mask wear
a mask wear a mask wear a mask wear a mask wear a mask wear a mask wear a mask” - NRC
negative: 21; Revised: 0
341

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Speculative Latinidades: imagining Latinx identities in science fiction and fantasy Media and activism

PDF

Multiracial politics or the politics of being multiracial?: Racial theory, civic engagement, and socio-political participation in a contemporary society

PDF

Voices unheard, stories untold: Black women, police violence and political participation

PDF

Feeling polarized: emotional partisan polarization and its reinforcement through campaign advertising

PDF

Bureaucratic politics and power building in the administrative state

PDF

Politics is something we do together: identity and institutions in U.S. elections

PDF

A question of identity: the rhetoric and argument of conspiracy theories

PDF

Neo-liberal hegemony and re-construction of Islamic community discourses in Turkey: the cases of the Gülen and Erenköy communities

PDF

Identity salience and political violence: centering social identity in the study of intrastate conflict

PDF

Refugee protection and empowerment joint venture: understanding how and why public and private actors cooperate to tackle the global refugee crisis

PDF

Socially-informed content analysis of online human behavior

PDF

Japanese-American basketball: constructing gender, ethnicity, and community

PDF

Identity and regional integration: case studies on central Asia and southeast Asia

PDF

Minority stress and intersectionality: how are identity, discrimination, and biological stress related to depression and suicidality among multiple minority subgroups?

PDF

The role of skin tone on candidate evaluation: an analysis of racial socialization among Latino immigrants

PDF

Window Seat: examining public space, politics, and social identity through urban public transportation

PDF

Creating cities and citizens: municipal boundaries, place entrepreneurs, and the production of race in Los Angeles county, 1926-1978

PDF

The representation of LGBT interests in the United States: a multimethod approach

PDF

The politics of transformation: mass media and the Northern Ireland peace process

PDF

The strategy of firms towards political elites and consumers in polarized western democracies

Asset Metadata

Creator Hua, Whitney (author)

Core Title Signaling identity: how race and gender shape what representatives say online

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Political Science and International Relations

Degree Conferral Date 2023-12

Publication Date 11/21/2023

Defense Date 11/02/2023

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag computational methods,Congress,emotions,gender politics,legislative politics,OAI-PMH Harvest,political communication,political psychology,race and ethnic politics,social media,text analysis

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Crigler, Ann (committee chair), Goodnight, G. Thomas (committee member), Grose, Christian (committee member)

Creator Email huawhitn@usc.edu,whitneyhua@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113777747

Unique identifier UC113777747

Identifier etd-HuaWhitney-12485.pdf (filename)

Legacy Identifier etd-HuaWhitney-12485

Document Type Dissertation

Format theses (aat)

Rights Hua, Whitney

Internet Media Type application/pdf

Type texts

Source 20231127-usctheses-batch-1108 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu