You are on page 1of 16

University of Utah

Emotions, Campaigns, and Political Participation


Author(s): Christopher Weber
Source: Political Research Quarterly, Vol. 66, No. 2 (JUNE 2013), pp. 414-428
Published by: Sage Publications, Inc. on behalf of the University of Utah
Stable URL: http://www.jstor.org/stable/23563153
Accessed: 13-05-2017 18:01 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Sage Publications, Inc., University of Utah are collaborating with JSTOR to digitize, preserve and
extend access to Political Research Quarterly

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Political Research Quarterly
66(2)414-428
Emotions, Campaigns, and © 2012 University of Utah
Reprints and permissions:

Political Participation
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1065912912449697
prq.sagepub.com

(DSAGE

Christopher Weber

Abstract

There has been a scarcity of work examining the political consequences of discrete emotions.This article exam
political effects of several emotions—anger, sadness,fear, and enthusiasm. Emotional ads should influence whet
become politically active.To test this, two experiments were administered.The first examines emotional re
campaign messages; the second tests whether emotions influence political participation.The results indicate
mobilizing, by increasing participatory intentions and factors related to participate.This result is then replic
ad-tracking data.The findings indicate that emotions are an important factor in studying campaign effects.

Keywords
emotions, affect, political campaigns, advertising

Emotions are a staple of American political campaigns, Lau and Pomper 2004; Brooks 2006), with the
and one need not look far to find examples of emo- being that under particular conditions attack ad
tional campaign ads. Both scholars and political pundits turnout. Yet much of this work fails to consider t
describe campaigns in emotion-centered terms, noting of emotions.
how candidates connect with voters, how a cold, emo- In this article I examine the consequences
tionless persona alienates voters, and how campaign tional appeals in campaign ads by (1) demonstra
advertisements elicit a host of emotions. Consider individuals experience an array of emotions in
Hillary Clinton's "3 AM" advertisement released during to campaign advertisements and (2) disaggreg
the 2008 Democratic Party primaries, or prototypical tions of the same valence; specific emotions h
spots, such as Lyndon Johnson's "Daisy Girl" ad in and important political consequences. Using tw
which an unaware girl picking apart a daisy precedes ments and ad-tracking data, I demonstrate that
images of a nuclear explosion. Since the onslaught of are a vital characteristic to consider when descr
television advertising, candidates have drawn on many consequences of political campaigns. In these
emotions (Brader 2006). anger emerges as a mobilizing force. Anger-evoking
While political advertisers attempt to elicit many emo- political messages heighten part
tions in the public (Brader 2006), there has been a scarcity consistent with previous research s
of scholarly work examining whether specific emotions associated with behavioral approach (L
aroused in political campaigns influence political behav- 2006; C. Harmon-Jones et al. 201
ior. Much of the literature on campaign advertising has the consequences of emotional ca
examined the implications of emotional valence— several theoretical approaches,
positive versus negative advertising (Ansolabehere et al.
1994) or ads that elicit enthusiasm versus fear (Brader
2005,2006). Moreover, one of the more contentious issues
to emerge in the Study of American campaigns is whether 'Louisiana State University, Baton Rouge, LA
negative ads affect political interest and participation. In
the decade and a half following Ansolabehere et al.'s Corresponding Author:
, , , , Christopher Weber, Department of Political Science, Manship School
(1994) seminal finding that attack advertisements reduce of Mass Communication, Louisiana State Univ
turnout, there has been a flurry of academic and nonaca- 70803, USA.
demie interest in the topic (Clinton and Lapinski 2004; Email: crweber@lsu.edu

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 415

Dimensional and Discrete what they call "aversion," aligns with the disposition sys
Models of Emotion tem- "Aversion, including feelings of anger, disgust, con
tempt, and hatred, signals the need to confront an
Research on affect has exerted a growing influence on the adversary. When familiar aversive stimuli are encou
study of political behavior, and affect is an essential com- tered, people rely on previously learned routines to
ponent in social judgment and behavior (see Davidson, age these situations, just as they do for familiar rewardin
Scherer, and Goldsmith 2003). The political implications circumstances" (MacKuen et al. 2010, 441; for cont
of emotion are pervasive, having been empirically linked evidence, see Watson 2009). Whether one operates in t
to perceptions of candidates and vote choice (Lodge and disposition or surveillance system depends on how
Taber 2005; Brader 2005), media consumption and learn- interprets threatening stimuli. Familiar threats activat
ing (Marcus and MacKuen 1993; Huddy et al. 2005), anger, whereas unfamiliar threats trigger anxiety. Dra
policy attitudes (Huddy, Feldman, and Cassese 2007), and on these findings, campaign ads that elicit anger or en
political participation (Valentino et al. 2011; Rudolph, siasm may have the behavioral consequence of elev
Gangl, and Stevens 2000). participation, whereas ads that elicit sadness or fear
Comparatively little research has examined the moti- should depress participation,
vational and persuasive impact of specific emotions in the A parallel approach in the study of emotion
context of a political campaign (the exception being discrete emotions. Much of this research
Valentino, Gregorowicz, and Groenendyk 2009; Valentino appraisal theory, which centers on how the e
et al. 2011). Much of the extant work in political science a stimulus shapes one's emotional experie
has advocated a dimensional approach to the study of appraisals have been consistently cited to u
affect, with one dimension being defined by positive emo- tional reactions: certainty, valence, control
tions and a second by negative emotions (Marcus 2003). and responsibility (Smith and Ellsworth 1985; C
Consider the pioneering work of George Marcus and Harmon-Jones 2009a, 2009b). Anger, for examp
colleagues and their theory of affective intelligence (AIT; from evaluating a stimulus as negative; it acc
Marcus and MacKuen 1993; Marcus, Neuman, and heightened perception of personal control th
MacKuen 2000; Neuman et al. 2007). According to this effectively cope with the event. Anger als
approach, emotional responses fall within one of two approach motivation and attributing blame
systems—disposition and surveillance. These systems ual (Lazarus 1991; Berkowitz 1990; Be
provide guidance as to whether habits are sufficient for Harmon-Jones 2004; Ellsworth and Sc
judgment, or whether greater attention needs to be paid to Mackie, Devos and Smith 2000). Sadness and
the surrounding environment (Marcus, Neuman, and other hand, stem from beliefs that a negative eve
MacKuen 2000). Positive emotions promote operating in be prevented, blame is not easy to assign, and o
the disposition system and a reliance on habit. On the as easily cope with the situation (Keltner, Ellswo
other hand, when the surveillance system is activated, Edwards 1993; Lerner and Keltner 2001).
usually through a particular threat, the individual relies Because anger and enthusiasm involve a
less on predispositions and habit. Anxiety is central to sense of personal control, these emotions facilit
this theoretical perspective since it is the emotion that ioral approach over withdrawal (Marsh, A
moves people to operate in the surveillance system. Kleck 2005). Fear and sadness, however, pr
Scholarly work has sought to understand which emo- ceptions of reduced personal control and
tions are rooted in these two systems, given evidence that quently lead to behavioral avoidance. P
negative emotions sometimes have effects similar to posi- findings comport with these expectations, w
tive emotions. For example, anger and anxiety are highly crete emotions correlate with particular actio
correlated emotions yet entail different physiological, psy- (Levenson 1992; Frijda, Kuipers, and ter Sch
chological, and motivational consequences (MacKuen As Frijda, Kuipers, and ter Schure (1989,213)
et al. 2010). Unlike anxiety, anger tends to accompany tions motivate "approach or attentional int
behavioral approach (Harmon-Jones et al. 2011) and deter- and withdrawal, and opposition and attack ar
mination (C. Harmon-Jones et al. 2011). Anger also acti- manifest behavior resulting from these impuls
vates left anterior areas of the cerebral cortex, much like Despite the theoretical differences bet
positive emotions (E. Harmon-Jones et al. 2004). Moreover, sional and appraisal models, in some instance
anger leads to information-processing strategies qualita- predict similar outcomes. Both approaches con
tively similar to positive affect (Bodenhausen, Sheppard, anger should facilitate behavioral approach (ho
and Kramer 1994). Watson 2009). Similarly, Tellegen and colleagues intro
Recent work by MacKuen and colleagues (2010) duce an integrated model of affect that consid
incorporates anger into AIT by arguing that anger, or from both approaches (Tellegen, Watson, and Clark

