Facing Political Leaders : a study on how personality traits and political preferences can influence the effect of facial dominance on leadership evaluations in political candidates

(1)

Facing Political Leaders

A study on how personality traits and political preferences can influence the effect of facial dominance on leadership evaluations in political candidates.

Daphne Spreeuwenberg - 11889225 Master Thesis

Graduate School of Communication – University of Amsterdam Political Communication and Journalism

28/06/2019

Supervisor: Franziska Marquart Word count: 7496

(2)

Abstract

The role of political candidates is becoming more important. Voters increasingly focus more on choosing between leaders than parties. They also tend to base their political decisions on intuitive cues, like appearance. Facial characteristics are an important factor in evaluating political candidates on leadership abilities. Previous research has shown a relation between political orientation and the evaluation of facial dominance. It has also been argued that political orientation emanates from personality traits. Even though facial dominance has been widely studied, a joined investigation was still missing. The current study extended previous findings in order to embed these into a larger, overarching research design. In an online experiment, Dutch participants were shown either a dominantly or non-dominantly morphed picture of a political candidate. They were asked to evaluate these on eight different

leadership characteristics, and participants were also tested for the personality traits openness and conscientiousness. Results showed no significant relation between personality traits and political preferences or candidate evaluation. There was no interaction effect found of facial dominance and personality traits, or facial dominance and political preference, on both candidate evaluations and voting intention. Overall, the dominant stimulus was evaluated more positively on the item ‘good leader’, regardless of the political orientation of the participants. A mediation analysis, aiming to test an overarching model combining multiple findings, showed no indirect effect of facial dominance on voting intention through candidate evaluation and no moderation effect of political preference.

(3)

1. Introduction

The most important and well-known aspect of democracies is the right to vote. Voters have three choices to make in order to make a substantiate vote. First, citizens will have to decide what party they prefer. To do so, they will have to, second, take into account the parties’ policies and, third, its leader, who eventually will enact the policies. It is questionable whether these factors are all equally important. Aarts, Blais and Schmitt (2011) argue that the role of party leaders has been largely neglected in literature on voting behaviour, whereas there is a clear dominance of citizens’ party identification and ideology. But is this justified? And even if the role of party leaders only plays a small role in voting decisions, how does it work?

Political parties are primarily represented by their political candidates. They often function as figureheads that communicate their party’s position on issues to citizens. Major political candidates and party leaders are recognized more prominently than others,

suggesting they play an important role (McAllister, 2007). Political leaders have become more important over time (McAllister, 2007), partly due to the enormous changes within political communication in the past decades. Strömbäck (2008) showed how media play an increasingly relevant role for politics, and how political leaders adapt to the media logic, thereby getting more dependent on media to reach their voters. With these so-called ‘media-centred democracies’, political election campaigns have developed as well. A result of this is the increasing personalization of politics (Swanson & Mancini 1996; Schmitt-Beck & Farrell, 2002, as stated in Ohr, 2013), in which personal qualities of candidates are more prominent in both campaign materials and media coverage (Ohr, 2013). However, this partly depends on the type of democracy: In presidential democracies, like the United States, there is a much

(5)

towards the personalisation of politics (Heffernan &Webb, 2005; Mughan, 2000; Ohr, 2013). Mughan (2000) stated, in line with the findings above, that within parliamentary democracies, the focus shifts from the party to the party leader. Voters increasingly focus more on choosing between leaders than on choosing between parties, the so-called presidentialization of

parliamentary democracies. This confirms the increasing role of political leaders. A party’s image and support thus highly depend on their political candidates. In addition, a candidate’s image plays an important role in establishing the party’s image, which is built from citizens’ perceptions of the candidate (Garzia, 2013). Citizens’ impressions of candidates are based on different types of candidate information. Different cues like media performance and the record as a political leader, but also less conscious factors like physical characteristics, contribute to this information (Van Zuydam & Hendriks, 2015). De

Landtsheer and colleagues (2008) argue that citizens tend to base their political decisions on intuitive cues like language styles, non-verbal behaviour, and appearance characteristics. This means that these subconscious factors play an important role in voters’ decision-making process. The current study will focus on one of these subconscious cues in political candidates’ appearance, facial dominance, and aims to examine what role they play in candidate evaluations and the decision-making process in voting intentions. Moreover, the current study examines what other factors, like political orientation or personality, could possibly influence the effects.

1.1 Facial characteristics in leadership judgements

Facial characteristics are an example of intuitive cues and have been widely studied. Willis and Todorov (2006) showed that it takes individuals only 100 milliseconds to make trait judgements based on a person’s face. Facial characteristics are also an important factor in the

(6)

unknown to its audience, or little information is available about the candidate, citizens are likely to make assumptions on their leadership qualities based on the information that is available, such as appearance (Antonakis & Eubanks, 2017). This was confirmed by an experiment of Ballew and Todorov (2007) that showed how subjects were able to predict election outcomes by judging candidates’ faces on competence after 100 millisecond exposures.

During the 2004 U.S. Senate elections, Todorov et al. (2005) showed that inferences of competence based on facial characteristics were predictive for the outcome of the elections. They also showed how additional information on the candidate would slightly dilute but not completely overrule the effect of facial characteristics. This implies that voters are anchored in first impressions. Antonakis and Dalgas (2009) delved deeper into this implication, and suggested that voters still use the same heuristic path in the decision-making process as children do. They showed this with two experiments, one in which adults had to rate pairs of candidate’s faces on competence, and one in which children had to do a similar task with the same stimuli, but easier and more playfully explained. The results show that both children and adults were successful in predicting the election winner based on competence evaluation. Moreover, children’s ratings strongly predicted adults’ ratings. This confirmed assumptions that adults still use the same heuristics as children do when it comes to evaluating political candidates.

Ballew and Todorov (2007) examined this more extensively with three experiments, by showing participants pairs of political candidates’ faces (winner and runner up), and having the participants select the candidate that looked most competent in their opinion. The

(7)

picture for maximum of two seconds. These findings confirm that when leaders’ facial appearances are evaluated positively on competence, they are most likely to win elections. It also suggests competence to be a predictive characteristic for leadership abilities.

