Facial traits and appearance-based inferences about political candidates are found to predict election outcomes above chance in various countries with different electoral systems and institutions. Research over the last decade demonstrates two different versions of this finding. First and most thoroughly studied, candidates who from their mere faces are evaluated as more competent get more votes on Election Day. Second, recent research finds that the ideological leanings of candidates and the voters they cater to also matter: Rightwing and conservative candidates receive more votes if they look more dominant, while liberal candidates lose votes when looking dominant and masculine. In this article, we investigate if these patterns extend to candidate selection and support within parties as determined by party organizations. We test this through an original combination of naïve respondents’ trait ratings of candidates for Danish local elections and these candidates’ positions on the ballot as decided by nomination processes within local party organizations. The results strongly support that the conclusions in previous studies extend to dynamics within the party among party members: Danish local party organizations tend to nominate facially competent candidates at the top of the ballot regardless of their ideological leaning. Moreover, liberal and conservative parties position dominant looking candidates significantly different on the ballot with liberal parties being less likely to assign facially dominant candidates to top ballot positions. These results add important new insights about the underlying psychological processes causing appearance-based voting and relates to the ongoing discussion about the quality of public opinion formation.
A number of recent studies have shown that individuals’ preferences for politicians are influenced by the physical features of the politician such as their facial appearance or the pitch of their voice. In actual electoral races, for example, candidates with facial and vocal traits that are perceived as “competent” gain more voters (e.g. Todorov et al., 2005; Olivola & Todorov, 2010; Berggren Jordahl & Poutvaara, 2010; Laustsen 2014; Klofstad, 2016). Also facial and vocal traits that make a person appear “dominant” have been found to influence election results but in more complex ways: Conservative candidates gain more votes when they look or sound dominant, while liberal candidates in contrast lose votes when evoking such dominant impressions (Laustsen & Petersen 2015, 2016; Laustsen, Klofstad & Petersen 2015).
Political scientists have viewed these effects against the background of democratic ideals. These ideals emphasize how citizens should vote for a candidate that best represents their preferred set of policies. In this regard, the physical appearance of the candidate arguably ought to play no role. In light of these ideals, political scientists have argued that voters’ reliance on the physical traits of political candidates is yet another demonstration of how irrational and uninformed most voters are (cf. Lenz & Lawson, 2011; Banducci et al., 2008; Olivola & Todorov, 2010. See also Healy et al., 2010; Rutchick 2010; Achen & Bartels, 2004). In a nutshell, these findings have been taken as yet another verification of Churchill’s statement that “the best argument against democracy is a five-minute conversation with the average voter.”
Evolutionary psychologists, in contrast, have viewed these findings in another light. Physical features in the face or the voice have been evolutionarily recurrent cues to ancestrally relevant traits such as prowess and strength (Tigue et al., 2012; Laustsen & Petersen, 2015, 2017; Spisak et al. 2012; van Vugt & Grabo, 2015; von Rueden & van Vugt, 2015). According to an emerging view, it is therefore plausible that the effects of physical features in modern elections reflect a psychological system of adaptive followership designed to generate preferences for leaders with traits that were ancestrally relevant for problem-solving (van Vugt & Grabo, 2015; von Rueden & van Vugt, 2015; Laustsen & Petersen, 2015; Spisak et al., 2012a). In this view, the reliance on the physical traits of political candidates does not reflect a lack of political sophistication. Rather it reflects the activation of an exceptionally sophisticated psychology in a modern context (i.e., political elections) that resembles its environment of evolutionary adaptedness (i.e., followership decisions).
In this article, we seek to add to this discussion by analyzing the effects of the facial traits of political candidates among a population who cannot be accused of being politically uninformed or unaware: Active members of political parties. We analyze how facial traits shape the degree of internal organizational support a candidate receives as indexed by a candidate’s position on local election ballots in Denmark. These ballot positions are determined through nomination processes within the local party organizations and are important assets for the candidates during the actual election (Blom-Hansen, 2016; Meredith & Salant, 2013; Ho & Imai, 2008). We focus on the effects of the two most well-studied physical traits, facial traits related to perceived competence and dominance, and demonstrate that active members of political parties are influenced by these traits in ways comparable to the general public. These findings not only provide additional evidence for the existence of these effects but also provide evidence that they are unlikely to emerge from lack of political engagement or sophistication.