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
416 Political Research Quarterly 66(2)

Huddy, Feldman, and Cassese 2007). Their "hie


model" posits that affect is arrayed hierarc
discrete emotions at the first level, a negative
affect factor at the second level, and a bipolar
versus unhappiness factor at the third level (T
Watson, and Clark 1999). The model does not u
the existence of discrete emotions, nor does it
the notion of a single
The Experiments
dimension of affect. Rathe
and dimensional approaches are synthesized
model (Huddy, Feldman, and Cassese 200
Considering these approaches, and the obs
that scholars typically restrict their analysis
versus negative ads, I examine whether four e
aroused in campaign ads influence political
anger, sadness, fear, and enthusiasm have uni
ferentiated effects? On entering the lab, participants were seated at a com
puter terminal and instructed that they were to view a
campaign ad followed with a short survey. Participants
Hypotheses and Design were asked to watch and react to a political web adver
An underexplored question is how negative emotions tisement released from a previous congress
differ from one another in their political consequences. tion. The stimulus materials consisted of videos c
Given the unique relationship between emotions and using professional video-editing equipment, an
particular action tendencies, four emotions aroused in script for several of the messages consisted o
campaign ads—anger, sadness, fear, and enthusiasm— modified versions of those used by Ansolabe
should have important and differentiated consequences (1994) and Brader (2005, 2006) as well as s
for political behavior. In this study, I focus on these from existing political ads. Participants were
emotions because, when taken together, they constitute would view an ad drawn from a congressi
many of the emotions appealed to in political cam- between two candidates, John Wilkins and D
paigns (Brader 2006). As such, anger in the political All participants watched the advertisement at
domain should be tightly bound to perceptions of effi- computers equipped with headphones. Each p
cacy. However, since sadness and fear correspond to was randomly assigned to view only one of
apathy and withdrawal, these emotions should reduce after which point he or she completed a sho
efficacy and participatory intentions. On the other hand, After completing the survey, participants wer
fear and sadness, two emotions associated with help- see the researcher, at which time the study wa
lessness and withdrawal, should lead to political disin- and participants were thanked.
terest and abstention. In the first study, participants were randomly assigned
My two expectations are that (1) anger, fear, sadness, to view one of four ads designed to
and enthusiasm should be unique and differentiated con- emotion—anger, sadness, fear, or enthu
structs that (2) have varying consequences on political these ads all focused on the same substanti
participation and factors associated with participation. they did vary considerably in the types o
Specifically, anger and enthusiasm should be relatively In the anger ad, the focus was drugs, in t
strong predictors of both political efficacy and the desire fear ads the focus was violent crime, and
to participate in politics, whereas sadness and fear should ad the focus was the crime rate in the
lead to comparatively less participation. was intentional, as the purpose was to max
Three studies were conducted to test these hypotheses. ferences in experienced emotions across
The first study focuses on whether participants can dif- examine whether emotional responses a
ferentiate their emotional reactions felt in response to ferentiated. The purpose of the manipula
campaign ads. In study 1, participants were randomly ate an emotional reaction and examine whe
assigned to view one of four ads designed to elicit a dis- are able to differentiate specific emotio
crete emotion. The stimulus materials from this study campaign messages.1
were then used in study 2, where the goal was to measure After demonstrating differentia
the effects of emotional advertisements on participation. tions to these messages, a second study
The results indicate that anger increases the desire to be address the "stimulus equivalence" limit
politically active. Finally, because of the external validity and examine the behavioral ramificatio

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 417

campaign
ness ads, w
creating b
informatio
ciated with
ment foll
anger or s
whether e
cal backdro
the narrat
were absen
to the ad,
tions about
participate in politics.2 whether these emotions are indicators of a general posi
Emotion. To ascertain whether the manipulation tive versus negative affective state,
evoked an emotion, several questions were asked regard- A single factor model was first estimated. This model
ing the participant's emotional state. In study 1, several tests whether anger, disgust, fear, fright, hope, happiness,
questions were asked regarding the participant's levels and enthusiasm all load on a single bipolar dimension,
of enthusiasm, anger, fear, and sadness. Enthusiasm was with "positive" affect on one end and "negative" affect at
measured with three items—hopefulness, happiness, and the opposing end. On the whole, the one-factor model pro
optimism. Fear was measured with two items (afraid, vided an extremely poor fit to data (comparative fit index
fearful), as was anger (angry, disgusted). Sadness was [CFI] = .694, Tucker-Lewis index [TLI] = .591, root mean
measured with two items (sad, depressed). In study 2, the square error of approximation [RMSEA] = .392).
focus is on contrasting anger and sadness. Sadness was A bivalent model was estimated next, in which nega
measured with the same two items (sad, depressed), as tive emotions load on one dimension and positive emo
was anger (anger, disgust). tions on a separate dimension. This is the model most
Participatory intentions. Intention to participate in poli- frequently relied on in work that advances the dimen
tics was operationalized in a number of ways and is the sional approach to emotion (Watson and Clark 1992), and
focus of study 2. An internal efficacy scale was created it is closely aligned with the affective intelligence litera
from six questions that were combined into a scale rang- ture. There is little empirical support for this model, as it,
ing from 0 to 1 (a = .80). The statements were similar to too, does not provide an acceptable fit to data (CFI
those consistently asked in the American National Elec- .947, TLI = .927, RMSEA = .166).
tion Studies, for instance, "People like me don't have a Next, two hybrid models were estimated. The first model
say in what the government does" and "I feel that I have specifies that anger is defined by anger indicators (i.e.,
a pretty good understanding of the important political anger, disgust), fear by fear-related indicators (i.e., fearfu
issues facing our country." External political efficacy afraid), and happiness versus sadness by enthusiasm indica
was measured with a single item: "I don't think public tors (i.e., hope, optimism, happiness) and sadness indicators
officials care much about what people like me think"; (i.e., sad, depressed). This three-factor model provided
responses range from 1 (strongly disagree) to 4 (stron poor fit (CFI = .759, TLI = .639, RMSEA = .369). A seco
gly agree). Several participation questions were also hybrid model was estimated where negative affect was
included: voting ("How likely is it that you'll vote in the defined by the anger and fear indicators, and a positive
next election?") and volunteerism ("How likely is it that affect factor was defined by the enthusiasm and sadness
you'll volunteer in the next election?"). Finally, an indicators. This model also provided a very poor fit to da
agree-disagree civic duty statement was analyzed: (CFI = .750, TLI = .654, RMSEA = .361)
"People should vote whenever there is an election." The The best fitting model was a four-factor model, suc
ad scripts for studies 1 and 2 and the analyzed questions that anger and disgust load on an anger dimension; sad
are available in the online supplementary material (at ness, depression, and happiness define a sadness dime
http://prq.sagepub.com/supplemental/). sion; afraid and fearful items define the fear dimensio
and hope, optimism, and happiness load on an enthusiasm
dimension (CFI = .995, TLI = .9991, RMSEA =.059).3 In
Study I :The Structure Other words, the best fitting factor model was one where
of Emotions specific emotion terms load on four specific emotions
Dimensional models contend that emotions collapse to constructs. The estimates for the four-factor model are
one or two dimensions; discrete models argue that the presented in Table 1.