Besides competence, what features are considered important in a political leader? Mattes and colleagues (2010) examined how the characteristics deceitfulness, competence, attractiveness, and threatening appearances would fit political leaders. Participants were shown pictures of the faces of political candidates that ran against each other and made them judge them on the four traits. For each pair they had to indicate which politician would be most likely to have/show the characteristic. The results showed that competence,

attractiveness and threatening appearances were predictive for election outcomes. High attractiveness and threatening evaluations were predictive for losing candidates. Election winners were more likely to be rated highly on competence. This also confirms the important role of competence for political leader.

Another study that included multiple items to measure leadership evaluations was a study of Hayes (2005). He argued that parties and candidates often make use of trait ownership: By profiling a party on certain political issues and the way a party or candidate would enact on them, certain traits become associated with certain actions, and voters tend to associate these traits directly with party candidates. Hayes found that, on average, Republican candidates were rated higher on the characteristics ‘strong leader’ and ‘moral’, while

Democrats were rated higher on assessments of ‘compassion’ and ‘empathy’. No big

differences were found when judging Republican and Democrat candidates on their decency, intelligence, knowledge, or inspiration. Nonetheless, as both leaders scored high on these characteristics, they can still function as good leader indicators as well. These results suggest that it is important to measure leadership evaluations with multiple items. While Hayes (2005)

(8)

whether there are differences in how voters evaluate left- and right-leaning candidates on these items.

1.2 Facial dominance

Research has shown that certain facial features can be linked to perceived leadership qualities. But what kind of face makes for a good leader? Previous studies found that candidates with a more dominant appearance are more likely to be evaluated as a good leader (Spisak, Homan, Grabo & Van Vugt, 2012). The origin of facial characteristics and leadership abilities can be traced back to primates, where dominance plays an important role in hierarchy (Petersen, Dubuc & Higham, 2018). Petersen and colleagues (2008) mentioned three main facial aspects that contribute to dominance in non-human species: facial expressions, colour signals, and facial shape. He argues that a specific ratio between the width and the height of a face is an important indicator for alpha male status, assertiveness, and aggression. This same ratio is also applicable to humans and can be associated with testosterone levels (Lefevre et al., 2013), since high levels of testosterone cause aggression and status-striving behaviour. Pound, Penton-Voak and Surridge (2009) demonstrated that men with masculine facial characteristics have significantly higher testosterone levels when they experienced success, e.g. winning a competition. Watkins, Jones and DeBruine (2010) confirmed that dominance is often linked to facial masculinity. This facial masculinity can be found in features like thinner eyes, a stronger jaw, and notable brows (Pound et al., 2009).

Facial dominance in political leaders has been studied extensively. However, most studies examined facial preferences in different political contexts. For instance, Van Vugt and

(9)

2014; Little, Burriss, Jones & Roberts, 2007; Re et al., 2013). As these studies simulated political war/peace contexts, it is uncertain whether these could also be applied to the real world. Berinsky, Chatfield and Lenz (2019) tried to replicate the findings in real-world wartime and peacetime elections. Their findings suggested that candidates with dominant facial characteristics were preferred more often in both types of elections. In addition, one could question whether real-world elections can always be defined as wartime of peacetime. It might be more realistic to state that during elections, it is possible that certain topics evoke conflict whilst others evoke cohesion at the same time. Therefore, it would be interesting to examine facial dominance in current political times.

1.3 Political preferences

Research has shown that there is a relationship between facial characteristics and candidate evaluations, and also between evaluations and the political ideology of the candidate (Hayes 2005). This raises the question whether voters’ political preferences influence their candidate evaluations based on politicians’ faces as well. Few studies have focussed on this particular perspective.

Laustsen (2017) found a relation between political preferences and perceived political context. He argued that liberals perceive society as a safe place, defined by cooperation, whereas conservatives perceive society as threatening and dangerous. He states that it is likely that these differences in political preferences also lead to different preferences in candidate characteristics and leadership. According to Laustsen (2017), conservatives are likely to prefer strong leadership and powerful candidates, whereas liberals are likely to value warmth as a candidate characteristic. This could be in line with the findings mentioned earlier, where dominant faces are preferred in time of conflict and non-dominant faces in time of peace. In

(10)

this way, conservative voters’ preferences could be linked to more dominant faces and liberal voters’ preferences to less dominant faces.

Laustsen and Petersen (2016) tested this relationship with multiple studies and confirmed a relationship between voters’ political ideology and a preference for facial dominance. They showed how in the Danish context dominant faces are preferred by conservative voters, while liberals prefer non-dominant faces for their leaders. In contrast, liberal candidates were found to gain significantly less votes when their faces had dominant characteristics. This also shows a direct relationship between facial characteristics and voting behaviour (see also Laustsen, 2017; Laustsen & Petersen, 2018). Evidence for the differences was also provided by the authors, as they did one study on the degree in which the Danish politicians were recognized by subjects. When politicians were highly recognised, effects of the facial dominance disappeared. This makes it necessary to take candidates’ level of prominence into account as well.

While facial dominance has been studied extensively, its combination with political preferences is hardly investigated. Furthermore, most of the studies that focus on this relationship (e.g., Laustsen, 2017) were conducted in Denmark, and it is valuable to extend the research context to other countries as well. The current study will build upon these findings, aiming to make these more generally applicable amongst other northern European countries.

1.4 Personality traits

(11)

also play an important role (e.g., Bakker, Rooduijn & Schumacher, 2016). Research has agreed on the ‘Big Five’ personality traits as core characteristics that substantiate someone’s personality. These Big Five contain the traits extraversion, neuroticism, agreeableness, conscientiousness, and intellect or openness (Digman, 1990). Personality traits are relatively stable (Avery, Lester and Yang, 2015) and can predict political behaviour and political attitudes (Gerber, Huber, Doherty, Dowling, & Ha, 2010). Bakker and De Vreese (2016) confirmed that each personality trait can be predictive for different aspects of political behaviour. Bakker et al. (2016) showed that low levels of agreeableness lead to a more populistic and extreme political ideology, either left-leaning or right-leaning. People with high levels of agreeableness were more likely to vote for the establishment.

