Gender and Perceptions of Candidate Competency

This article was downloaded by: [sue thomas]On: 15 May 2012, At: 17:12Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Women, Politics & PolicyPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/wwap20

Gender and Perceptions of CandidateCompetencyRebekah Herrick a , Jeanette Mendez a , Sue Thomas b & AmandaWilkerson ca Oklahoma State University, Stillwater, Oklahoma, USAb Pacific Institute for Research and Evaluation (PIRE), Santa Cruz,California, USAc Public Outreach Fundraising, Seattle, Washington, USA

Available online: 11 May 2012

To cite this article: Rebekah Herrick, Jeanette Mendez, Sue Thomas & Amanda Wilkerson (2012):Gender and Perceptions of Candidate Competency, Journal of Women, Politics & Policy, 33:2, 126-150

To link to this article: http://dx.doi.org/10.1080/1554477X.2012.667748

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representationthat the contents will be complete or accurate or up to date. The accuracy of anyinstructions, formulae, and drug doses should be independently verified with primarysources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.

http://www.tandfonline.com/loi/wwap20

http://dx.doi.org/10.1080/1554477X.2012.667748

http://www.tandfonline.com/page/terms-and-conditions

Journal of Women, Politics & Policy, 33:126–150, 2012Copyright © Taylor & Francis Group, LLCISSN: 1554-477X print/1554-4788 onlineDOI: 10.1080/1554477X.2012.667748

Gender and Perceptions of CandidateCompetency

REBEKAH HERRICK and JEANETTE MENDEZOklahoma State University, Stillwater, Oklahoma, USA

SUE THOMASPacific Institute for Research and Evaluation (PIRE),

Santa Cruz, California, USA

AMANDA WILKERSONPublic Outreach Fundraising, Seattle, Washington, USA

Following an innovative study showing rapid inferences about thecompetence of candidates based on photos correlated with elec-toral success, we examine the effects of the sex of the subjects andcandidates on these results. Our results indicate that the relation-ship between inferences of competence and electoral success aremore complex than previously believed. We found a gender gapin evaluations of competence and maturity of candidate facesand in support for women candidates; however, an overall pref-erence among all subjects for men candidates. Additionally, therelationship between competence and victory is affected by a hostof variables unconsidered previously.

KEYWORDS facial appearance, candidate competency, genderand campaigns, attractiveness, maturity, competence

Do citizens judge political candidates, including candidates for US Congress,based on perceptions of competence? A great deal of political scienceresearch indicates that they do (Fenno 1978; Davidson, Oleszek, and Lee2008; Jacobson 2008). Can judgments about competence be based on criteriaas seemingly unrelated to that important trait as facial appearance and struc-ture? And, if so, do those judgments correlate with electoral victory? Todorov

Address correspondence to Jeanette Mendez, Department of Political Science, OklahomaState University, 213 Murray Hall, Stillwater, OK 74078. E-mail: [email protected]

126

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2

Gender and Perceptions of Candidate Competency 127

and colleagues (2005) suggest that the answer to the latter two questionsis yes. In a study published in Science magazine, these scholars found that“inferences of competence based solely on facial appearance predicated theoutcomes of US congressional elections better than chance” (Todorov et al.2005, 1623).

In light of these startling results, we seek to extend the Todorov and col-leagues (2005) study to investigate how enduring findings in gender politicsresearch affect the Todorov model. In specific, we are interested in findingsindicating that although women candidates for political office in the UnitedStates are as likely as men to win electoral races, they face different setsof challenges and opportunities than do men candidates (see, for example,Herrick 2001; Lawless and Fox 2005; Thomas 2002, 2005). In this study, weask the following questions: (1) Are women candidates for Congress judgedas equally, more, or less competent than men candidates based on facialstructure? (2) Are there differences in competence ratings of women andmen candidates depending on whether candidates are rated by female ormale subjects? (3) Do gender-based perceptions of candidate competenceand sex of the candidates affect success in simulated and true electionsfor the US Congress? Answers to these questions will extend the rapidlygrowing literature on the public’s interest in and willingness to increase theproportions of women serving in the federal legislature.