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
418 Political Research Quarterly 66(2)

Table I. Emotional Reactions to Advertisements, Study


and fear to zero did not worsen the overall model fit, x2(l)
Anger Fear Sadness = 1.509, ns, nor did constraining to zero the correlation
between sadness and enthusiasm, ^(l) = 0.006, ns. The
I (.00)
.9/2 correlation between anger and enthusiasm, however, was
.937 (.047) significant, x2(l) = 6.208,/? < .01. While positive and neg
.855 ative emotions are largely orthogonal, the model supports
I (.00) the notion that anger, sadness, and fear should be treated
.947
as separate, albeit correlated, constructs.
.993 (.028)
Study 1 also allows for a preliminary test of whether
.941
the manipulations elicit the expected emotions. For
I (.00)
.884 instance, does the anger ad elicit anger or some other
.964 (.041) emotion? Since the emotions are a consequence of the
.852 treatment conditions, I extend Table 1 by estimating a
multiple-indicator multiple-cause (MIMIC) model. The
four factors in Table 1 were predicted by dummy vari
-.315 (.073)
ables corresponding to the participant's assignment to a
-.278
treatment condition.

The point estimates and 95 percent confidence inter


Correlations vals in Figure 1 illustrate that felt enthusiasm was high
.715 est in the enthusiasm condition. The estimates for the
^anger.fear
anger,sad
.750
treatment variables leading to felt enthusiasm are all
.830
fear,sad negative, and the confidence intervals do not overlap
-.148
enth, anger with zero.
.076
enth.fear
.005
Contrasting the negative emotions, Figure 1 demon
enth,sad
Model fit strates that experienced anger was highest in the anger
CFI .995 condition relative to the fear condition, %2(1) = 6.340,
TLI .991 p < .01, and the sadness condition, x2(l) = 6.486,/? < .01.
RMSEA .059
However, the effect of the fear manipulation was not
nearly as clean. The anger manipulation elicited approxi
CFI = comparative fit index; RMSEA = root mean square error of ap
mately equal levels of fear relative to the fear manipula
proximation;TLI = Tucker-Lewis index. Four-factor confirmatory factor
tion, x2(l) = 0.171, ns, which is not uncommon given the
model. Entries are means and variance-adjusted weighted least squares
estimates.The factor loading for the first indicator is constrained toheightened
I co-occurrence of anger, fear, and anxiety
for model identification. Unstandardized coefficients with standard er
rors in parentheses and standardized coefficients in italics. Correlations
(Diener and Iran-Nejad 1986). However, the fear manipu
represent the standardized covariances between factors.Thresholds lation did elicit marginally more fear than the sadness
and factor variances have been excluded from the table for parsimony manipulation, x2(l) = 3.903,/? < .05. Regarding sadness,
but can be obtained from the author on request.
felt sadness was higher in the sadness condition relative
to both the anger condition, x2(l) = 5.639,/? < .01, and the
fear condition, x2(l) = 3.358,/? < .07.4
While the best fitting model is one where anger, sad These findings point to heterogeneity in emotional
ness, fear, and enthusiasm are separate constructs, the con reactions felt in response to campaign messages, echoing
structs are correlated. Emotions co-occur, particularly work that demonstrates variation in the consequences of
emotions of the same valence. For instance, the correla political ads of the same tone (Kahn and Kenney 1999;
tion between anger and fear was substantial (r = .715), as Brooks and Geer 2007). Several key findings emerge
were the correlations between sadness and anger (r = .750) from this study. First, emotions—particularly negative
and between sadness and fear (r = .830). However, the emotions—can be differentiated. While several theoreti
correlations between the three negative emotions and cal approaches contend positive emotions reside on one
enthusiasm were smaller (r = -.148; dimension and negative emotions on a separate dimen
Anger,Enthusiasm
rr_ ^ , = .076; r , „ , . = .005), which is con sion, the results do not support this view. Rather, sadness,
Fear,Enthusiasm # Sadness,Enthusiasm
fear, anger, and enthusiasm appear to be unique con
gruent with previous work demonstrating that emotions of
the same tone are heavily correlated (Diener and Iran
structs, despite being heavily correlated. Second, the
Nejad 1986), whereas positive and negative affect are
study illustrates that three of the four manipulations were
orthogonal (Marcus, Neuman, and MacKuen 2000). effective in eliciting a specific emotion. Sadness was
Indeed, constraining the correlation between enthusiasm
most prevalent in the sadness condition relative to all

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 419

-f/
'0 00 1 78,
FFear />.
y\;

•$

s.^

»<{ igf

1 f - < 87yC F.Sad

1^1
Vv \ •'.

3v

Figure I. Structural equation MIMIC model.


The effect of the experiment on the four emotion factors from Table I. Entries are unstandardized estimates with 95 percent confidence
intervals in parentheses. Item thresholds, factor covariances, residual variances, and standard errors have been excluded for clarity.

other conditions, anger was higher in the anger condition Study 2: Does Anger Mobilize?
relative to all other conditions, and enthusiasm was
greater in the enthusiasm condition relative to all other A concern in political communication research is whether
conditions. Because the primary motivation is to contrast a documented effect can be attributed to emotions, cogni
emotions of the same valence, in study 21 slightly modify tion, or some combination of the two. The complexity of
the anger and sadness ads to better document the electoral political messages—ranging from evocative cues to edit
consequences of these emotions. ing techniques and the use of coded language—renders it

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
420 Political Research Quarterly 66(2)

Table 2. Participation Results, Study 2.