Other personality traits can be linked directly to the left-to-right political spectrum. Furnham and Fenton-O’Greevy (2018) showed that openness positively predicts left-wing political orientation, explaining that openness is often associated with curiosity and being open for change. Conscientiousness was shown to be predictive for right-wing political orientation, since conscientiousness is often associated with conservatism. Avery et al. (2015) confirmed these findings: They showed that openness and conscientiousness had an indirect effect on Obama’s vote share, where higher levels of conscientiousness led to lower support and higher levels of openness led to more support.

Overall, several studies have confirmed that the personality trait ‘openness’ can predict political liberalism and preferences for left-wing parties, whereas ‘conscientiousness’ can predict conservatism and preferences for right-wing parties (Schoen & Schumann, 2007; Vecchione, Schoen, Castro, Cieciuch, Pavlopoulos & Caprara, 2011; Xu, Xiaowen &

(12)

Since research has shown that personality traits can predict political preferences, and political preferences can predict preferences in facial dominance and candidate evaluations, it would be interesting to examine whether personality traits can also directly affect preferences in facial dominance above and beyond the influence of ideological preference.

While many of the aspects mentioned in this theoretical framework have been addressed in previous studies, a joined investigation on their influence on candidate

evaluations and voting preferences is still missing. The current study will aim at replicating previous studies and embed their findings into a larger research design. All assumptions are visualised in a conceptual framework (Figure 1).

By adding multiple items of candidate evaluations, the current study combines findings of Laustsen and Petersen (2016; 2018; see also Laustsen 2017) and Antonakis and Dalgas (2009), Ballew and Todorov (2007) and Todorov et al. (2005), which makes it possible to detect what aspects substantiate facial preferences. This not only adds value to already existing findings, but also extends these. The current study will extend previous findings and elaborate on these by an overarching framework, aiming to provide a clear overview of all connecting pathways.

(13)

Figure 1. Conceptual framework

The study’s general research question asks:

To what extend do the personality traits openness and conscientiousness predict preferences for dominant or non-dominant facial characteristics, and how does this preference relate to political preference, candidate evaluations and voting intention?

Based on this research question, the following hypotheses are formulated:

H1a: Participants with a high level of the personality trait openness will be significantly more left-wing

H1b: Participants with a high level of conscientiousness will be significantly more right-wing.

(14)

H2a: Participants with a high level of the personality trait openness will have significantly more preference for non-dominant faces

H2b: Participants with a high level of conscientiousness will have significantly more preference for dominant faces.

H3: Participants with a high level of the personality trait openness will evaluate the candidates significantly different than participants with a high level of

conscientiousness.

H4a: Participants with more left-wing political preferences will evaluate non-dominant faces more positively than dominant faces.

H4b: Participants with more right-wing political preferences will evaluate dominant faces more positively than non-dominant faces.

H5a: Dominant faces will be evaluated more highly on the traits strong leader and moral H5b: Non-dominant faces will be evaluated more highly on the traits compassionate and

empathetic.

H6: Participants that give positive evaluations to a political candidate are more likely to have significantly higher voting intentions than participants that give negative evaluations.

(15)

2. Method

2.1 Design

The current study aims to answer the research question and hypotheses through an online experiment with one factor (facial dominance). The other independent variable (personality traits) cannot be manipulated as it captures respondents’ individual differences.

The manipulated factor resulted in two groups; participants were randomly assigned to see either a dominant or non-dominant face (between-subjects factor). Before exposure to the experimental manipulation, respondents were tested on two personality traits: openness and conscientiousness. These were taken into consideration as a second independent variable and treated as a moderator in the analyses.

2.2 Pre-test

In order to select appropriate stimuli, a pre-test was conducted. To do so, existing stimuli were taken from a study of Laustsen and Petersen (2016; provided by the authors). The original stimuli consisted of three pictures of three Danish politicians each. They showed the original picture as well as two morphed pictures. The pictures were morphed with a software called Psychomorph, and were changed approximately 40% along a dominance dimension. One was morphed into a more dominant (+40) version and one into a less dominant (-40) version. This dominance dimension was created by Laustsen and Petersen themselves. In total, six pictures were obtained. The goal of the pre-test was to see whether the manipulation would also be successful amongst a Dutch sample. Since the main experiment only included one politician, the most successful manipulation from the pre-test would be used in the actual experiment. In addition, the pre-test was used to check whether the politicians were

(16)

picture were shown, and the original pictures were disregarded. All stimuli are shown in Figure 1 in the appendix.

The pre-test was distributed through social media. In total, 22 people completed the pre-test. Participants were aged 21-57, and 9 of them were male (40.9%). The majority of the participants in the pre-test were highly educated (72.7%).

The pre-test showed three Danish politicians: Morten Bødskov, Ole Hækkerup, and Troels Ravn. Participants were shown three pictures (either the dominant or non-dominant version), one of each politician. Both the version and the order of the politicians were randomised between or within respondents, respectively. Participants were informed about the goal of the study and gave their consent in advance and were debriefed about the manipulation

afterwards. Before showing the pictures, participants were informed that the people in the pictures were political candidates. The party or ideology of the political candidates were not mentioned in the pre-test, and participants were asked to answer the questions solely based on their perception of the picture they were shown. After seeing each photo, participants were asked to rate the level of dominance of that political candidate on a scale of 1 to 100. This was done for all three photos separately. At the end of the survey, participants were asked whether they recognised one of the politicians and, if this was the case, which one.

The majority of the participants did not recognise any of the politicians. Two participants recognised Morten Bødskov, and one of these also recognised Ole Hækkerup. After an informal conversation with these participants, it turned out they both thought these were Dutch politicians. Morten Bødskov was compared to the Dutch politician Alexander Pechtold. This implied that one of the other two politicians might be a better choice for the main

(17)

Ole Hækkerup (t (20) = 2.17, p = .043). Participants that were shown the dominant version (M = 52.90, SD = 23.31) evaluated this picture as significantly more dominant than participants that were shown the non-dominant version (M = 30.33, SD = 25.14). There was no significant difference between the respective dominant and non-dominant pictures of Morten Bødskov (p = .566) or Troels Ravn (p = .440). The descriptives of the pre-test can be found in Table 1 in the appendix.

