The Affective Effect on Political Judgment

Abstract

This study examines the impact of affect on candidate evaluation and voting intention by conducting an experiment using three treatments: positive, negative, and neutral nonverbal expressions of a fictional congressional office-seeker. Three issues were addressed in the TV interviews. Results show that candidate image exerts a stronger influence on viewers’ voting intention than the candidate’s stance on issues, controlling for viewers’ prior attitudes toward those issues. In addition, negative affect is more powerful than positive, reinforcing the belief that making a good impression will not help a candidate as much as a bad impression will hurt.

Keywords

affect agenda setting communication effects political communication quantitative methodology

Recent scholarly attention has taken note of the fact that humans are fundamentally emotional in democratic practice.¹ Even though voters may be educated and informed and may intentionally endeavor to be rational, their political judgment still can be clouded with subjective feelings and personal penchants. It is common for individuals to attach affective information to candidates;² moreover, the most readily accessible information voters have is their own emotional responses to political figures. Given this, the attributes of political candidates can be a decisive factor in determining votes.³ People can get a glimpse of a candidate and immediately form crucial, affective opinions. Sometimes their judgments rely exclusively on mediated images or advice from others without careful study of the candidate’s issues and platforms. It is, therefore, not surprising that seasoned political consultants advise public office-seekers to be cautious about their appearance and demeanor.⁴

Of course, images of public office-seekers are not the only thing covered in the media. Lengthy, detailed discussions about important issues also are fodder for the current 24/7 media system. The stances political candidates take (or do not take) toward key issues, particularly the thorny ones, are unquestionably newsworthy. But it is challenging to gauge the extent to which issues—and which of the key issues—actually determine whether and how people cast their votes. Furthermore, it is unclear which of the two forces—attributes or issue stance—has a stronger impact when people make a decision at the ballot box. This study seeks answers to these questions using an experiment with three levels of affect for the same candidate. The affective manipulation is operationalized by positive, negative, and neutral nonverbal behaviors of a professional actor playing a congressional office-seeker in televised news interviews. Three issues—the environment, economy, and crime—were addressed in different episodes of the interviews recorded in a TV studio. The results of this experiment shed light on political communication theories of second-level agenda setting⁵ and the affect effect,⁶ and they expand our knowledge about how individuals process political information.

Literature Review

Agenda Setting

The theory that speaks most directly to our question of whether attributes or issues play a stronger role in people’s political decision-making is agenda setting. The first level of the theory focuses on the impact of issue salience, and the second level hones in on affective impressions, such as the kind provided by candidates’ appearance attributes. Since McCombs and Shaw’s seminal piece on the agenda-setting effect,⁷ hundreds of studies have documented the impact of salient media coverage on the public’s perceived importance of issues⁸ and attributes of issues.⁹ Both levels have been shown to be effective in predicting people’s perceptions.¹⁰ However, no research has explored the influence from these competing sources in a controlled experimental setting; consequently, direct evidence of causality is absent. That is the first aim of this study, to determine whether issues (first level) or attributes (second level) have a greater influence.

In these studies, many different terms are used to refer to the same or similar concepts, including the terms image, attributes, affect, affective traits, and various combinations of these words. In the interest of clarity, this study uses the term attribute or appearance attribute to refer to the concept of the image or impression a candidate conveys to voters by way of his or her positive, negative, and neutral nonverbal behaviors, including expressions, gestures, and tone of voice.

Priming

The agenda-setting effect itself does not address the question of whether the more the media cover an issue, the more the public knows about it and subsequently uses issue stance to evaluate the candidate, so we turn to priming theory. Evidence of priming effects enables researchers to conclude that the prominent issues do provide voters with a criterion to evaluate politicians, which directly or indirectly supports the impact of salient agendas in a campaign. Political science literature also has demonstrated the importance of candidates’ issue stances on voter decisions, while some show that a significant proportion of voters base their votes on a single issue.¹¹ Before comparing the influence of issues and attributes on political judgment, it is imperative for us to examine whether the salient agendas lead viewers to make decisions about the candidates. It is important to establish that issues are important before comparing their impact with that of appearance attributes. Therefore, the following hypothesis is formed.

H1: The extent to which viewers care about an issue addressed by the candidate (issue congruence) is positively related to their support for the candidate.