Variables Internal External Vote Volunteer Civic Duty

Cue -0.03 -0.37 0.20 -0.46* 0.08

(0.03) (0.30) (0.30) (0.31) (0.34)


Anger 0.01 -0.32 0.27 -0.04 -0.14

(0.04) (0.31) (0.31) (0.32) (0.34)


Anger x cue 0.19** 0.65* 0.28 J 14*** 0.20
(0.05) (0.45) (0.46) (0.47) (0.51)
Female -0.11*** 0.14 -0.02 -0.15 0.01

(0.03) (0.24) (0.25) (0.26) (0.27)


Constant 0.61***

(0.03)
-1.96 -0.87 -0.31 -3.97
|

-0.05 0.48 1.59 -2.64


2
2.02 2.86 -0.58
3
N 274 274 270 271 274

Anger, sadness, and political participation. Internal efficacy column contains ordinary least squares estimates with standard errors in parentheses.
The remaining columns contain maximum likelihood estimates from ordered logistic regression with standard errors in parentheses,
fp < . 10, one-tailed. **p < ,05,***p < .01, two-tailed.

even more difficult to isolate the distinct effects of cogni- mobilizing, sadness demobilizing. Table 2 reveals the
tion and emotions. effects of the manipulation on internal and external politi
In the previous experiment, while all ads focused on cal efficacy, volunteerism, voting, and civic duty. T
the substantive issue of crime, somewhat different aspects internal efficacy column contains ordinary least squar
of crime were discussed. The sadness ad drew on violent coefficients, whereas the remaining columns are m
crime, whereas the anger ad emphasized mandatory sen- mum likelihood estimates from ordered logistic regres
tencing and drug policy. Because the anger and sadness sion. Given gender differences in political participation, I
manipulations were relatively clean with respect to the also control for the effect of gender in the models.6
manipulated emotion, slight modifications were made to Given the manner in which the variables are coded
these ads to explore whether anger mobilizes and sadness Table 2, the effect of the "cue" variable contrasts the sa
demobilizes. ness ad with the cue to the sadness ad without the cue.
In one condition, participants were exposed to an The cue variable does not
advertisement that included emotionally evocative ele- statistical significance fo
ments; specifically, an ominous voiceover, music, and p < .20, one-tailed), civ
threatening images accompanied the message. In a sec- tailed), voting (z = 0.68,
ond condition, these cues were absent. While the infor- efficacy (z = 0.24, p < .41, on
mation was held constant—that is, the information was marginally significant for v
essentially equivalent—the ads varied in the degree to one-tailed).
which a particular emotion should have been evoked. To contrast the anger ad
This method is similar to the approach used by Brader without the cue, the e
(2005, 2006). recoded, such that 0 = anger ad and 1 = sadness ad. This
As such, participants were assigned to one of four con- variable was then interacted with the cue man
ditions following from a 2 (emotion ad: anger vs. sad- dummy. Contrasting the anger ad with and
ness) x 2 (cues: present vs. absent) between-subjects cue leads to a significant difference for intern
design.5 This allowed for a test of whether emotional ads (t = 1.98, p < .025, one-tailed), voting (z = 1.36,
(relative to a baseline condition) affect participation. one-tailed), and volunteerism (z = 1.95, p <
A two-way emotion ad x audiovisual cues interaction tailed). The contrast was nonsignificant for exter
suggests that the emotion advertisements vary depending cacy and civic duty, however.
on whether the emotional cues are present or absent. To facilitate interpretation of several of th
Specifically, the anger ad with the audiovisual cues predicted values generated from Table 2 are
should elicit a greater desire to participate (relative to the Figure 2. Examining the upper-left plot in Figur
less emotional baseline); the opposite should occur for with emotionally evocative elements increase
the sadness ad. This would suggest that anger is relative to the same ad lacking emotionally

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 421

Anger Experiment Sadness Experiment

S^
8 8
*= 3=
LLjin - tu1^ -

, , _T_

Anger Cues No Cues Sadness Cues No Cues


Intemal —♦— External Internal —♦— External

00 - CO _
C ■ c ~
o o

co<9- • co<9
Q. g.

i
o
Tf -
t
co
CL CN1

o O
~r _r_

Anger Cues No Cues Sadness Cues No Cues


Vote —♦— Volunteer Vote —♦— Volunteer

Predicted Values from Table 2.

Figure 2. Effects of treatment conditions on efficacy, volunteerism and voting, study 2.

cues. The light gray line represents external efficacy, the ad/cue condition was 0.38 versus 0.43
dark gray line internal efficacy. On a scale of 0 to 1, the no cue condition.
predicted internal efficacy mean for participants in the The strongest effect emerged for volu
anger/emotion cues present condition was 0.58. Absent increased the likelihood of stating one w
these cues, the expected value was 0.51. The opposite an increase from 0.15 to 0.26. The sad
pattern emerged for sadness. The sadness ad with the emotion cue led to a greater likelih
emotional cues leads to an expected value of 0.47, (0.15) relative to the ad with th
whereas the same ad without the emotional cues leads to together, these results imply tha
an expected value of 0.49. While anger increases internal more specifically, negative emot
efficacy, sadness decreases efficacy. The same counter- have uniform effects. Anger seem
vailing pattern was found for external efficacy. For the Sadness has a less consistent effect. W
anger ad, including the emotionally evocative cues of the effect points to demobilizati
increases external efficacy from 0.54 to 0.44. The sadness between the sadness ad with the cues
cues decrease external efficacy from 0.54 to 0.46. without the cues were consistently n
Predicted probabilities were also generated based on in the case of volunteerism did sadn
the respondents' reported intention to participate in tional levels of statistical significan
upcoming elections. Specifically, the probabilities that While these results are informati
one would "definitely" vote as well as whether one would an effect of anger on participat
"definitely" or "probably" volunteer in the next election lates of participation, the data are lim
were calculated in all treatment conditions. For voting, that the college student participan
anger again boosts participatory intentions. The probabil- subset of the population (Sears 1
ity of voting is 0.56 in the anger ad/emotional cue condi- laboratory environment is sufficien
tion, whereas the probability is 0.44 in the anger ad/no the typical "living room" setting in w
cue condition. The probability of voting in the sadness are typically consumed. Study 3 a

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
422 Political Research Quarterly 66(2)