The results of the pretest indicate that the picture of Ole Hækkerup would be most suitable to use as stimulus material in the main experiment, and confirm the successful manipulation of facial dominance for Hækkerup.

2.3 Main study

2.3.1 Procedure

The main experiment was conducted online, and the link to the online survey was distributed among Dutch citizens through Facebook and WhatsApp. All subjects participated voluntarily, and no incentives were offered. All subjects gave their consent in the beginning of the survey and were debriefed about the manipulations and focus of the study at the end. To ensure comparability among participants, the introduction texts that were shown prior to the stimuli exposure were adapted to the participants’ party ideology. Subjects that did not vote for any of the major Dutch parties were led to the end of the survey and excluded from participation. In order to create a realistic environment, the introductory texted referred to the upcoming European Elections (May 2019). As the European Elections are a second order election in the Netherlands (Lefevere & Van Aelst, 2014), it was deemed likely that participants have little knowledge about the elections’ candidates and might therefore be more willing to assume actual existence of the politician in the stimuli. In line with existing theories, this could also

(18)

contribute to the heuristic path of decision-making, having participants depend more on intuitive cues such as facial appearance.

2.3.2 Data cleaning and subjects

All responses with missing data were deleted. Initially, a total of 242 participants completed the online survey. The duration of the survey, in seconds (M = 788.33, SD = 3567.56), showed some unreliable recordings and the presence of outliers. To optimize the validity of the responses, outliers that took more than one hour or less than one minute to complete the questionnaire were deleted. This resulted in a total of 234 participants that were included in the analyses.

122 of the participants (52.1%) were male and 111 were female (47.4%). One participant (0.4%) did not identify with either one of the gender options. The age of the participants varied from 18 to 83 years old (M = 30.68; SD = 13.82). Most participants completed an academic (n = 78; 33.9%) or higher education (n = 78; 33.3%). Others completed a vocational education (n = 34; 14.5%). All other educational levels were completed by less than 10

percent of the participants.

Most participants indicated to be politically interested. This was measured on a 7-point scale (1= Not interested at all; 7= Very interested). 53.4% of the participants answered 5 or more (M = 4.46, SD = 1.66). The sample also showed clear preferences for certain parties. 61 of the participants (26.1%) voted for D66, 54 participants (23.1%) voted for GroenLinks and 47 participants (20.1%) voted for VVD. The other parties were represented each by less than 10 percent of the participants.

(19)

2.3.3 Materials

Independent variables

Facial dominance relates to the manipulated conditions and is the independent variable of this study. Two conditions were made out of a picture of the Danish politician Ole Hækkerup. Personality traits was included as a second independent variable in the current study. Respondents’ personality traits were measured using the Dutch mini IPIP (Bakker & De Vreese, 2016), with four items measuring openness and conscientiousness, respectively. The items that were included can be found in Table 2 in the appendix. All items were shown in randomized order. Participants were to indicate to what extend the statement reflected their own personality (1 = Very incorrect, 7 = Very correct).

Even though the mini IPIP is an existing scale that has been proven to be successful, a factor analysis was done to confirm this. A PCA-factor analysis with varimax rotation was

computed for the items “fantasy” (1), “interest in abstract ideas” (2), “difficulty with abstract ideas” (3), “imagination” (4), “chores” (5), “order” (6), “messy” (7), “forget to put things back” (8). The eigenvalue and screen plot confirm that three components can be made out of these eight items. No items were disregarded due to a minimum factor load below? .40. The first component consists of item 5, 6, 7 and 8, and explains 29.18% of the variance. The second component consists of item 2 and 3 and explains 17.17% of the variance and the third component, consisting of item 1 and 4 explains 14.14%. These findings are partly in line with the mini IPIP scale. The first component contains all four items that were intended to measure the personality trait conscientiousness. The second component contains two items of the personality trait openness that focus on the participant’s interest and understanding of abstract ideas. The third component contains two items of the trait openness that focus on the

(20)

The reliability of the variable conscientiousness (M = 5.06, SD =.87), comprising four items, was acceptable: α = .75.

Due to the results of the factor analyses and the scale reliability of the two components

(component 2: α = .20; component 3: α = .49; overall reliability both openness components: α = .31), the trait openness could not be reliably measured and therefore had to be excluded from further analyses. This means that for answering the hypotheses, solely conscientiousness was taken into account.

Participants were asked what party they voted for during the last national elections, or, in case they had not voted, what party they would have most likely voted for. This was measured through one single multiple-choice question, containing the major Dutch parties. People that voted for different parties were excluded from the experiment. The reason to do this was to match the condition to their political ideology. Where previous studies found that political ideology goes beyond personality (Bakker et al., 2016), it was important to align the conditions with participants’ ideology to create a more realistic situation, provide the same conditions for all participants, and exclude differences in political biases.

Political preference was measured by a single item asking, on a ten-point scale (1 = very left, 10 = very right, M = 5.01, SD = 1.99) where participants would place themselves.

Participants’ political interest was measured by a single seven-point scale on which

(21)

Dependent variables

Candidate evaluation was measured by using an existing 7-point scale, containing eight items (ANES-scale, as stated in Hayes, 2005). These items were freely translated and both versions can be found in Table 3 in the appendix. As it was not clear if any components existed, an exploratory factor analysis was done.

A PCA-factor analysis with varimax rotation showed that two factors form candidate

evaluations. The scree plot confirms this. No items were disregarded due to a minimum factor load below? .40. The first component consists of the items “leader”, “moral”,

“compassionate”, “empathetic”, and “inspiring”, and explains 53.98% of the variance. The second component consists of the items “decent”, “intelligent” and “knowledgeable” and explains 15% of the variance. The reliability of the first component, computed as variable candidate evaluation personality, comprised of the first five items, was good (α = .851); the reliability of the second component, computed as variable candidate evaluation competence (three items), was good as well (α = .832).