THE LITERATURE

The Todorov and colleagues (2005) study was based on social psychologi-cal research demonstrating that faces are an important source of informationabout others and that people extrapolate from appearances to reach conclu-sions about “personal dispositions.” The Todorov team conducted a seriesof experiments. In the first, subjects were shown pairs of photos of realcandidates for the US House of Representatives and US Senate in the earlypart of this decade. If subjects recognized any of the faces, those pairs wereexcluded from analysis so that results based on prior judgments rather thanon perceptions of competence alone were avoided. In this phase of theresearch, the candidates perceived as more competent won in more than70% of the Senate races and slightly more than 66% of the House races. In asecond round of studies, time constraints were introduced to the process(one second per pair of faces) to gauge impressions that are not overriddenby desires to provide any type of socially appropriate response. These rapidjudgments predicted slightly more than 67% of the races. In yet anotherexperimental round, to address alternative hypotheses about trait inferences,subjects were asked to make judgments on seven trait dimensions basedon photos of faces.1 The results: only the judgments related to competencepredicted electoral outcomes. Finally, after rating candidates on the seven

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2

128 R. Herrick et al.

aforementioned dimensions, subjects were shown pairs of photos and askedfor whom they would have voted in the races. In both simulated and trueelections, voting preferences were correlated with judgments of competencederived from facial appearance.

The most remarkable aspect of this research is that the subjects did notrecognize the photo pairs that the researchers used; therefore, the subjectshad no opportunity to form preconceived notions of candidate competencebased on other factors such as familiarity, incumbency, political party, issuestances, and the like. To be sure, voting in real rather than simulated electionsprovides opportunities for voters to revise initial impressions of candidatesbased on appearance or other factors. Certainly, party identification, ideol-ogy, issue positions, experience, and the like can and do matter. Todorov andcolleagues (2005) argue, however, that “additional information may weakenthe relation between inferences from faces and decision, but may not changethe nature of the relation” (1625). Another way to understand this approachis that visual cues about candidates may form the foundation upon whichaffective or cognitive cues are built. The resulting structure (voting decisions)will reflect both affective and cognitive dimensions, but cognitive cues maynot eradicate the impact or direction of affective perceptions.

Bolstering the Todorov team’s approach is a fairly robust trove ofresearch on correlates of facial appearance in the political realm andbeyond. For example, across a series of studies that touched on the politicalimplications of facial appearances, Rosenberg and associates found that can-didates’ likableness, leadership, competence, and congressional demeanorwere inferred from photos of the faces of mock candidates, and that theseinferences affected voter preferences (Rosenberg et al. 1986; Rosenberg,Kahn, and Tran 1991; Rosenberg and McCafferty 1987).2 Lewis and Bierly(1990) found that judgments of House members’ competence were affectedby perceptions of their attractiveness—and that female subjects exhibited a“pro-female” sentiment when rating women representatives.

Candidate appearances matter even when presented with informationadditional to photos. For example, in simulated elections, Budesheim andDePaola (1994) found that:

(a) physical appearance (attractiveness in this case) influenced evalua-tions even when individuating personality information was provided; (b)subjects’ evaluations were less influenced by their agreement with thecandidates’ issue positions when image information was presented thanwhen it was not; and (c) subjects’ evaluations were less influenced byissue agreement when a candidate’s image was evaluatively mixed thanwhen it was evaluatively consistent.3 (339)

Candidate appearance, including facial traits, can even be manipu-lated to affect candidate evaluations and viability. For example, Rosenberg

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


and McCafferty (1987) found that two different photos of the same mockmen candidates resulted in different assessments of the candidate. Similarly,Rosenberg, Kahn, and Tran (1991) report that the electoral chances of womencandidates were improved by having makeup artists and photographers useinformation about positive facial features, including eye shape, eyebrowwidth, skin texture, hair style, and facial shape to create favorable photos.4

Another facial dimension likely to affect perceptions of competence inpolitics and beyond may be called the maturity and baby-face dimension.Baby faces are round, with large eyes, small noses, high foreheads, smallchins, and high eyebrows, while mature faces are the opposite (Berry andMcArthur 1985; Zebrowitz and Montepare 2005, 1565). Baby-faced individ-uals are perceived to be less competent than others, most likely becausepeople overgeneralize and attribute other babylike characteristics to them(Zebrowitz and Montepare 2005).