Modern Elections and Evolved Followership Psychology
Studies have identified at least two sets of robust findings that link candidates’ facial appearance to election results. First, and most well-established, studies document that candidates—regardless of their gender and partisan affiliation—stand a better chance of getting elected the more competent they appear (e.g. Todorov et al., 2005; Hall et al., 2009; Olivola & Todorov, 2010; Laustsen 2014). This finding most often emerges from a research design in which naïve raters—individuals completely unfamiliar with the candidates—indicated their competence perception of each of the candidates based solely on immediate and spontaneous perceptions from face photos of the candidates. Next, these ratings are averaged across raters yielding average candidate scores of facial competence which are then used to predict the real electoral success of the candidates (usually also controlling for factors such as gender, age, incumbency etc.).
Second, using survey and laboratory experiments, candidates who appear dominant and masculine are found to be preferred above their feminine and non-dominant appearing counterparts in experimental contexts of social conflict (Little et al., 2007; Spisak et al. 2012a, 2012b; Laustsen & Petersen 2015, 2016, 2017). Importantly, the effects of the degree of conflict are paralleled by effects of individual differences that relate to perceptions of conflict such as political ideology. Conservatives are found to prefer dominant appearing leaders more than liberals and this tendency has been linked to psychological tendencies to favor group dominance and hierarchy as measured by Social Dominance Orientation (Laustsen & Petersen 2015, 2016, 2017; Laustsen, Petersen & Klofstad 2015). These findings have been extended to actual elections (utilizing the above described research design) such that conservative candidates (i.e., candidates representing conservative parties and, consequently, catering to conservative audiences) are found to receive more votes when they look dominant whereas dominant-looking, liberal candidates receive fewer votes (Laustsen & Petersen, 2016).1
These findings can be, and have been, interpreted as the reflection of an adaptive psychological system of followership (van Vugt & Grabo, 2015; von Rueden & van Vugt, 2015; Laustsen & Petersen, 2015, 2017; Spisak et al., 2012b). Over the vast majority of evolutionary history, humans lived in small-scale societies consisting of a few hundred individuals. While group-living exerted significant fitness advantages, it also created a series of social problems related to the coordination and assurance of contributions to collective action (e.g. Cosmides & Tooby, 1992; Tooby, Cosmides & Price 2006; van Vugt 2006). One way to solve problems of coordination is by handing decision-making authority to a single person, the leader (van Vugt & Ahuja, 2010; van Vugt, Hogan & Kaiser, 2008; van Vugt 2006; Spisak et al. 2012b). Therefore, according to researchers of evolutionary models of leadership, the forces of natural selection selected for psychological mechanisms regulating whom to follow and grant decision-making authority to. These psychological mechanisms as a whole comprise what is referred to as the followership psychology or the psychological system of adaptive followership (see for instance Laustsen & Petersen, 2015; van Vugt & Grabo, 2015; von Rueden & van Vugt, 2015).
While being composed of a large number of specific mechanisms, researchers (Spisak et al. 2012a, 2012b) recently presented the “biosocial leadership categorization model” which divides these mechanisms into two broad categories related to general and context-specific leadership evaluations. This model proposes that followership decisions are two-step processes: First, followers categorize potential leaders from non-leaders through assessments of each potential leader against a cognitive template or mental prototype of general leadership abilities. This first processing step identifies the pool of individuals who could potentially lead the group. It is likely that this first step focuses on general correlates of competence. Being competent in the sense of being intelligent, able to catch the attention of people, able to “get the job done” is, for example, something that should be valued in leaders in general. Because of the small-scale nature of ancestral groups, the set of ancestrally recurrent cues to these traits would obviously be rich. At the same time, it is plausible that this set would include physical cues such as physical attractiveness and physical cues to determination including posture and gaze (although, it should be noted that research has yet to establish an empirical association between such cues and actual leadership abilities).
Second, after initial identification, the model proposes that followers factor in the specific set of problems facing the group and, hence, seek to identify the leader most optimal under present circumstances. At this second stage, followers again evaluate a potential leader against an existing template, but at this stage the template specifically relates to the current problem context (Spisak et al., 2012b; for similar non-evolutionary models of implicit and contingency based leadership theories see for instance Lord et al. 1984; Schyns & Meindl, 2005). Past studies have focused on a particular type of problem that our ancestors recurrently faced: Social conflict. Moreover, it has been argued that preferences for leaders should depend on the extent to which such conflicts are present as, for example, reflected in the historical traditions of Native Americans of having both a war and a peace chief (Hoebel, 1954; Price & van Vugt, 2013). Between-group conflict is one of the most cooperative endeavors that exists and it requires intensely coordinated action to prevail against other groups (Laustsen & Petersen 2015; von Rueden et al. 2014). In situations of social conflict, followers should therefore put premium on traits that predict abilities to enforce collective action such as dominance. Also at this stage, the set of relevant ancestrally recurrent cues could include physical ones such as bodily or facial features relating to masculinity and strength (see Sell et al. 2009; Sell et al., 2010; Tigue et al. 2012). Under peaceful situations, in contrast, followers should generally avoid assertive and potentially exploitative leaders (Boehm, 2000; Hibbing & Alford, 2004; van Vugt, Hogan & Kaiser, 2008; van Vugt & Ahuja, 2010; Price & van Vugt, 2013; Bøggild & Laustsen, 2016).