Voting intention was measured by a single question in which participants had to indicate how likely they were to vote for the candidate during the European Elections in May 2019. A 100-point slider scale was used (M = 41.79, SD = 27.44).

2.3.4 Randomization checks

In order to check whether participants were successfully randomized between the

experimental conditions, multiple analyses were computed. A one-way analysis of variance showed no significant differences between the conditions in terms of respondents’ age (non-dominant: M = 30.30; (non-dominant: M = 31.06, (F (1, 23) = .18, p = .675). The results of a Chi-square test showed no significant relation between condition and gender (Χ2 (2) = 1.04, p =

(22)

non-dominant and dominant condition with regard to respondents’ personality trait conscientiousness (t (232) = .42, p = .203). In addition, a chi-square test showed no

significant relationship between the two conditions and political ideology/past vote (Χ2 (10) =

11.432, p = .325).

Independent samples t-tests showed no significant differences between the non-dominant and dominant condition regarding political preference (t (232) = .066, p = .948) or political interest (t (232) = 1.03, p = .306). Since participants were successfully randomized on all relevant (control) variables, the main analyses did not take into them account.

2.3.5 Statistical treatment

The study will make use of a mediation analysis and regression analyses.

3. Results

As the reliability test on the component openness was unacceptable, H1a, H2a and H3 could not be tested.

H1b assumed that participants with a high level of conscientiousness will be significantly more right-leaning. To test whether a relationship exists between conscientiousness and political preferences, a correlation analysis was done. While people with higher levels of conscientiousness are more likely to have right-wing preferences (r (234) = .26, p = .074), this relationship is not significant. Therefore, H1b was rejected.

(23)

conscientiousness. The regression analysis showed that the interaction variable was not predictive for the two candidate evaluation scales personality (β = -.375, p = .320) and competence (β = -.406, p = .284). When people with low level of conscientiousness were shown the non-dominant condition, they could have been more likely to evaluate them more negatively (see also figure 2 and 3 in the appendix). Yet again, this relationship was not significant.

Preferences cannot only be measured by evaluations but also by voting intention. Therefore, a similar regression analysis was done on voting intention. This showed no significant predictive effect of the interaction variable on voting intention (β = -.099, p = .793; see Figure 4 in the appendix). Concluding, H2b was rejected.

H4a and H4b expected left-leaning participants to evaluate the non-dominant stimulus more positively and right-leaning participants to evaluate the dominant stimulus more positively. In order to test whether a relationship between political preferences and candidate evaluation exists, an interaction variable was computed for condition and political preference. A regression analysis showed that this interaction variable is no significant predictor when evaluating candidates’ personality (β = .09, p = .744) or competence (β = -.411, p = .117). When people have right-wing preferences and were shown a dominant stimulus, they were more likely to evaluate the politician positively on personality and negatively on competence items (See also Figure 3 below and 5 in the appendix). However, again these differences are not significant, and H4a and H4b are rejected.

(24)

Figure 3. Interaction plot of political preference, facial dominance and personality evaluation

*Political preference was split into 3 categories (10-point scale: 1 – 3 = left, n = 37; 4-7 = centred, n = 99; 8-10 = right, n = 52).

H5a and H5b assumed that dominant faces will be evaluated more highly on the traits ‘strong leader’ and ‘moral’, whereas non-dominant faces will be evaluated more highly on the traits ‘compassionate’ and ‘empathetic’. The latter three items are all in the personality component of the candidate evaluations. Leader is the only item that belongs to the competence

component. Independent samples t-tests were conducted for the four items of candidate evaluation (leader, moral, compassionate and empathetic) and facial dominance. These were tested for each item separately, as previous research showed that individual items of the scale

(25)

2.70, p = .007). Participants that were shown the dominant stimulus (M = 4.70, SD = 1.16) evaluated this higher on leadership than participants that were shown the non-dominant version (M = 4.28, SD = 1.20).

T-test on the items moral (t (232) = .06, p = .954), compassionate (t (232) = 1.33, p = .185) and empathetic (t (232) = 1.36, p = .176) with regard to facial dominance all showed no significant differences between the two conditions. Therefore, H5a was only partly confirmed and H5b was rejected.

To test the path of H5 as a whole, a regression analysis was done with facial dominance and both personality and competence candidate evaluations. The regression analysis showed no significant predictive qualities of facial dominance on both personality (β = .02, p = .778) or competence (β = -.011, p = .861) evaluations.

H6 assumed that participants that evaluate a political candidate positively would be more likely to turn out to vote in the EP 2019 election than participants who evaluated the candidate less positively. In order to test this, a regression analysis was done on the variables candidate evaluation and voting intention. It showed that candidate evaluations regarding personality explained 20.4% of the variance in voting intention (F (1, 232) = 60.60, p < .001) is a significant predictor for voting intention (β = .46, p < .001). Candidate evaluation regarding competence is also shown to be a significant predictor for voting intention (β = .44, p < .001) and explains 18.8% of the variance: The higher candidates are evaluated on leadership qualities, the more likely participants are to vote for them. Therefore, H6 was confirmed.

So far, we’ve tested the following direct lines of the conceptual model, shown in Figure 5 below.

(26)

Figure 5. The hypotheses implemented in the conceptual framework

In order to be able to test the complete model, the remaining direct and indirect paths will be tested as well.

An independent samples t-test showed no significant differences between the two conditions regarding voting intention (t (232 = 1.28, p = .203). Therefore, no direct effect of facial dominance on voting intention can be confirmed.

Regression analyses showed no significant predictive effects of political preferences for candidates’ personality evaluation (β = .10, p = .145) or competence (β = .13, p = .051). The latter was, however, very close to being significant. The positive coefficients imply that participants with right-leaning preferences tend to evaluate candidates more positively,

(27)

A regression analysis showed that the interaction variable of condition and political

preference was no significant predictor for voting intention (β = -.025, p = .924): People that have left-wing preferences and were shown a dominant stimulus were morelikely to vote for the politician, but across all levels of political preference, respondents who saw the dominant face were more likely to vote for the candidate. See also Figure 6.