Even though the nexus between issue agenda setting and voting preference is not well documented, the connection between candidates’ attributes and voting decision seems more straightforward. For example, researchers¹² using data from the 2000 elections in Poland and the United States found that advertised images of candidates influenced voting intention. Also, second-level agenda-setting effects have been found in various election studies in different countries.¹³ Typically, the traditional manipulation of candidate attributes is limited to the textual description of the candidate in a news story. As studies¹⁴ have noted, it is important to encompass other domains that provide direct information about a candidate’s image. The nonverbal dimension of affect is rarely tackled in the political communication literature, yet it is well established in psychology that nonverbal properties can be objectively judged and that the nonverbal channel of communication is an excellent conveyor of affective information, especially emotions.¹⁵ The present study aims to fill this gap by examining the impact of positive and negative nonverbal behaviors delivered by the same candidate on viewers’ assessment of that candidate.

As discussed above, the attributes of candidates can be crucial cues for the electorate. One review of the literature¹⁶ concluded that voters’ perceptions of candidates can substantially sway their votes. Boyd¹⁷ found that 38.6% of the variance in voting defection can be accounted for by perceptions of candidate appearance. Stamm’s team¹⁸ went even further to say that issues have nothing to do with elections and that the candidate’s image is the decisive element in any election. Although more evidence supports the impact of candidate attributes on voting, the comparison between the influences derived from issue and attributes is scarce—but needed. Kiousis and McCombs¹⁹ were first to tackle the comparison between first- and second-level agenda setting. Their findings ushered in the idea that the second-level, or affective agenda setting, might exert a larger influence on people’s attitudes. One study²⁰ using survey data from the 2004 presidential election confirmed that, indeed, the second-level agenda has a greater impact on the public’s perception. Based on the above literature, the following hypotheses were formed:

H2: The candidate’s positive or negative appearance attributes will affect viewers’ affective evaluation of the candidate.

H3: Viewers’ affective evaluation of a political candidate is associated with their voting intention; the more positive the affective evaluation, the more likely the viewer is to vote for the candidate.

H4: The candidate’s appearance attributes will influence viewers’ voting more than his issue stance.

Affective Intelligence

Affective intelligence theory²¹ maintains that emotions are critical because they produce more attention to political issues, and people are more likely to be receptive to negative cues when forming their political opinions. Marcus, Neuman, and MacKuen²² even contend that “emotions enhance citizen rationality,” which rebuts the conventional thinking that people’s emotions can cloud judgment and lead to irrational behavior. Their series of studies found that specific negative emotions led to arousal, which stimulated cognition that resulted in considered judgment. One particular study²³ found such effects based on candidates’ personal attributes.

There is already evidence in agenda-setting literature of negative information’s predominant influence over its positive counterpart. Wanta and colleagues²⁴ found that positive coverage of a nation generated no impact on the public, but negative coverage led to negative perceptions toward that foreign nation. In another relevant study, negative economic news could predict consumer expectations, whereas positive news did not.²⁵ In addition, some research indicates that negative information is more important than positive information in political judgments, plays a greater role in voting behavior,²⁶ and that individuals rely on it more than positive information when forming impressions of others.²⁷ Therefore, we develop the following hypothesis to examine the impact of the valence of a candidate’s appearance attributes:

H5: The candidate’s negative appearance attributes will influence viewers’ assessments of the candidate more than the positive counterparts.

Method

Stimuli

This experiment used three distinct stimuli to test the impact of nonverbally based appearance attributes on perception and voting intention: positive nonverbal video, negative nonverbal video, and a neutral audio-only recording for the control group in which subjects were not exposed to a video. The varied footages were all produced in a digital TV studio to resemble an in-depth interview conducted on a typical news program. A professional actor played the candidate; the actor was a white male in his early forties who specializes in facial and bodily movement, has years of professional experience in films and theater, and teaches at a state university in the South. A forty-year-old female doctoral student who had worked as an anchor in various TV stations played the anchor role. The fictional candidate is running for a congressional seat in Kansas, which the participants—from the South and Southwest—would know little about. Such a manipulation is imperative for the research because it deprives the participants of using their existing knowledge or familiar cues to form hasty electoral decisions and thus facilitates deliberative choices.²⁸

As a repetition factor, the scripts of the interviews covered three common campaign issues that consistently are reported in public opinion polls as among the most important problems in the country: the environment, the economy, and crime. The three issues were chosen primarily because they represent common issues in political campaigns. However, they did not represent “typical” Republican or Democratic issues, allowing us to avoid attaching a party label to the candidate. Thus, it was less likely participants would rely on party cue and would more likely use the appearance attributes when assessing the candidate. Each of the interview episodes focused on one issue and ran about eight minutes, typical for such a TV interview format.