In terms of real-world effects, Berry and Zebrowitz-McArthur (1988)found that baby-facedness affected sentence and conviction patterns in sim-ulated criminal trials. Mature-faced individuals were held more accountablefor their intentional actions. Zebrowitz and McDonald (1991) found, amongother results, that, in small-claims cases, the more baby-faced the defen-dants, the more likely they were to lose in negligence cases and to win inintentional cases. Central to evaluations of candidates is Brownlow’s (1992)research on how subjects responded to a baby-faced or mature-faced femalespeaker delivering a persuasive communication. She found that subjects weremore likely to be persuaded by baby-faced speakers when trust was an issue,but were more likely to be persuaded by matured-faced speakers whenexpertise was in question.

Directly related to the political realm, researchers have found that arti-ficially altering the maturity levels of photos of presidents’ faces affectedsubjects’ evaluations of the presidents. For example, increasing PresidentClinton’s babyish features resulted in him being judged more attractive, com-passionate, and honest than was the case in either an unaltered pictureor a picture altered to appear more mature. Comparing altered photos ofPresidents Kennedy, Reagan, and Clinton that increased maturity resulted inKennedy and Reagan being seen as more powerful, dominant, strong, andcunning than they were in unaltered photos or photos altered to increasebaby-facedness—although increasing maturity in Clinton’s photo did not alterpower ratings of him (Keating, Randall, and Kendrick 1999). Directly tyingmaturity levels of faces to perceptions of politicians’ competence, Poutvaara,Jordahl, and Berggren (2009) find that baby-facedness is negatively relatedto inferred competence in Finnish politicians.

Of central concern to the present study, the maturity of faces has alsobeen found to interact with individuals’ sex to affect perceptions of peo-ple. Berry and McArthur (1985) found that baby-faced males were judgedas more naïve than their mature-looking male counterparts. Baby-faced men

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


were also seen as less powerful and warmer (Berry 1991). The effects ofsuch perceptions are borne out in a host of research studies. Brownlow andZebrowitz (1990) found that baby-faced females were cast in television com-mercials because they were seen as the least expert but most trustworthyspokespersons, while matured-faced men were considered to be the mostexpert but least trustworthy spokespersons. Friedman and Zebrowitz (1992)report that when baby-faced females were compared with mature-facedmales, sex-role stereotypes regarding warmth and power were present; thesestereotypes were reversed when mature female faces were compared tobaby-faced males. Since there is an overlap among gender, baby-facedness,and sex-role stereotypes, these scholars argue that sex differences in facialmaturity levels are directly related to sex-role stereotypes.

In sum, facial type and appearance has been demonstrated to corre-late with ability to be credible and to be successful in the public realm,even in light of additional information. Most important for the present study,facial maturity and gender may interact to affect women candidates’ polit-ical electability in ways different than their male counterparts. It is to theexploration of these possibilities that we now turn.

HYPOTHESES, DATA, AND METHODS

Derived from the array of interdisciplinary social science literatures, and rest-ing heavily on political science, as discussed throughout this article, our studyrests on three sets of hypotheses that deal with perceptions of candidates byour total subject sample, perceptions of candidates by separate female andmale samples, and, finally, the extent to which gender-based relationshipsamong perceptions of competence, sex of candidates, and electoral victoryexist. The sets of hypotheses are discussed in the following sections.

Hypotheses for Subjects’ Perceptions of Candidates

1. Subjects will be more likely to rate baby faces as less competent thanmature faces.5

2. Subjects will be more likely to rate women’s faces as less competent thanmen’s faces.

3. Subjects will be more likely to rate women’s faces as less mature thanmen’s faces.

Although baby faces are associated with both positive and negativeattributes, we hypothesize that congressional candidates with baby faces,most notably women, will be disadvantaged compared to their counterparts.A deep well of research findings, introduced previously, indicating that baby