Figure 6. Interaction plot of political preference, facial dominance and voting intention

Even though not all separate paths were confirmed by the analyses, it would still be interesting to test the model as a whole. As it is impossible to include every path in one analysis, the basis of the model can be traced back into a model 8 mediation analyses

(PROCESS, Hayes, 2017) consisting of a mediation analysis with a moderating variable. The variable personality traits will be excluded from the analysis.

(28)

The mediation analysis showed no indirect effect of facial dominance on voting intention through candidate evaluation:

Competence:

Political preference 3 (left-leaning): E=1.31, 95% CI (-1.43; 4.44). Political preference 7 (right-leaning): E=-1.64, 95% CI (-5.39; .87).

Personality:

Political preference 3 (left-leaning): E=.65, 95% CI (-2.54; 63.97). Political preference 7 (right-leaning): E=-.05, 95% CI (-3.91; 3.35).

The mediation analysis confirmed earlier findings and showed no moderation effect of political preference on both competence (Index = -.74, 95% CI (-2.14; .21) and personality evaluations (Index = -.18, 95% CI (-1.63; 1.10) as mediator. No interaction effects were found for political preference and facial dominance on candidate evaluations (personality: b = -.02, 95% CI (-.15; ,11), t = -.33, p = .745; competence: b = -.09, 95% CI (-.20; .02), t = -1.57, p = .117) and voting intention (b = .74, 95% CI (-2.37; -3.86), t = .47, p = .639).

4. Discussion

The aim of the current study was to examine whether personality traits and political preferences could predict preferences in facial dominance with regard to candidate

(29)

on candidate evaluations and voting intention. The main goal was to see how personality traits and political preferences can influence the effect of facial dominance on leadership

evaluations in political candidates and voting intention.

Facial dominance was found to have no direct effect on candidate evaluations (H5) and voting intention. Regarding the separate items of candidate evaluations, it was shown that the

dominant stimulus was evaluated significantly higher on the item ‘good leader’. This is in line with findings of Spisak et al. (2012) and implies that dominant facial characteristics are more likely to be evaluated as leader and associated with leadership. No further significant

differences in evaluation items were found. However, as this is hardly examined before, it would still be valuable to do further research to these items. Since Hayes’ (2005) study included very large U.S. datasets, with thousands of respondents, it would be valuable to test this again amongst a similarly sized sample from a parliamentary democracy.

Candidate evaluations were shown to be a significant predictor for voting intention (H6). As previous results showed the relationship between evaluations and election outcomes (Todorov et al., 2005), this was in line with the expectations. When participants evaluated the candidate positively, they were more likely to vote for the politician.

Political preferences and facial dominance did not have an interaction effect on candidate evaluations (H4) and voting intention. The direction of the results showed

interesting findings, as the direction in H4a was not in line with previous research (Laustsen & Petersen, 2016). It was shown that left-leaning participants evaluated the dominant stimulus more positively than the non-dominant stimulus. Results of H4b showed the expected direction as the dominant stimulus was evaluated more positively by right-leaning participants.

(30)

Right-leaning participants evaluated the stimuli very differently on personality evaluations and very similarly on competence evaluations. For competence evaluations, this direct effect of political preference was almost significant. Regarding left-leaning participants, the differences between evaluations of the stimuli were almost unnoticeable. These results indicate that facial characteristics can be evaluated differently on the separate items of leadership ability. As these findings show mixed results, it would be interesting do examine this more extensively in future research.

Political preferences were predictive for voting intention. Right-leaning participants were more likely to vote for the politician, regardless of what version they were shown. An explanation might lay in differences in voting turnout amongst right- and left-leaning

participants. To investigate this, future research should focus on overall voting turnout instead of voting intention for one politician.

The current study expected a relation between political preferences and personality traits (H1). Not all hypotheses concerning personality traits could be tested, as the personality trait ‘openness’ could not be reliably measured. As the mini IPIP scale has been proven to be successful before, this was quite surprising. Some participants contacted the researcher with questions about the items that included the term abstract. As two of four items of the

component openness focussed on abstract ideas, wrong interpretation of these items might explain the failed reliability test. When future studies examine only one or two personality traits, it is recommended to measure these with the original IPIP scale, which contains 10 items per trait. If some items are interpreted incorrectly, chances are bigger that the scale

(31)

The relation between political preferences and conscientiousness was close to being

significant, implying that participants with high conscientiousness were likely to have right-wing preferences. As results were insignificant, this was not in line with previous research, however, the assumed direction of the relationship was (e.g., Avery et al., 2015; Furnham & Fenton-O’Greevy, 2018).

The study tested a relationship between personality traits and preferences in facial dominance, measured by candidate evaluations and voting intention (H2). The direction of the – insignificant – findings was rather unexpected, as it showed that subjects with low

conscientiousness that were shown the non-dominant stimulus evaluated these more

negatively. As previous research has shown that low conscientiousness is linked to left-wing political orientation (e.g., Gerber et al., 2010 Schoen & Schumann, 2007) and left-wing orientation is linked to non-dominant preferences (Laustsen, 2017; Laustsen & Petersen, 2016; 2018), it is interesting that a predictive relation of conscientiousness on facial

preferences implied the opposite. One reason for this unexpected result may be found in this study’s sample, which, on average, scored high on conscientiousness, causing skewed results (see also Figure 2 in the appendix). A study of Allik and McCrea (2004) found that

personality traits can be related to cultural and geographical differences. The Dutch score high on conscientiousness, explaining the skewed sample. Further research could prevent this by using a sample which is either equally distributed amongst the trait, or which allows you to compare groups (e.g. participants from different countries) that are strongly related to certain traits.

An overall limitation of the study concerns the stimulus materials. As the current study used existing stimuli that had only been examined with a Danish sample, this could have

(32)

could still be differences in how Dutch would evaluate the stimuli compared to Danish subjects. Therefore, it is would be interesting to replicate this study with both Danish and Dutch participants, in order to see if any differences occur.