The scripts about the environment, the economy, and crime for both positive and negative conditions were identical; also the candidate and the news anchor were identical. Therefore, the only thing that differed was the way the political candidate delivered the message in a positive or negative manner; a neutral audio-only recording provided the control condition where the tone of voice was neutral and no appearance attributes could be seen without video. The manipulative component is the nonverbal treatment—the professional actor who played the candidate exhibited positive or negative gestures, facial expressions, body movement, and tone of voice. For example, the political candidate, played by the professional actor, leaned toward the camera, and thus the viewer, and spoke in a pleasant tone of voice when delivering positive appearance attributes. In contrast, he leaned back with arms crossed, did not make eye contact with the camera, and spoke angrily or impatiently when he delivered negative appearance attributes.

Manipulation Check

A manipulation check showed that the positive, negative, and neutral stimuli were, indeed, perceived as intended by participants. Ninety students from a large southwestern university not used in the main study participated in a check of the nonverbal manipulations. Thirty-five saw the positive version of the video, thirty-eight saw the negative version, and seventeen heard the control version with audio only. The appearance attribute variables—facial expressions, body posture, gesturing, and overall appearance—were measured on 7-point Likert-type scales ranging from 1 = extremely negative to 7 = extremely positive and indexed (α = .91). There were significant differences between the positive and negative indexes in the predicted direction (F = 70.43, df = 1, 71, p < .001). The mean for the positive version was significantly higher (more positive) at 17.69 (SD = 4.80) than the mean of 9.55 for the negative version (SD = 3.39). There also were significant differences in the tone of voice variable, also measured on a 7-point Likert-type scale, (F = 40.55, df = 2, 88, p < .001); the positive version was rated significantly more positively (M = 5.17, SD = 1.40) than the negative version (M = 2.42, SD = 1.44) and the neutral version (M = 3.53, SD = .51); the negative version was rated significantly more negative than the neutral version.

Procedures

Participants were recruited from undergraduate mass communication classes in two state universities in the South and Southwest. This was a between-subjects design with each participant given all three issue episodes in one of the three conditions, either the positive, negative, or neutral stimuli. The order of the three issue episodes (environment, economy, and crime) was randomly rotated to reduce potential order bias. Participants were asked to fill out a pre-test that included questions about their opinion on the most important issues facing the country, their attitude toward elections and politics (five questions, including “Voting is a hassle” and “Politics are dirty”), and their views on the three issues, which were used to control for pre-existing attitudes toward the issues. After each viewing, participants were asked to assess the candidate’s appearance attributes, evaluate the candidate’s stance on the three issues, and gauge their likelihood of voting for the candidate. A total of seventeen attributes were measured (α = .96): overall appearance, tone of voice, appeal, liking, honesty, intelligence, reliability, qualified, trustworthy, competent, experienced, compassionate, approachable, visionary, moral, good leader, and cares about people like me.

Results

The experiment included 230 participants, 84 in the positive group, 81 in the negative group, and 65 in the control group who heard the neutral interview recording without seeing video. Of the participants in three sessions, 34% were male, 69% white, and 14% African American. Thirty-six percent of the subjects reported they were Republicans, 34% said they were Democrats, and the rest were either independent or undecided.

This study used three different issues to examine the nexus between exposure, concern level, and voting likelihood. A pre-test (7-point scale: 1 as least important and 7 as most important) found that participants’ prior evaluations of the three issues differed significantly (F = 6.19, df = 2, 687, p = .002). The economy was the most important issue (M = 6.01, SD = 1.22), followed by crime (M = 5.74, SD = 1.21). The environment issue was rated least important (M = 5.59, SD = 1.46). The difference between the ratings toward the economy and environment was statistically significant.