these limitations by employing public opi


emotion-based ad-tracking data to examine
quences of campaign advertising. From the
construct a measure of emotional ad exposure, which is _ , ,, . , ,
jt j• , ,. . , , , Dependentvariables
used to predict participatory behavior. r
Political participation was constructed by su
responses to four items—whether one tried to in
Study 3: Emotions in the 2004 another's vote, attended meetings, did other work
Presidential Campaign candidates or parties, contributed mo
To examine the consequences of emotional ads during an yard sign (kr-20 = .65). Efficacy
actual campaign, independent coders classified all adver- single item: "People like me do n
tisements included in the 2004 WISCAD presidential government does." I also analy
advertising data set (Goldstein and Rivlin 2007). The ads respondent closely follows
were classified according to their degree of anger, fear, which is averaged from two que
hope, and sadness. I use hope as a general measure of measure of political discussion was c
enthusiasm. Although hope and enthusiasm are clearly items regarding how much the re
distinct emotions, I rely on hope—as opposed to enthusi- tics with friends, coworkers, a
asm or other positive emotions—for the following reasons: Given the observational natur
First, scholars have demonstrated that positive emotions are important control variables w
typically more difficult to differentiate and that positive Media consumption was appro
emotions often have relatively similar information process- number of days the individu
ing consequences (Fredrickson 1998, 2001). Second, AIT work news, and late-night comedy
has generally used "hope" and "pride" as measures of = .50). Knowledge was generat
enthusiasm. Finally, from an empirical perspective, I age number of correct respon
expected that enthusiasm and pride, relative to hope, would edge test (kr-20 = .88). Respo
be difficult to detect in campaign ads, thus leading to lower by 1,000), gender, race, educa
interrater reliability estimates. were also included in the models. Finally, I include the
I rely on the 2004 WISCAD data set, which includes natural logarithm of the number of ads released
ad storyboards and detailed tracking information of ads media market and cable television through the tim
released in the 2004 presidential campaign (Goldstein respondent's interview.
and Rivlin 2007). Coders were first trained to categorize To estimate the consequences of emotion ad e
the ad storyboards according to the ad's emotional tenor. I estimated a two-level hierarchical model,
Ads were coded as 1 if the ad made an appeal to the respondents were modeled as nested within thei
emotion, 0 otherwise.7 A total of 656 advertisements market. Discussing politics and following politics/cam
were coded by three trained coders. To calculate inter- were treated as continuous dependent variables; effi
coder reliability, 96 ads (14.6 percent) were randomly and participation were treated as ordinal depend
selected and coded by two coders. Intercoder reliability ables. Accordingly, linear random intercept models
was good for anger and enthusiasm, marginal for fear, estimated for the continuous dependent variables, w
and somewhat low for sadness, indicating that sadness proportional odds models with random interce
is somewhat more difficult to reliably detect in cam- estimated for the ordinal variables.9
paign advertisements.8 Table 3 displays the estimates for the four models.
Across these four models, anger emerged as mobilizing
Ad Exposure emotion, enhancing interest, discussion, participation, and
" efficacy. F ear, however, had the opposite effect, by decreas
The coded ads were next merged with the 2004 NAES ing political discussion, interest in politics, and par
data. Respondent-level measures of anger, sadness, fear, tion. The effects of enthusiasm and sadness were some
and enthusiasm ad exposure were created from the num- contrary to expectations. Enthusiasm was not related
ber of presidential ads released in a respondent's media interest in politics but decreased participation in politic
market, weighted by the respondent's media consump- discussing politics, and efficacy. Sadness, similarly
tion habits. The weighting procedure, which is explained inconsistent effects; it enhanced discussion, interest,
in the online supplementary materials, is a variant of the participation but led to a nonsignificant effect on effi
strategy developed by Freedman, Franz, and Goldstein and political interest. With respect to sadness, the estim
(2004). The exposure variables were all recoded to vary should be interpreted with caution, as the interrater re
from 0 to 1 (anger exposure: M = 0.66, SD = 0.25; sad ability for sadness was somewhat low.

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 423

Table 3. Adve

Interest Discussion Participation Efficacy


Anger ad 0.186*** 1.379*** 1.511*** 0.581***
exposure (0.021) (0.150) (0.386) (0.212)
Sadness ad 0.045** 0.521*** 0.916*** 0.020
exposure (0.014) (0.099) (0.256) (0.132)
Fear ad -0.143*** -1.141*** -0.805*** -0.177
exposure (0.015) (0.110) (0.262) (0.154)
Enthad 0.019 -0.609 -0.775 -0.352*
exposure (0.020) (0.147) (0.363) (0.203)
Media 0.293*** 2.111*** 0.657*** 0.408***
consumption (0.008) (0.059) (0.176) (0.078)
Female -0.059*** -0.350*** -0.048 -0.466***
(0.002) (0.017) (0.053) (0.022)
College 0.074*** 0.411*** 0.359*** 0.825***
(0.003) (0.020) (0.062) (0.026)
Political 0.074*** 0.271*** 0.327*** 0.493***

knowledge (0.003) (0.024) (0.068) (0.030)


Income 0.063*** 0.682*** 0.417*** 0.715***
(0.003) (0.024) (0.073) (0.013)
Minority -0.017*** -0.025*** 0.010 -0.053*
(0.003) (0.024) (0.076) (0.032)
Age 0.002*** -0.017*** -0.003* -0.008***
(0.001) (0.001) (0.002) (0.001)
Registered 0.118*** 0.595*** 0.911*** 0.328***
(0.004) (0.026) (0.092) (0.034)
Total ads 0.014 1.146*** 1.598*** 0.316***

(0.009) (0.065) (0.192) (0.088)


Constant 0.248 0.573

x, 2.94 -0.137
x2 4.45 1.347
x3 5.43 1.366
x 6.50 2.344

x5 7.65
Variance ( 0.05922 3.23345
Variance 0.00163 0.02344 0.00809 0.01498
Markets
N 40,4013 44,233 5,669 27,614
N 88 88 88 87
Markets

Ad exposure and participation.The likelihood of being politically active and discussing politics, interest in politics, and levels of efficacy were modeled
such that individuals were nested within media markets. Political interest and discussion were modeled as a linear random intercept model; the re
maining models are random intercept proportional odds. Data are from the 2004 National Annenberg Election Study (NAES) and 2004 content-an
alyzed Wisconsin Advertising Project (WISCAD) data set, where storyboards were coded and merged with the tracking data.The coefficients and
standard errors associated with income have been multiplied by 100. Entries are unstandardized coefficients with standard errors in parentheses.
*p < . 10, **p < .05, ***p <.01, two-tailed.

To clarify the results from Table 3, the predicted lev- Similar to Table 3, Figure 3 attests to the participatory
els of discussion, interest, efficacy, and participation consequences of anger.
were estimated across the range of anger ad exposure. A consistent finding across these three studies is that
The remaining variables were specified to remain at anger is mobilizing by elevating the desire and impor
their respective means and modes. Because the inter- tance attached to political participation. Negative emo
cepts vary across media markets, simulated values were tions do not have uniform consequences; anger is unique
generated in four media markets—New York; Los relative to other negative emotions. Across these studies,
Angeles; Mobile, Alabama; and Des Moines, Iowa. the analyses demonstrate that anger generates heightened

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
424 Political Research Quarterly 66(2)

Anger and Political Participation


Interest Efficacy
New York New York
Des Moines 10
• - - Des Moines
O) to _
CD Los Angeles o
Los Angeles
Mobile — Mobile
o
CO to
O o

to

O o

CO
o

in
IT)
co -
o o

Political Discussion Political Participation


in _
o New York
co ~j —— New York
0 — Des Moines Des Moines
co ~ Los Angeles Los Angeles
to
Mobile
o
co _| Mobile
c\i

o
c\i
CM
d

to
o

o
o
! 1 1 1 r § -t 1 1 1 1 r
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Anger Exposure

Figure 3. Anger and participation.