Most participants in this study’s sample indicated to be politically interested. As the questionnaire was framed on the upcoming European elections, this could have influenced the effect of the manipulation. Some participants indicated that they already knew the running candidates of their favourable party and therefore knew the stimulus was not real. This shows that the EU election frame and the matching introduction text could have influenced the manipulation.

The current study’s sample was unequally divided amongst the political parties. The majority of the sample voted for liberal or left-wing parties. Regarding political preferences, the sample was much more equally divided. This shows how parliamentary democracies can be quite complicated. An example of a party that is contradictory to many findings is the VVD, which is associated with liberalism and right-wing political orientation (Van Kersbergen & Krouwel, 2008). This makes it more complicated to interpret results.

An additional recommendation concerns the scale that was used to measure candidate evaluations. The current study used an existing ANES scale, but it would be good to develop a new scale especially for the multiple aspects of good leadership. Contrary to what most studies state about the prominent role of competence as a leadership trait (Antonakis & Dalgas, 2009; Ballew & Todorov, 2007; Todorov et al., 2005), there are studies that argue that ‘warmth’ is a more important trait than competence (Laustsen & Bor, 2017). This shows that it is necessary to examine specific leadership traits more extensively.

(33)

findings, however, the first step has been taken towards an overarching framework. The study initiated the beginning movement towards a better understanding of underlying relationships between existing research in this area.

(34)

References

Aarts, K., Blais, A., Schmitt, H. & European Consortium for Political Research (2011). Political leaders and democratic elections (Comparative politics). Oxford: Oxford University Press.

Allik, J. & McCrae, R. R. (2004). Toward a geography of personality traits: Patterns of profiles across 36 cultures. Journal of Cross-Cultural Psychology, 35(1), pp. 13–28. Antonakis J. & Dalgas O. (2009). Predicting elections: Child’s play! Journal of Science, 323,

pp. 1183.

Antonakis, J. & Eubanks, D. L. (2017). Looking leadership in the face. Current Directions in Psychological Science, 26 (3), pp. 270–275.

Avery, J., Lester, D. & Yang, B. (2015). State-level personality and presidential vote share in the 2008 and 2012 presidential elections. The Social Science Journal, 52(2), pp. 112– 122.

Bakker, B. & De Vreese, C. (2016). Personality and European Union attitudes: Relationships across European Union attitude dimensions. Journal of European Union Politics, 17(1), pp. 25–45.

Bakker, B.N., Rooduijn, M. & Schumacher, G. (2016). The psychological roots of populist voting: Evidence from the United States, the Netherlands and Germany. European Journal of Political Research, 55(2), pp. 302–320.

Ballew C.C. & Todorov A. (2007). Predicting political elections from rapid and unreflective face judgments. Proceedings of the National Academy of Sciences, 104 (46), pp. 17948-17953. DOI: 10.1073/pnas.0705435104

(35)

different risks of exploitation shape preferences for leader facial dominance. The Leadership Quarterly 27(6), pp. 820-37.

De Landtsheer, C., De Vries, P. & Vertessen, D. (2008). Political impression management: How metaphors, sound bites, appearance effectiveness, and personality traits can win elections. Journal of Political Marketing, 7(3–4), pp. 217–238.

Digman, J.M. (1990). Personality structure: Emergence of the five-factor model. Annual Review of Psychology 41(1), pp. 417–440.

Furnham, A. & Fenton-O’Creevy, M. (2018). Personality and political orientation. Personality and Individual Differences, 129, pp. 88–91.

Garzia, D. (2013). Can candidates' image win elections? A counterfactual

assessment of leader effects in the Second Italian Republic. Journal of Political Marketing, 12(4), pp. 348-361. DOI: 10.1080/15377857.2013.837303

Gerber, A., Huber, G., Doherty, D., Dowling, C. & Ha, S. (2010). Personality and political attitudes: Relationships across issue domains and political contexts. American Political Science Review, 104(1), pp. 111-133. DOI:10.1017/S0003055410000031 Hayes, D. (2005). Candidate qualities through a partisan lens: A theory of trait ownership. American Journal of Political Science, 49(4), pp. 908-923. DOI:10.2307/3647705 Heffernan, R. & Webb, P. (2005). The British Prime Minister: Much more than “First among

Equals”, in Poguntke, T. & Webb, P. (eds), The Presidentialization of Politics: A comparative study of modern democracies. Oxford: Oxford University Press, pp. 26– 62.

Laustsen, L. (2017). Choosing the right candidate: Observational and experimental evidence that conservatives and liberals prefer powerful and warm candidate personalities, respectively. Journal of Political Behavior 39(4), pp. 883-908.

(36)

evaluations: Warmth is more important than competence, leadership and integrity. Electoral Studies, 49, pp. 96–107

Laustsen, L. & Petersen, M.B. (2016). Winning faces vary by ideology: How nonverbal source cues influence election and communication success in politics. Journal of Political Communication 33(2), pp. 188-211.

Laustsen, L. & Petersen, M. (2018). When the party decides: The effects of facial competence and dominance on internal nominations of political candidates.

Evolutionary Psychology: An International Journal of Evolutionary Approaches to Psychology and Behaviour, 16(2). DOI:10.1177/1474704917732005

Lefevre, C. E., Lewis, G. J., Perrett, D. I. & Penke, L. (2013). Telling facial metrics: Facial width is associated with testosterone levels in men. Journal of Evolution and Human Behavior, 34(4), pp. 273–279.

Little, A.C. (2014). Facial appearance and leader choice in different contexts: Evidence for task contingent selection based on implicit and learned face-behaviour/face-ability associations. The Leadership Quarterly 25(5), pp. 865-74.

Little, A., Burriss, R., Jones, B. & Roberts, S. (2007). Facial appearance affects voting decisions. Journal of Evolution and Human Behaviour, 28(1), pp. 18-27. DOI:10.1016/j.evolhumbehav.2006.09.002

Mattes, K., Spezio, M., Kim, H., Todorov, A., Adolphs, R. & Alvarez, R. (2010). Predicting election outcomes from positive and negative trait assessments of candidate images. Journal of Political Psychology, 31(1), pp. 41-58.