As described above, a number of studies have found evidence consistent with this and demonstrated that preferences for dominant and masculine facial features increase in the face of war and conflict (Little et al., 2007; Hall et al., 2009; Little, Roberts, Jones & DeBruine, 2012; Little & Roberts, 2012; Spisak et al., 2012a, 2012b; Laustsen & Petersen, 2015, 2016, 2017; Tigue et al., 2012). As also noted, individual differences in perceptions of conflict predict preferences for dominant facial features. This research has in particular focused on political ideology. Hence, conservatives and liberals differ in their fundamental world views with conservatives generally seeing the social world as more competitive, conflict-ridden and threatening than liberals who, in contrast, tend to perceive the world as more safe, secure and friendly (Jost et al., 2003; Duckitt & Sibley, 2010; Hibbing et al., 2013; Oxley et al., 2008). Consistent with this, conservatives tend to have a stronger preference for facially dominant individuals as leaders than liberals (Laustsen & Petersen 2015, 2016; for a similar test with respect to preferences for vocal cues to dominance see Laustsen et al., 2015)2.
This model, we argue, provide an integrative explanation for the previous findings on the role of facial traits related to perceived candidate competence and dominance in modern elections. Modern elections are essentially contests between political candidates for leadership and, hence, should activate any psychology designed for followership in voters. In contrast to individuals in ancestral groups, voters in modern environments are rarely personal acquaintances with their potential leaders. However, because of the mediatized nature of modern politics, modern voters do have significant information about the physical characteristics of candidates and would be able to form impressions of competence and dominance on this basis. Consistent with the notion that competence enters at the first stage of followership decisions, all voters should prefer candidates who appear competent and, in support of this, preferences for competent candidates does not seem to differ across the ideological spectrum. At the second stage, in contrast, preferences for candidates that appear dominant should indeed vary with perceptions of conflict and individuals who do not believe that society faces social conflict—such as liberals—should prefer non-dominant leaders (see Laustsen & Petersen 2015, 2016, 2017; Laustsen 2016).
From the perspective of evolved followership psychology, these effects reveal that modern voters utilize shallow but potentially relevant cues when making voting decisions. Within political science, in contrast, it has been argued that these effects reflect irrational voters who lack sophistication and political engagement. To substantiate this claim, Lenz & Lawson (2011) demonstrate that the reliance on facial cues of competence are greatest among individuals who know the least about politics in the United States and watch a lot of television exposing them to candidates’ visual appearances. Moreover, a set of studies find that trait judgments based on only rapid exposures to faces (as short as 100 ms) are highly correlated with judgments made without any time constraints (Willis & Todorov, 2006; Olivola & Todorov, 2010). In direct relation to electoral contests, Ballew & Todorov (2007) show that the effects of facial competence on voting decisions are largest when raters are not asked to deliberate and make good competence judgments of the faces, and that very brief exposures to candidate faces (100-250 ms) are as good at predicting election results as are judgments made without time constraints. In sum, this suggests that face-based character judgments constitute a heuristic device made obsolete by more systematic thought. Still, at present it is safe to say that the evidence for the assertion that physical cues are primarily used by political unsophisticated individuals is mixed. Hence, another recent study, while well-powered, failed to find any evidence of the interaction between sophistication and the tendency to rely on facial information among Spanish subjects (Brusattin, 2012). In this study, individuals irrespective of sophistication were influenced by facial cues of competence. Similarly, Laustsen & Petersen (2016) found that the effects of facial dominance on electoral success obtains both for candidates who have and have not been in office before, even while voters would be expected to have more and alternative information about incumbent candidates.
One potential reason for the mixed findings is that, in general, political scientist have found it difficult to obtain individual difference measures in the mass public that clearly distinguish between those who approach politics in rational, “unbiased” ways and those that do not (compare, for example, Kam, 2005; Taber & Lodge, 2006). In this study, we therefore follow an alternative research strategy. We focus directly on a subject population that clearly is highly engaged in politics and have important stakes in making the right political decision. As population, we focus on active members of party organizations, who make decisions about which candidates can run for office, and we examine these party organizations’ actual choices of candidates in real-world elections. Specifically, we focus on how local Danish party organizations order candidates on election ballots.