The effects of anger on discussion, interest, efficacy, and p
probability of engaging in three or more participatory acts
strongly disagreeing with the statement "people like me hav
effects of anger ad exposure are simulated in four media

interest and involvement, attesting to the mobilizing, much of this dissatisfaction has been a
approach-oriented nature of this emotion. increase in spots better suited to elicit an e
inform. Yet as Brader (2006, 47) aptly notes, "A trem
dous gap exists between what most people believe abo
Discussion and Conclusion
the importance of emotional appeals (the 'convention
With the onslaught of television advertising in the 1950s, wisdom') and what is actually kno
scholars and pundits alike feared the adverse impact of appeals."
televised campaigns, noting that advertisements offered Despite conventional wisdom that em
little substantive information about political candidates cally important, a scarcity of scholarly wo
and would displace more personal modes of campaigning cally assessed the consequences of em
(Ansolabehere and Iyengar 1997; Geer 2006). And while ads. A notable exception is the pioneer
the electorate has grown increasingly dissatisfied with the Brader (2005, 2006). In several exp
state of American campaigns by staying away from the (2005) finds that fear-evoking messages st
polls in large numbers and showing distrust of politicians, attention to information, whereas enth

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 425

larger relian
retical expe
majority of
positive mes
ing whether
and knowled
and Goldst
finding that
sensus seem
among part
increase electoral participation. relation among specific emotions may be taken as evi
A potentially fruitful way to understand the conse- dence discrediting the discrete model (Marcus, Neuman,
quences of negative and positive campaign messages is to and MacKuen 2000). Yet as noted by Izard (1972), "Most
consider an ad's emotional appeal (Brader 2005, 2006). theorists who deal with discrete emotions have suggest
The studies in this article were designed to examine the that existence of pure emotion, such as pure fear or pure
behavioral impact of four commonly appealed-to emo- guilt, is probably fairly rare in day-to-day living and virtu
tions in advertisements—anger, sadness, fear, and enthu- ally impossible to obtain in the laboratory or in any other
siasm. In line with the theoretical expectations, the results research setting. I share this position" (cited in Watson an
showed that anger is unique when compared to other Clark 1992, 499). Study 1 demonstrates that while ange
negative emotions. Anger, an emotion marked by ele- sadness, fear, and enthusiasm are heavily correlate
vated perceptions of control, promotes the propensity to factors—in line with the dimensional perspective—these
become politically active and leads to an augmented emotions are sufficiently unique to be treated as separat
belief that one's behaviors are politically consequential. constructs.
On the other hand, fear and sadness had different effects, Setting aside the correlations among emotions, d
by lessening civic engagement. crete emotions have unique patterns of physiological,
neural, cognitive, behavioral, and motivational tenden
cies (Ekman, Levenson, and Friesen 1983). And in con
Toward an Integrated Model of trasting emotions of the same tone—particularly anger
Emotion in Politics
and fear—these emotions have unique electoral conse
These findings have both psychological and political quences (Valentino et al. 2011). Studies 2 and 3
ramifications, as they advance our understanding of the this position by demonstrating these emo
political consequences of emotion. While much work pits unique and differentiated consequences fo
negative and positive emotions against one another, there behavior.
is nuance in the consequences of emotions of the same The findings in these three studies point to
valence. The findings in these three studies underscore tance of an integrated model of emotion—e
the importance of discrete emotions. The emotions retaining the necessary complexity and nuance
explored were conceptually distinct, with unique behav- crete perspective and the parsimony of the dim
ioral consequences. perspective. The findings are congruent with the hierar
The question thus remains, does the dimensional or dis- chical model of affect developed by
crete approach better account for the nature of emotional and Clark (1999; see also Watson an
reactions felt in response to campaign messages? While 2009; Huddy, Feldman, and Casse
these results appear to support the discrete emotion per- this approach, affect is arrayed hie
spective, dimensional approaches have addressed the para- crete emotions at the first level, p
doxical effects of anger. Affective intelligence theorists affect at the second level, and a bipo
have introduced "aversion" as a subcomponent of the dis- sus "unhappiness" dimension at the
position system, noting anger's similarities to positive Watson, and Clark 1999). In some
affect. While the term aversion would seem to denote with- "core affect" or general positivity
drawal and avoidance, Marcus, Neuman, and MacKuen be sufficient; in other circumstan
(2000) find that anger is qualitatively similar to enthusi- consider negative and positive
asm. However, scholars disagree on whether anger is (Lavine 2001); and in other circu
approach or avoidance oriented. Carver and Harmon-Jones expects differentiation among discre
(2009b), for instance, contend that anger is a negative, of analysis should be specific emo
approach-oriented emotion; Lemer and Tiedens (2006) cific emotions need not preclude e
argue that anger should be considered a positive emotion alities between emotional states. Giv

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
426 Political Research Quarterly 66(2)

Notes to campaign ads, a par


of emotional responses
advantageous strategy to understand campaign
rests on better understanding thea number
1. The first study also included consequence
of political particip
crete emotions. tion questions. There were significant differences in polit
While there are clear consequences of emotion for cal participation across the four conditions. However,
civic engagement, an important extension of this finding because a nontrivial amount of content varies across condi

is to simultaneously consider how emotions moderate tions, these results were not included. Cell sizes are as fol
information-processing strategies. While anger may have lows: enthusiasm n = 66; sadness n = 74; fear n = 73; anger
the normatively desirable consequences of producing a « = 74.