(37)

Media and the presidentialization of parliamentary elections, pp. 22-51. American History in Depth. London, UK: Palgrave Macmillan.

Ohr, D. (2013). Changing patterns in political communication. In Aarts, K., Blais, A. & Schmitt, H. (Eds.), Political leaders and democratic elections. Oxford, UK: Oxford Scholarship Online.

Petersen, R.M., Dubuc, C. & Higham, P.J. (2018). Facial displays of dominance in non human primates. In Senior, C. (Ed.), The facial displays of leaders, pp. 123-143. Birmingham, UK: Palgrave Macmillan, Cham.

Pound, N., Penton-Voak, I. S. & Surridge, A. K. (2009). Testosterone responses to

competition in men are related to facial masculinity. Proceedings of the Royal Society B: Biological Sciences, 276, pp. 153-159.

Re, D. E., DeBruine, L. M., Jones, B. C. & Perrett, D. I. (2013). Facial cues to perceived height influence leadership choices in simulated war and peace contexts. Journal of Evolutionary Psychology, 11(1), pp. 89-103.

Schoen, H., & Schumann, S. (2007). Personality traits, partisan attitudes, and voting

behaviour. Evidence from Germany. Journal of Political Psychology, 28(4), pp. 471-498.

Spisak B. R., Homan A. C., Grabo A. & Van Vugt M (2012). Facing the situation: Testing a biosocial contingency model of leadership in intergroup relations using masculine and feminine faces. The Leadership Quarterly, 23(2), pp. 273–280.

Strömbäck, J. (2008). Four phases of mediatization: An analysis of the mediatization of politics. The International Journal of Press/Politics 13(3), pp. 228-46

Todorov A., Mandisodza A. N., Goren A. & Hall C. C. (2005). Inferences of competence from faces predict election outcomes. Journal of Science, 308, pp. 1623–1626.

(38)

and the 'foreigners issue'. Journal of European Public Policy, 15(3), pp. 398-414. Van Vugt, M. & Grabo, A. E. (2015). The many faces of leadership: An evolutionary

psychology approach. Current Directions in Psychological Science, 24(6), pp. 484– 489. DOI:10.1177/0963721415601971

Van Zuydam, S. & Hendriks, F. (2018). Credibility enacted: Understanding the meaning of credible political leadership in the Dutch parliamentary election campaign of 2010. Journal of Political Marketing, 17(3), pp. 258-281.

DOI:10.1080/15377857.2015.1039747

Vecchione, M., Schoen, H., Castro, J.L.G., Cieciuch, J. Pavlopoulos, V. & Caprara J.V. (2011). Personality correlates of party preference: The Big Five in five big European countries. Journal of Personality and Individual Differences 51(6), pp. 737-42. Watkins, C. D., Jones, B. C. & DeBruine, L. M. (2010). Individual differences in dominance

perception: Dominant men are less sensitive to facial cues of male dominance. Journal of Personality and Individual Differences, 49(8), pp. 967-971.

Willis J., & Todorov, A. (2006). First impressions: Making up your mind after a 100-ms exposure to a face. Journal of Psychological Science, 17, pp. 592–598.

Xu, X., & Peterson, J. B. (2017). "Differences in media preference mediate the

link between personality and political orientation." Journal of Political Psychology, 38(1), pp. 55-72.

(39)

Appendix

Figure 1. Original facial stimuli from Laustsen and Petersen (2016)

Non-dominant Dominant

Morten Bødskov

Ole Hækkerup

(40)

Table 1. Descriptives and t-values of pre-test materials

M SD N t p

Morten Bødskov (1): Dominant 35.71 13.88 14 .58 .566 Non-dominant 32.13 13.90 8

Ole Hækkerup (2): Dominant 52.90 23.31 10 2.17 .043 Non-dominant 30.33 25.14 12

Troels Ravn (3): Dominant 42.55 12.12 11 .79 .440 Non-dominant 36.90 20.42 11

(41)

Table 2. Item wording mini IPIP, and Dutch translation

Personality trait Items

Openness: English 1. I have a vivid imagination.

2. I am not interested in abstract ideas. *

3. I have difficulty understanding abstract ideas. * 4. I do not have a good imagination. *

Dutch 1. Ik heb een levendige fantasie.

2. Ik ben niet geïnteresseerd in abstracte ideeën. * 3. Ik heb moeite om abstracte ideeën te begrijpen. * 4. Ik heb geen goede verbeelding. *

Conscientiousness: English 5. I get chores done right away. 6. I Like order.

7. I make a mess of things. *

8. I often forget to put things back in their proper place. *

Dutch 1. Ik doe karweitjes meteen. 2. Ik houd van orde.

3. Ik maak een puinhoop van dingen. *

4. Ik vergeet vaak om dingen op de juiste plaats terug te zetten. *

(42)

Table 3. Item wording ANES-scale (Hayes, 2015), and the translation

Item Translated item

Strong leader Sterke leider

Moral Moreel Compassionate Meelevend Empathetic* Empathisch Decent Geschikt Intelligent Intelligent Knowledgeable Deskundig Inspiring Inspirerend

*) Item was measured with the statement ‘cares for people like me’, translated as ‘geeft om mensen zoals ik’.

(43)

Figure 2. Interaction plot for conscientiousness, facial dominance and personality evaluations

*Conscientiousness was split into 3 categories (7-point index: 1 – 2.33 = low, n = 6; 2.34 – 4.66 = medium, n = 67; 4.67 – 7 = high, n = 161).

** It must be noted that the group sizes differ much. Therefore, the figure is helpful for interpretation of the results, but not reliable.

(44)

Figure 3. Interaction plot for conscientiousness, facial dominance and competence evaluations

(45)

Figure 4. Interaction plot for conscientiousness, facial dominance and voting intention

(46)

Figure 5. Interaction plot of political preference, facial dominance and competence evaluation

Facing Political Leaders : a study on how personality traits and political preferences can influence the effect of facial dominance on leadership evaluations in political candidates