If these decisions are shaped by evolved followership psychology, we should expect two predictions to hold true within these organizations. First, in connection to the evaluation of general leader potential, we predict that all party organizations—regardless of their ideological leaning—will put a premium on candidates who look competent. This is captured by the facial competence prediction: Facial competence will positively predict candidates’ nomination success for all candidates regardless of ideological leanings. Second, in connection to the evaluation of problem-specific leader potential, we expect that conservative party organizations will evaluate dominant looking candidates more positively than their liberal counterparts leading conservative constituencies to put a premium on facial dominance while liberal constituencies should prefer non-dominant looking candidates. This leads to the facial dominance prediction: Facial dominance will be a relatively more positive predictor of nomination success for conservative than for liberal candidates. Methods and Materials
To test our predictions we follow existing studies on face-based voting and combine naïve respondents’ ratings of candidates’ traits—i.e., competence and dominance—from mere photos of the candidates with actual data on the candidates’ success. Prior work has provided evidence that ordinary voters’ electoral behavior and preferences correlate with face-based trait inferences of the candidates. In the below analyses we change the focus from such party external processes to party internal processes of nomination of candidates linking candidates’ visual appearance to a valid measure of the support from within their own party: Ballot position (see Blom-Hansen et al. 2016).
Data was collected from the 2009 Danish local elections in the three municipalities Brønderslev, Frederikshavn and Mariager Fjord which provides a total of 268 candidates of whom 257 were running for one of eight main Danish parties (the remaining 11 candidates were running for “local parties” campaigning on local matters only. Consequently, these candidates are not possible to position in either a liberal or a conservative party block). While these data have previously been used to investigate the relationship between candidates’ facial competence and dominance, respectively, and their electoral success (see Laustsen 2014, Laustsen & Petersen 2016) they have not been used to explore relationships with candidates’ nomination success as measured by their ballot positions.
Danish local elections are held each fourth year. They are proportionally representative elections with multiple candidates nominated from, most often, eight or nine parties in each municipality or district. Voters can vote for either the party or for a given candidate from a party with approximately 75 percent of the voters choosing the latter (Elklit, 2013: 50). Media coverage and public attention is typically lower on local than on national elections in Denmark, and this attention and interest difference is also apparent in turnouts for the last local and national election, respectively: 71.9 percent of the registered Danish voters took part in the 2013 local elections (65.8 percent in the 2009 local elections), while—in comparison—85.9 percent voted in the 2015 national election.
To conduct our analyses, we gathered data on all candidates running in three different municipalities where altogether 25 party organizations nominated on average 19 candidates to their ballots (range = 1 – 31; median = 22). In Denmark, it is the local party organizations that determine how many candidates are running for seats in the municipality council (from their given party) and the order in which they are listed on the ballot. However, no formal rules exist regarding the exact procedures for intra-party nomination processes. This leaves some room for procedural differences between parties. According to Blom-Hansen et al. (2016: p. 174), “(T)he order of the candidates on the list is normally decided in two steps, in some – but certainly not in all – cases by a ballot among local party members”. The top ballot position is usually allocated first and reflects who will be the party’s mayoral candidate. In the second step, the order of the remaining candidates is determined. In this way, the final ranking of the candidates reflects “the order in which [a given party] would like the candidates elected” (Blom-Hansen et al. 2016; p. 174). Consistent with this, existing research shows that this signal is used by voters such that candidates higher on the ballot receives more votes than candidates lower on the ballot (see Ho & Imai, 2008; Meredith & Salant, 2013). Blom-Hansen et al. (2016) characterize the details of the nomination process as “an internal party matter” where local party committees construct their lists in different ways. However, for our present purpose, the most important feature of these intra-party nominations is that they—despite differences in exact procedures—all express local party organizations’ and party members’ candidate preferences; often formed on the basis of intra-party debates and speeches from the candidates.3 Overall, a candidate’s position vis-à-vis fellow partisan candidates constitutes a valid and direct measure of her/his success in party internal nomination contests.