H1 examined the extent to which viewers care about an issue addressed by the candidate and its relationship to their support for the candidate. All three issues showed significant and positive associations between concern and voting (environment r = .75, n = 228, p < .01; economy r = .71, n = 228, p < .01; crime r = .71, n = 230, p < .01). H1 was supported.

Having established that concern about the issues is related to support for the candidate, this research turned to its primary focus of the impact of the candidate’s varied appearance attributes on viewers’ affective evaluations. Seventeen attributes were included in the post-test, which were subsequently extracted with the Varimax method into two principal components—candidate “image” and “capability.” These two were examined separately as two dependent variables in the following statistical tests. The candidate “capability” factor was loaded heavily on being qualified (.84), experienced (.80), intelligent (.71), and competent (.82), whereas the candidate “image” factor was loaded heavily on overall appearance (.75), being approachable (.82), tone of voice (.77), liking (.62), and appealing (.66). The entire test results for the third hypothesis can be seen in Tables 1 and 2. The six analyses of covariance (ANCOVA) tests (candidate “image” and “capability” across three issues, controlling for original issue concern) indicated that the varied nonverbal treatments led to strikingly different evaluations of the same candidate. The impact on candidate “image” evaluation under three issue conditions resulted in a significant finding (F = 6.92, df = 6, 450, p < .01, η² = .16); the finding also was significant for candidate “capability” (F = 11.80, df = 6, 450, p < .01, η² = .25).

Table 1.

Candidate “Image” Evaluation under Different Issues and Conditions.

	Appearance attribute manipulation
Issue	Positive	Negative	Control
Environment	.10_a	−.41_ab	.38_b
Economy	.12_c	−.33_cd	.26_d
Crime	−.03	−.17_e	.25_e

Note. Presented in the cells are means of the “image” factor scores. The means that share the same subscripts are different with statistical significance less than .05.

Table 2.

Candidate “Capability” Evaluation under Different Issues and Conditions.

	Appearance attribute manipulation
Issue tested	Positive	Negative	Control
Environment	−.26_a	−.06_b	.41_ab
Economy	.40_c	−.52_cd	.13_d
Crime	.39_e	−.51_ef	.14_f

Note. Presented in the cells are means of the “capability” factor scores. The means that share the same subscripts are different with statistical significance less than .05.

Post hoc tests showed that the majority of the pairs were significantly different. Looking at Table 1, we can see that the candidate’s “image” was perceived more positively when participants were exposed to positive appearance attributes than when they viewed negative appearance attributes. However, the candidate appeared to suffer somewhat when the subjects saw him—he was actually rated higher when the subjects only listened to what he said about the three issues. Evaluations of the candidate’s capability resulted in the direction expected (most positive evaluation for positive appearance attributes, followed by the evaluation of the control group, and then negative appearance attributes; see Table 2) when economy and crime were discussed; there was no significant effect of appearance attributes on the topic of the environment. It is worth pointing out again that the participants’ prior concern levels toward the environment were the lowest of all issues. Overall, the evidence supported H2 that the valence of appearance attributes leads to varied affective evaluations.

H3, that viewers’ affective evaluations of the political candidate will influence their support for him, was firmly supported. Partial correlations between viewers’ affective assessments (“image” and “capability”) of the candidate and their likely support for his congressional bid under the three issue conditions were all positive (albeit moderate) and significant (see Table 3).

Table 3.

Partial Correlations between Voting Intention and Candidate Appearance Attribute.

Voting intention for the candidate when he talks about:	Appearance attribute
Voting intention for the candidate when he talks about:	“Image”	“Capability”
Environment (N = 225)	.43	.35
Economy (N = 225)	.29	.45
Crime (N = 227)	.49	.34

Note. Presented in the table are Pearson product moment partial correlation coefficients (controlling for issue congruence), all of which are significant at .01 level.