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


faces are perceived as less competent, less powerful, and less expert, butmore naïve, warmer, and more trustworthy, underlie this prediction. Politicalscience studies indicate that the inferences of baby-faced people also dove-tail with stereotypes of women politicians and the traits and issues withwhich they are associated. For example, public perceptions of women leg-islators include that they are more compassionate, warmer, more honest,less assertive, more task oriented, and more interested in issues related toeducation, poverty, health care, the environment, and the welfare of chil-dren and families than are men legislators. Men, on the other hand, are seenas dominant, tough, decisive, assertive, in possession of technical expertise,and stronger on issues of crime suppression and punishment, economic per-formance, military and foreign policy, trade, taxes, agriculture, security, andterrorism (Dolan 2004a, 2004b; Eagly and Carli 2003; Huddy and Terkildsen1993a, 1993b; Rosenwasser and Dean 1989; Sanbonmatsu 2002). In addition,in national politics, strength, power, decisiveness, and maturity are valued.This is borne out in research by Huddy and Terkildsen (1993a) that finds vot-ers prefer candidates running for national office who have traits traditionallyassociated with men, such as a high level of instrumentality and competenceon issues that are typically seen as male issues, such as those just introduced.Since instrumentality and masculine traits are similar to those traits associ-ated with mature faces, voters may be more likely to perceive mature-facedcandidates and men candidates as more competent for federal office.

Hypotheses for Female and Male Subjects’ Perceptions of Candidates

4. Female subjects will be more likely than male subjects to judge baby facesand mature faces of candidates as equally competent.

5. Female subjects will be more likely than male subjects to judge womenand men candidates as at least equally competent.

6. Male subjects will be more likely to judge women candidates as lesscompetent than men candidates.

In the first set of hypotheses, we addressed investigations of the total groupof candidates for office undifferentiated by sex of the subjects. In this set,we direct attention to questions of whether subjects’ sex is correlated withjudgments about the maturity of the faces of all candidates, women and mencandidates, and assignments of competency. Here again, a host of socialscience research, especially political science studies on women in politics,provides the foundations for these expectations of differences in judgmentsof female and male subjects.

First, because women tend to be more supportive of compassion issuesthan men (Clark and Clark 1999; Huddy and Terkildsen 1993b; Whitaker2008), they may be less likely to perceive a large competence-related gapbetween baby faces and mature faces. The warmth associated with baby

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


faces may be a plus for women subjects as they judge competence. Second,although subjects’ sex may not have a significant effect on evaluationsof attractiveness (Jackson, Hunter, and Hodge 1995), female subjects mayexhibit a similarity bias. Because people prefer candidates and others whoare similar to themselves, women may prefer female faces whereas menmay prefer male faces. In fact, Lewis and Bierly (1990) found that femalesubjects tended to judge women House members as more competent thanmen House members, but no distinctions were evident among the responsesof male subjects. In any case, any general tendency of voters to prefersome male traits may be offset by these types of tendencies among femalesubjects.6

Hypotheses for the Relationship of Competence to Electoral Victory

7. All else equal, candidates judged as most competent will be more likelyto be selected by subjects as those for whom they would vote.

8. Candidates judged as the most competent will be more likely to havewon the true election.

9. Female subjects will be more likely to “vote” for women candidates thanwill male subjects in both elections.

10. Women candidates will be less electorally viable than men candidates,but controlling for competence will diminish the effects of sex of thecandidate on visibility.

In this set of expectations, we follow Todorov and colleagues (2005) inpredicting that judgments of competence will lead to vote choice and then,based on the extensive relevant literature in political science and psychologydiscussed throughout this article, extend their research by investigating theextent to which the sex of the candidate and sex of the subjects affects thisrelationship.

To test these hypotheses, we conducted three sets of surveys in sev-eral introductory courses in American government in a large Midwesternuniversity.7 In each survey, subjects were asked to evaluate photos or pairsof photos of real candidates who ran for US Congress in 2004 and 2006.Photos were selected from CNN’s Web site (http://www.cnn.com) and wereconverted from color to black and white to standardize them. In addition,to control for variance in the quality and backgrounds of the photos for theportion of the experiment in which candidates were paired, we were carefulto use photos that were similar on these dimensions to the greatest extentpossible. Completing each survey took 10 to 15 minutes.