more active electorate, anger often has undesirable conse 2. Cell sizes are as follows: sad/no cue n = 73; anger/no cue
quence of leading to a greater reliance on heuristics, « = 69; sad/cue n = 78; anger/cue n = 54.
biases, and reduced attention to contemporaneous infor 3. It is important to note that a model absent the happiness
mation (MacKuen et al. 2010; Valentino, Gregorowicz, cross-loading provided an acceptable, albeit somewhat infe
and Groenendyk 2009). Anger may facilitate participa rior, fit to data (CFI = .982, TLI = .969, RMSEA = .108).
tion, though not necessarily thoughtful participation. Fear 4. The effects of the manipulation on experienced emotions
and sadness may increase effortful information process using traditional ANOVA and focused contrasts can be
ing, but with the adverse consequence of demobilization. found in the online supplementary materials at http://prq.
It is important to underscore the limitations of this proj sagepub.com/ supplemental/.
ect. On the whole, the project explored only four emo 5. A manipulation check indicated that the correct emotions
tions, yet other emotions may be politically consequential, were elicited in each condition. Specifically, an anger scale
such as disgust. Future work should more precisely detail was created based on reported disgust and anger (/polychoric =
the consequences of these emotions. Candidates also .64); a sadness scale was created from sadness and depres
appeal to a broader array of emotions, and an advanta sion (r , , . = .65). A 2 x 2 ANOVA was conducted with
polychonc
geous avenue of future research is to detail the conse emotional appeal and cues entered as covariates. For felt
quences of more emotional appeals. Moreover, future anger, the two-way interaction approached significance,
work should establish whether particular emotions reso F(l, 270) = 2.08, p < .15; the main effect of cue appeal was
nate differently with certain voters. For example, does marginally significant, F( 1, 270) = 3.06, p < .10; the main
political expertise moderate the consequences of emo effect of the emotion manipulation approached signifi
tional campaign messages? Perhaps with greater empiri cance, F(l, 270) = 2.43,/) < .12. The reason for a nonsig
cal scrutiny of specific emotions, we will more fully nificant interaction was that while anger was highest in the
understand why some campaigns are so successful in high cue, anger appeal condition (M = 0.586, SD = 0.191),
meeting the dual goals of persuasion and mobilization. the remaining means were approximately equal (anger/no
cue: M= 0.486, SD = 0.277; sadness/cue: M= 0.491, SD =
Acknowledgments 0.266; sadness/no cue: M= 0.482, SD = 0.279). Anger was
An earlier version of this article was presented at the 2007 more pronounced in the cue anger condition relative to the
annual meeting of the American Political Science Association. no cue anger condition (t = 2.29, p < .01, one-tailed),
I would like to thank the anonymous reviewers, as well as higher than the cue sadness condition (t = 2.25, p < .025,
Stanley Feldman, Howard Lavine, John Geer, Nicholas one-tailed), and higher than anger in the no cue sadness
Valentino, Charles Taber, Leonie Huddy, Ted Brader, Erin condition (i = 2.38, p < .025, one-tailed). Regarding
Cassese, Johanna Dunaway, Jim Garand, W. Russell Neuman, sadness, neither the two-way cue x emotion interaction,
George Marcus, Christopher Federico, Cassie Black, Amy F( 1, 270) = 0.26, p < .62, nor the emotional manipulation
Bradley, and participants in the LSU and Stony Brook main effect were significant, F{1,270) = 0.72, p < .40, though
University Brown Bag series. An online appendix containing the main effect for the cue was significant, F(l, 270) = 8.99,
supporting material can be found at http://prq.sagepub.com/ p < .01. The sadness ad absent the cue elicited less sadness
supplemental/. (M = 0.41, SD = 0.25) relative to the same ad with the cue
manipulation (M = 0.49, SD = 0.27), a difference that is
Declaration of Conflicting Interests significant (i = 1.88,/) < .05, one-tailed). However, the
The author(s) declared no potential conflicts of interest with anger ad with the cue manipulation elicited comparable
respect to the research, authorship, and/or publication of this levels of sadness (M = 0.48, SD = 0.25; f = 0.23, ns),
article.
though the anger ad lacking the emotional cue manipulation
elicited less sadness (M = 0.36, SD = 0.28) relative to the
Funding cue, sad ad (/ = 2.71,/) < .01, one-tailed). While the emo
The author(s) received no financial support for the research, tion manipulation was not as clean as anticipated—perhaps
authorship, and/or publication of this article. because of the complexity of the stimuli—the subsequent

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
Weber 427

contrasts m
Brader, Ted. 2005. "Striking a Responsive Chord: How Political
ness ad elici
Ads Motivate and Persuade Voters by Appealing to Emo
ad, on the
tions." American ot
Journal of Political Science 49:388-405.
anger.
Brader, Ted. 2006. Campaigning
Becau for Hearts and Minds: How
Emotional Appeals in PoliticalI
effects, Ads Work. Chicago:
an Univer
tive sity of Chicago Press.
test of t
6. Because
Brooks, Deborah. 2006. "The Resilient Voter: Moving of
toward
der isClosure in includ
the Debate over Negative Campaigning and
tive conclus
Turnout." Journal of Politics 68:684-96.
from the
Brooks, Deborah, and John Geer. 2007. mod
"Beyond Negativity:
7. Coders wer
The Effects of Incivility on the Electorate." American Jour
ment nal of Political
based Science 51:1-16.
ily Carver, Charles, and Eddie
the emot Harmon-Jones. 2009a. "Anger and
were Approach:
told Reply to Watson (2009) and to Tomarken
to and
date or advertiser and make an inference about the emotion Wald (2009)." Psychological Bulletin 135:215-17.
the ad was designed to evoke. This is similar to the proce Carver, Charles, and Eddie Harmon-Jones. 2009b. "Anger Is an
dure employed by Brader (2006). All ads were coded on a Approach-Related Affect: Evidence and Implications." Psy
graded scale from 1 (no appeal) to 4 (strong appeal). Coders chological Bulletin 135:183-204.
then classified ads based on a number of emotions. I analyze Clinton, Joshua, and John Lapinski. 2004. "'Targeted'Advertis
the degree of "fear," "sadness," "hope," and "anger" in the ing and Voter Turnout: An Experimental Study of the 2000
ads. Because I am interested not in the extremity of the Presidential Election." Journal of Politics 66:69-96.
appeal but rather in a count of how many appeals occurred Davidson, Richard, Klaus Scherer, and H. Hill Goldsmith.
in a media market, the variable was recoded to be a binary 2003. Handbook of Affective Sciences. New York: Oxford
variable (1 = emotion present, 0 = absent) and a count vari University Press.
able was generated. Diener, Ed, and Ashgar Iran-Nejad. 1986. "The Relationship in
8. After recoding the variables to 1 = emotion present, 0 = emo Experience between Various Types of Affect." Journal of
tion absent: anger: intraclass correlation (ICC) = .66, Personality and Social Psychology 50:1031-38.

V'lychonc = Percentage agreement = 83.30, k = .66; Ekman, Paul, Robert Levenson, and Wallace Friesen. 1983.
"Autonomic Nervous System Activity Distinguishes Emo
fear: ICC = .36, 'polychoric = -62, percentage agreement = 75.00,
tion." Science 221:1208-10.
k = .36; sadness: ICC = .07, rpolychoric = -47, percentage agree
ment = 86.46, k = . 11 ; enthusiasm: ICC =polychonc
.66, r , , =.91, Ellsworth, Phoebe, and Klaus Scherer. 2003. "Appraisal Pro
percentage agreement = 86.46, k = .66. cesses in Emotion." In Handbook of Affective Sciences,
9. All models were estimated in Stata version 10 using gllamm edited by Richard Davidson, Klaus Scherer, and H. Hill
(Skrondal and Rabe-Hesketh 2004). Goldsmith, 572-95. New York: Oxford University Press.
Fredrickson, Barbara. 1998. "What Good are Positive Emo
References
tions?" Review of General Psychology 2: 300-319.
Ansolabehere, Stephen, and Shanto Iyengar. 1997. Going Fredrickson,
Nega Barbara. 2001. "The Role of Positive Emotions
tive: How Political Advertisements Shrink and Polarize the in Positive Psychology: The Broaden-and-Build Theory of
Electorate. New York: Free Press. Positive Emotions." American Psychologist 56: 218-226.
Ansolabehere, Stephen, Shanto Iyengar, Adam Simon, and Freedman, Paul, Michael Franz, and Kenneth Goldstein. 2004.
Nicholas Valentino. 1994. "Does Attack Advertising Demo "Campaign Advertising and Democratic Citizenship."
bilize the Electorate?" American Political Science Review American Journal of Political Science 48:723-41.
88:829-38. Frijda, Nico, Peter Kuipers, and Elisabeth ter Schure. 1989.
Berkowitz, Leonard. 1990. "On the Formation and Regulation"Relations among Emotion, Appraisal, and Emotional
of Anger and Aggression: A Cognitive-Neoassociationistic Action Readiness." Journal of Personality and Social Psy
Analysis." American Psychologist 45:494-503. chology 57:212-28.
Geer, John. 2006. In Defense of Negativity: Attack Ads in
Berkowitz, Leonard, and Eddie Harmon-Jones. 2004. "Toward
and Understanding of the Determinants of Anger." Emotion Presidential Campaigns. Chicago: University of Chicago
4:107-30. Press.