Facial traits. To get reliable measures of candidates’ facial traits, we first compiled standardized photos (200 × 250 pixels) of the 268 candidates (70 females and 198 males)4 running for city councils in the three northern Danish municipalities from an online database offered by the regional newspaper. Next, we recruited respondents that were fully unfamiliar with the candidates to rate facial competence and dominance (as well as other traits) from the photos. Specifically, we recruited 646 young Danish raters (approximately 16-20 years old) from high schools in other geographic regions than the municipalities in which the candidates were running for city councils.5 Importantly, rater age should not affect ratings of facial traits since even 3-4 year old children are demonstrated to make reliable and consistent judgments of competence and dominance from faces (Cogsdill et al. 2014). The raters were randomly assigned to a subset of eight or nine candidates and instructed to rate the candidate faces on seven different traits using 0–10 scales (“0” indicated minimal and “10” indicated maximum degree of a given trait): Dominance, physical strength, attractiveness, friendliness, accountability, competence and intelligence. This yielded a respondent-to-candidate ratio of 19 following other similar rating tasks in previous work (Banducci et al., 2008; Rosar et al. 2008). To reduce the number of rated traits, we conducted a Principal Component Factor analysis which showed that the traits cluster along two separate dimensions of competence (with the traits accountability, competence and intelligence loading the strongest) and dominance (with dominance and friendliness loading the strongest – and friendliness negatively so), respectively (see table 1 in Laustsen & Petersen, 2016 (p.198) for specific factor loadings and details on the Principal Component Factor analysis). In the following analyses we rely on the candidates’ independent factor scores on the competence and dominance dimensions as our measures of competence and dominance. To ease interpretation we recode both scales to a common 0-1 framework on which “0” and “1” reflect low and high competence (M = 0.58; SD = 0.18) and dominance (M = 0.37; SD = 0.17), respectively (for more information on the used facial materials and the rating task see Laustsen 2014; Laustsen & Petersen 2016).
Within-Party Support. As operationalization of our primary dependent variable, the support from members of the local party organization, we employ candidates’ positions on the ballot in the 2009 local election. As described above, local parties prioritize their candidates on the ballot providing a valid measure of party organizations’ candidate preferences. That is, the top candidate is put on top of the list, followed by the second-most preferred candidate, and ending with the least preferred candidate at the bottom of the list. Since a higher ballot position translates into more votes in actual elections (Ho & Imai, 2008; Meredith & Salant, 2013), a good ballot position is an important objective in within-party competition. Because the number of nominated candidates vary across parties (with more candidates nominated by larger parties), we cannot use candidates’ actual position on a given ballot since for instance being positioned as number 5 is different in a party nominating altogether 12 candidates compared to a party only nominating 5 candidates. Instead we employ a dichotomous measure of whether a candidate was located in the upper 20 percent of the ballot (coded “1”) or if (s)he was nominated below the top 20 percent (coded “0”). Importantly, because being positioned within or below the top 20 percent on a given ballot constitutes a somewhat arbitrary success criterion, we employ a series of alternative dichotomous measures of nomination success (positioned within or below the upper 25 percent, 30 percent, 40 percent and 50 percent). The results of these analyses are reported in the Online Supporting Materials.
Candidates’ ideology, age, sex, incumbency and municipality. We include the candidates’ ideology as moderator throughout the analyses. Due to the relatively low number of candidates from smaller parties in this sample, ideology is measured dichotomously based on candidates’ party affiliation and the block affiliation of that party (with either the liberal and left-leaning political block or the conservative and right-leaning political block). Candidates from the liberal block are coded ‘0’ while candidates from the conservative block are coded ‘1’.6
In the analyses we further control for candidates’ sex, age and incumbency. Because data on candidates actual age was unavailable we rely on a trichotomized categorical variable of the candidates’ perceived age (under 30 years; between 30 and 60 years; over 60 years of age) as rated by a group of raters (following Berggren, Jordahl, & Poutvaara, 2010).7 Finally, we control for incumbency measuring if a given candidate had already won a seat at the previous corresponding election or if (s)he held any regional or national seat while running for the city council. Finally, we control for municipality to cancel out potential differences across districts.8
Analyses. We test our prediction that candidates’ facial traits also relate to their ballot positions using logistical regressions for predictions of a top 20 percent ballot position. Following standard procedures we include robust standard errors clustered at the party level (cf., Berggren et al., 2010). Because only 257 of the candidates represent mainstream parties with a clear block affiliation (with either the rightwing and conservative or the leftwing and liberal Danish parties) models in the main text are based on these candidates (results for facial competence replicate using all 268 candidates; see Online Supporting Information S.I.8). We report regression coefficients with corresponding standard errors and level of significance, and to ease interpretation we illustrate main findings with predicted probabilities for being positioned in the top 20 percent of the ballot for low and high levels of facial competence (across all candidates) and for low and high levels of facial dominance (for liberal and conservative candidates separately).