H4 was that the candidate’s appearance attributes will influence viewers’ support for him more than his issue stance. This test aimed at finding the influence of appearance attributes on support for the candidate above and beyond the influence from the congruence of issue stances. In each of the regression models conducted, issue congruence as well as the two attribute factors were used as independent variables (see Table 4). On the environmental issue, the two extracted attribute factors (candidate “image” and “capability”) exerted significant influence (Beta values .48 and .37, respectively) on voting intention for the candidate. The influence from issue concern also was significant, but its Beta value is slightly smaller (.32). The same phenomenon took place when the economy was the issue. The two attribute factors represented an even larger force in predicting viewers’ support (.44 and .51, respectively), compared with the much smaller Beta value from correspondence of issue concern (.21). When the candidate talked about crime, the attribute factors also yielded a stronger impact on viewers’ voting intention than did agreement on issue concern. Their Beta values are .54 and .44, respectively, far larger than the issue concern (.20). It is worth noting that all three regression models are statistically significant with p value less than .001, and the adjusted R-square values range from .67 to .71. Therefore, H4 also was supported.

Table 4.

Predicting Support for Political Candidate by Issue Congruence and Candidate Image and Capability.

Issue	Predictor	B	R ²	F
Environment (N = 229)	Issue congruence	.32*
	Candidate image	.48*	.69	170.02*
	Candidate capability	.37*
Economy (N = 228)	Issue congruence	.21*
	Candidate image	.44*	.68	155.05*
	Candidate capability	.51*
Crime (N = 230)	Issue congruence	.20*
	Candidate image	.54*	.71	184.72*
	Candidate capability	.44*

p < .001.

H5 tried to verify the main thesis of the affective intelligence theory: people are more likely to be receptive to negative attributes than positive attributes in forming judgments of candidates. Affective intelligence theory predicts the candidate’s negative appearance attributes will influence viewers more than the positive counterparts. Multivariate analysis of variance (MANOVA) was used to see if the different treatments (positive, negative, and neutral) led to varied perceptions about the same candidate across the three issues. The overall model (including all issues and treatments) showed that the different attribute conditions, indeed, generated significantly different evaluations of the candidate—F = 8.42, df = 12, 444, p < .001, η² = .34. Six post hoc tests (Tukey HSD and Scheffe) also were used to see if negative attributes were significantly different from others. The twelve-pair comparisons (see Tables 1 and 2) indicated that, except for two comparisons where the negative treatment did not differ from the positive treatment, negative treatments were significantly different from the neutral as well as positive counterparts across the board. On the other hand, the influence of the positive treatment was significantly different from the neutral treatment only once. Thus, the majority of the evidence supports H5.

Discussion

The results of this experiment confirm all the proposed hypotheses and unveil an intriguing finding that supports what many scholars and political pundits have long suspected: A candidate’s attributes can be a decisive factor in election outcomes. Not only can a candidate’s appearance attributes lead to different perceptions of the candidate by viewers, but they also can predict voters’ support in the voting booth. Therefore, the appearance attribute hypothesis—drawn from both second-level agenda setting and affective intelligence theory—is confirmed with this study. Anecdotes about winning determinants from past elections include the successful candidate’s height, accent, body shape, or even hair as decisive factors, which could be considered for future research. In this study, those variables were held constant by having the same actor appear as the candidate in all the conditions; future studies could do the same using different aspects of a candidate’s appearance attributes.

Another significant finding of this study is that the impact of candidate attributes on voting intention is larger than that of issue congruence. Therefore, the investigation of the relative strengths of first-level and second-level agenda-setting results is a concrete conclusion. Despite its theoretical contribution, this finding may seem surprising, particularly to those who believe in rational discussion and participatory deliberation about crucial issues in political campaigns. What is found here departs from the ideal practice of a true democracy and unveils intriguing lessons for those who want to serve in the public affairs arena.

One portion of the affective intelligence theory that is supported explains that viewers’ exposure to negative appearance attributes has a larger influence on affective evaluations than positive appearance attributes. So, on one hand, political consultants’ persistent advice to candidates to attend to their appearance, wear suitable outfits, and make appropriate gestures, makes sense. On the other, evidence of more influence from negative attributes provides indirect empirical support for portraying one’s opponent negatively. It also reinforces the belief that making a good impression will not help a candidate as much as a bad impression will hurt.