Goldstein, Kenneth, and Joel Rivlin. 2007. "Presidential Adver


Bodenhausen, Galen, Lori Sheppard, and Geoffrey P. Kramer.
1994. "Negative Affect and Social Judgment: The Dilferential tising, 2003-2004 Combined File." Data set, final release.
Impact of Anger and Sadness." European Journal of Social Madison: University of Wisconsin Advertising Project,
Psychology 24:45-62. Department of Political Science.

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms
428 Political Research Quarterly 66(2)

Harmon-Jones, Cindy, MacKuen, Michael,Schmeichel,


Brandon Jennifer Wolak, Luke Keele, and George M
Eileen
Eddie Harmon-Jones. 2011.
Marcus. 2010. "The Expression
"Civic Engagements: of
Resolute Partisanship
nation: Similarities between
or Reflective Anger
Deliberation." American
and Journal
Approach
of Political
Science 54:440-58.
Positive Affect." Journal of Personality and So
chology 100:172-81. Marcus, George. 2003. "The Psychology of Emotion and Poli
Harmon-Jones, Eddie, tics."Vaughn-Scott,
Kate In The Oxford Handbook of Political
Sheri Psychology,
Mohr
Sigelman, and Cindy Harmon-Jones. 2004.
edited by David O. Sears, Leonie Huddy, "The
and Robert Jervis, E
Manipulated Sympathy 182-221.
andNew Anger on Left
York: Oxford University Press. and Rig
tal Cortical Activity." Marcus, George, and
Emotion Michael MacKuen. 1993. "Anxiety,
4:95-101.
Huddy, Leonie, Stanley Feldman, and
Enthusiasm, and the Vote: Erin
The Emotional Cassese
Underpinnings of
the Distinct Political Effects of
Learning and Anxiety
Involvement and
during Presidential Anger
Campaigns."
Affect Effect: American
Dynamics of Political Science Review 87:672-85.
Emotion in Political T
and Behavior, edited Marcus,
by George,
W. W.
Russell Neuman,
Russell Neuman, and Michael MacKuen. G
Marcus, Ann N. Crigler, and
2000. Affective Michael
Intelligence MacKuen,
and Political Judgment. Chicago:
Chicago: University of University
Chicago of Chicago Press.
Press.
Huddy, Leonie, Stanley Feldman,
Marsh, Charles
Abigail, Nalini Ambady, and Robert Kleck. Taber,
2005. "The
Lahav. 2005. "Threat, Effects of Fear
Anxiety, andand Anger Facial Expressions
Support foron Approach
Antite
Policies." American and Avoidance-Related
Journal Behaviors." EmotionScience
of Political 5:119-24. 49:5
Izard, Carroll Ellis. Muthén,
1972. Linda, and Bengt Muthén.
Patterns 1998-2010. Mplus User's A
of Emotions:
sis of Anxiety and Depression. San
Guide. 6th ed. Los Angeles: MuthénDiego,
and Muthén. CA: A
Press. Neuman, W. Russell, George E. Marcus, Ann N. Crigler, and
Kahn, Kim, and Patrick Kenney.
Michael 1999.
MacKuen, eds. 2007. "DoDynam
The Affect Effect: Nega
paigns Mobilize or Suppress Turnout?
ics of Emotion Clarifying
in Political Thinking and Behavior. Chicago:
tionship between Negativity and
University of Chicago Participation."
Press. A
Political Science Review 93:877-89. Rudolph, Thomas, Amy Gangl, and Dan Stevens. 2000. "The
Keltner, Dacher, Phoebe Ellsworth, and Kari Edwards. 1993. Effects of Efficacy and Emotions on Campaign Involve
"Beyond Simple Pessimism: Effects of Sadness and Anger ment." Journal of Politics 62:1189-97.
on Social Perception." Journal of Personality and Social
Sears, David. 1986. "College Sophomores in the Laboratory: Influ
Psychology 64:740-52. ences of a Narrow Data Base on Psychology's View of Human
Lau, Richard, and Gerald Pomper. 2004. Negative CampaignNature." Journal of Personality and Social Psychology 51:
ing: An Analysis of U.S. Senate Elections. Campaigning 515-530.
American Style. Lanham, MD: Rowman Littlefield. Skrondal, Anders, and Sophia Rabe-Hesketh. 2004. Generalized
Lavine, Howard. 2001. "The Electoral Consequences of AmbivLatent Variable Modeling: Multilevel, Longitudinal and Struc
alence toward Presidential Candidates." American Journal tural Equation Models. Boca Raton, FL: Chapman & Hall/
of Political Science 45:915-29. CRC.

Lazarus, Richard. 1991. Emotion and Adaptation. New York,Smith, Craig A., and Phoebe Ellsworth. 1985. "Patterns of Cog
NY: Oxford University Press. nitive Appraisal in Emotion." Journal of Personality and
Lerner, Jennifer, and Dacher Keltner. 2001. "Fear, Anger, Social Psychology 48:813-38.
and Risk." Journal of Personality and Social PsychologyTellegen, Auke, David Watson, and Lee Anna Clark. 1999.
81:146-59.
"Further Support for a Hierarchical Model of Affect: Reply
Lerner, Jennifer, and Larissa Tiedens. 2006. "Portrait ofto
the
Green and Salovey." Psychological Science 10:307-9.
Angry Decision Maker: How Appraisal TendenciesValentino,
Shape Nicholas A., Ted Brader, Eric W. Groenendyk,
Anger's Influence on Cognition." Journal of Behavioral
Krysha Gregorowicz, and Vincent L. Hutchings. 2011. "Elec
Decision Making 19:115-37. tion Night's Alright For Fighting: The Participatory Impact of
Levenson, Richard. 1992. "Autonomic Nervous SystemNegative Emotions." Journal of Politics 73:156-70.
Differences Among Emotions." Psychological Science
Valentino, Nicholas A., Krysha Gregorowicz, and Eric Groenendyk.
3:23-27.
2009. "Emotions, Efficacy, and the Habit of Participation."
Political Behavior 31:307-30.
Lodge, Milton, and Charles S. Taber. 2005. "Automaticity of
Affect for Political Candidates, Parties, and Issues:Watson,
An David. 2009. "Locating Anger in the Hierarchical
Experimental Test of the Hot Cognition Hypothesis." PolitiStructure of Affect: Comment on Carver and Harmon
cal Psychology 26:455-82. Jones." Psychological Bulletin 135:205-8.
Mackie, Diane, Thierry Devos, and Eliot Smith. 2000. "Inter
Watson, David, and Lee Ann Clark. 1992. "Affects Separable
group Emotions: Explaining Offensive Action Tendencies and Inseparable: On the Hierarchical Arrangement of the
in an Intergroup Context. " Journal of Personality and
Negative Affects." Journal of Personality and Social Psy
Social Psychology 79:602-16. chology 62:489-505.

This content downloaded from 207.162.240.147 on Sat, 13 May 2017 18:01:35 UTC
All use subject to http://about.jstor.org/terms

You might also like