In Survey 1, to allow us to separate perceptions of maturity fromperceptions of competence, 94 subjects were asked to estimate candi-dates’ maturity based on 156 photos of candidates presented individually.Specifically, the subjects were instructed to estimate the maturity of each

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


candidate on a five-point scale. To accomplish this, the subjects were giventhe following definition:

A “1” is “least mature,” meaning the candidate has a baby face, with roundlarge eyes, small noses, high foreheads, small chins, and high eyebrows.A “5” is “most mature,” meaning the candidate has a long or oblong faceshape, small eyes, medium to large nose, small forehead, and medium tolarge chin.

The need for and substance of this definition was based on the work of Berryand McArthur (1985) and Zebrowitz and Montepare (2005, 1565). In thosestudies, definitions were offered to ensure that students did not mistake theterm mature for aged.

The task for subjects in Survey 2 was evaluating candidate attractive-ness, because attractive individuals are often afforded “halo effects” forwhich they are attributed more positive traits than unattractive people (SeeZebrowitz 1997, chapter 7). In addition, candidate attractiveness can affectvoters’ evaluations of candidates and vote choice (Budesheim and DePaola1994; Poutvaara, Jordahl, and Berggren 2009; but see Mattes et al. 2009),although perhaps less so for women candidates (Sigelman, Sigelman, andFowler 1987).

To allow us to separate perceptions of attractiveness from perceptionsof competence, we executed a separate survey with a set of 78 subjects.That way, we were able to avoid contamination with the results obtainedin Survey 1. Subjects were asked to evaluate the attractiveness of each indi-vidual candidate on a five-point scale. Photos were the same as those usedin Survey 1 except that we removed 24 of the photos when some studentsrecognized their own members of Congress (MC), or because a particular MCwas prominently covered in the news at the time of the experiment (such asformer Indiana Representative Julia Carson, who died within the same timeframe the experiment was conducted, and whose photo was prominentlydisplayed in the news). That left us with 132 total photos. Subjects were toldto “rate the physical attractiveness of each candidate on a scale of 1–5 with‘1’ as the ‘least attractive,’ and ‘5’ as the ‘most attractive.”

The task for subjects in Survey 3 was evaluating candidate competenceand electability. As we did not want evaluations of maturity or attractivenessto influence evaluations of competence, we used a third set of subjects forthis survey (583 subjects), and then randomly split the subjects into twogroups (280 subjects in Group A, 303 subjects in Group B). Each wasasked to perform two tasks: to evaluate competence and to decide forwhom to vote. For the first group, the subjects evaluated the competenceof Candidate Group A. In the second group, subjects evaluated the compe-tence of Candidate Group B. Subjects in both groups identified their votechoice of all candidates. The photos were the same photos from Survey 2.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


For the first task of Survey 3, the subjects evaluated photos of individ-ual candidates’ competence using a five-point scale.8 As was the case in theTodorov and colleagues (2005) piece, we explicitly did not define the con-cept “competence.” To construct data for the second task pertaining to votechoice and to replicate the Todorov team’s 2005 study, we used the samephotos from the competence portion of Survey 3 and created 53 pairs ofsimilar-quality photos (meaning levels of resolution and clarity) with at leastone woman candidate. Of these, 49 were woman-man pairs, and 4 werewoman-woman pairs. We next identified similar-quality photos of man-manpairs for a total of 13 pairs. In all, there were 66 pairs of candidates. Subjectswere then shown the pairs of candidates and asked for which candidate theywould be most likely to vote.

A key concern with our research design was to avoid the possibility thatsubjects would attempt to override initial, intuitive reactions to the photos, sowe limited the time they had to make their evaluations. The research litera-ture suggests that two processes can be used to evaluate facial appearance orother stimuli, System 1 and System 2 processes: “System 1 processes are fastautomatic, effortless, associative, implicit (not available to introspection), andoften emotionally charged. . . . Difficult to control . . . System 2 are slowerserial, effortful and more likely to be consciously monitored and deliber-ately controlled” (Kahneman 2003, 698). System 2 processes can be used toreact to or control System 1 processes. For example, one may see a womancandidate, initially judge her to be of lesser competence, but override thatjudgment to avoid appearing sexist. A key to activating System 1 processes isto limit the amount of time subjects have to make an evaluation (Sczesny andKuhnen 2004). Because limiting the time subjects have to evaluate photoshas been shown to decrease System 2 corrections, following Todorov andcolleagues (2005), subjects in both Group A and Group B were given onesecond to make the evaluation.