The three issues under investigation generate slightly different results—most notably with the issue of the environment deviating from the other two. People’s initial attitude toward an issue and their ideological orientation could be at work. Based on the pre-test, the environment issue did not register for our subjects, whereas the issue of the economy was rated overwhelmingly high. Given this, an intricate interaction between issue consideration and candidate appearance attributes could have taken place—the issue deemed less important, such as the environment, is more likely to challenge viewers on a cognitive level and, therefore, more likely to be affected by valenced exposure. On the other hand, the highly rated issue, such as the economy, is less likely to be influenced by valenced candidate attributes. Therefore, voters’ prior issue stance could be a key conditional factor on level of concern, which should merit further investigation.

The study has limitations that should be addressed. First, as mentioned above, the choice of issues should be expanded and diversified. Also, the attributes of the actor (middle-aged white male) used in the experiment might limit the findings, although most political candidates still fit this demographic. Future studies might consider using actors of different gender or races and of varied charismatic levels to test the impact on viewers. In addition, other components of candidate affect—other than appearance attributes—should be taken into consideration as well. For example, using candidates of different height or body shape could produce dramatic difference in evaluations—although it would be technically challenging for researchers to hold many factors constant.

Conclusion

Voters are not the highly rational animals that democratic theory supposes them to be. Feelings, emotions, and even first impressions are more important than most of us would like to admit.²⁹ People form crucial impressions based on something as simple as whether a candidate makes a good impression or a bad one, and those impressions weigh heavily in their voting decisions. A common criticism of the media is that they cover a candidate’s image more than the issues. Just give voters the information they need to make an informed choice, critics say; tell them where the candidates differ on their platforms and policies, and people will use that information to make better-informed political choices. This study shows that is not necessarily so. Even when given rational, issue-based information, the participants in this controlled experiment used their affective impressions of a candidate’s appearance attributes to a greater extent in their voting decisions. It mattered less what the candidate said than how he said it. The image a candidate conveyed in his appearance and nonverbal behavior had a stronger impact than his stance on three different issues.

The main focus of this study was to assess the hierarchy of the two levels of agenda setting—issues, at the first level, and affect at the second level. The predominance of affective information over substantive information has previously been³⁰ based on survey data; however, that study was correlational and could not establish order or claim causality. It offered a good basis for proposing that affective information was more powerful than substantive information on issues in forming voting decisions, and this experimental study gives us the direct evidence of such a cause-and-effect relationship. Other evidence also points to the superiority of affective information over issues in voters’ decision making,³¹ but the strength of this study is in controlling for extraneous factors and establishing a causal connection. This is not to say that substantive information on issues exerts no effect on voters—it does indeed. But its power appears to be substantially less than that of affective information, at least in this experimental setting.

In addition, this study confirmed the hypothesis that the negative exerts a more powerful force than does the positive. The idea of a bias toward negative information is not new, but this study shows that if a candidate makes a negative impression, it will be more powerful than if he or she makes a positive one. Just as with Howard Dean’s discovery in the 2004 campaign with his “Dean Scream,” or with Tina Fey’s impersonation of Sarah Palin on Saturday Night Live in 2008, negative appearances can be all that people remember.

Agenda-setting theory can benefit from incorporating these theoretical advances. Specifically, the second level of agenda setting, which deals with people’s affective impressions, should be recognized as having a more powerful influence on voters than the first level, which deals with substantive issues. Second, within that affective second level, negative information exerts a stronger influence than positive information. These two theoretical ideas should be incorporated into the theory for a fuller understanding of how the media transfer their agendas, both affective and substantive, to the public. These theoretical statements suggest a more complex theory of agenda setting that adds to our understanding of and our ability to predict the media’s effects on people’s political knowledge. Our more complex account of the relative impact of affect and emotion on voters offers new insights into how the media shape political attitudes and behavior. What are the implications for the media and democracy from these findings? Rather than take this as cause for despair, we suggest that it enhances our understanding of emotion and rationality in the political process. Journalists, scholars, and voters should not resign themselves to being manipulated by campaign managers who know how to present a candidate for maximum effect. Rather, this information can be used to educate voters as well as the media. The processes by which affective information makes impressions on people are implicit ones—that is, people are not always aware of how appearance attributes of candidates are affecting their thinking. Information about implicit versus explicit information and cognitive versus affective processing should form the basis of media literacy courses in schools and should be taught in even more depth in journalism schools and mid-career workshops. Currently, topics dealing with appearance attributes are the purview of interpersonal and political communication courses, not journalism programs. This should be changed.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.