Once all data were gathered and coded, we estimated regression modelsto test Hypotheses 1 through 6 first. For analyses of perceptions of com-petence, the independent variables in our models included perceptions offacial maturity, attractiveness of candidates, sex of the candidate, candidateage, and whether a flag appeared in the photos. For analyses of perceptionsof maturity, the same set of independent variables was used. Maturity wasnot included, as that was the dependent variable in these equations. Foreach of these dependent variables, analyses were performed for the entiresubject population as well as separate analyses for male and female popu-lations. The dependent variable in each of these models was the perceivedcompetence of the candidate. This was measured as the mean score of theresponses of all subjects for each candidate.9

To test Hypotheses 7 through 10, we estimated regression models topredict election outcomes for a simulated election (which candidate waschosen the winner by the subjects?) and for the true election (did the winners

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


of the real-life election match the winners “elected” by the subjects?). Thefinal models of the study measured electoral victory with and without thecompetence measure.

RESULTS

To test Hypotheses 1 and 2 regarding the relationship of baby faces andmature faces to perceptions of competence, and the relationship betweencompetence and sex of the candidates, we estimated ordinary least squares(OLS) regression equations of competence with and without the maturityindependent variable.

The results from Table 1 confirm both hypotheses. First, as maturityscores increased by 1 point on the 1 to 5 scale, candidates’ competencescores increased by 0.39 points. In light of the overall mean of 2.90, a changeof 0.39 is substantial. Thus, more mature faces are rated as more competentcandidates by our subjects.

Next, as predicted with Hypothesis 2, sex of the candidate had a signif-icant effect on evaluations of competence—and in the expected direction.Men candidates were rated 0.31 points higher on the competence scale than

TABLE 1 Regression Analysis of Competence by Maturity, CandidateSex, Flag in Picture, Candidate Age, and Attractiveness

Competence factor With maturity Without maturity

Attractiveness 0.53∗∗∗ 0.57∗∗∗

(0.06) (0.06)Maturity 0.39∗∗∗

(0.07)Candidate sex 0.31∗∗∗ 0.41∗∗∗

(0.04) (0.43)Flag in picture 0.14∗∗∗ 0.15∗∗∗

(0.04) (0.05)Candidate age −0.002 0.01∗∗∗

(0.003) (0.002)Constant 0.23 0.85

(0.24) (0.24)R-squared 0.64 0.55N 127 127

∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05.Note: Standard errors are shown in parentheses.Competence: Mean score of all subjects for each candidate based on 5-pointLikert scale, 1 = “least competent”; 5 = “most competent.” Maturity: Meanscore of all subjects for each candidate based on 5-point Likert scale, 1 =“least mature”; 5 = “most mature.” Candidate sex: 1 = male candidate; 0= female candidate. Flag in picture: 1 if the background in the picturehas a flag; 0 all others. Candidate age: candidate age at time of election.Attractiveness: Mean score of all subjects for each candidate based on 5-pointLikert scale, 1 = “least attractive”; 5 = “most attractive.”

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


were women candidates. Results of the same equation without the matu-rity variable show the same pattern, but this time the sex of the candidatewas even stronger, at 0.41; each result was statistically significant. To fur-ther investigate the import of the difference in candidate gender coefficientsin the equations with and without the maturity variable, we performed aSobel test with maturity as the mediating variable. Results confirmed thatthe mediating effect of maturity is statistically significant at 0.01. Thus,as expected, the maturity level of faces affected perceptions of compe-tence of candidates and accounted for some, but not all, of the effects ofcandidate sex.

One other statistically significant finding in the equations in Table 1illuminates the nature of the relationship between maturity levels of facesand competence. All else equal, candidates perceived as attractive were morelikely to be seen as competent by subjects. This holds true with and withoutthe maturity variable included in the estimates, although the effect is smallerwhen maturity is included.

In Hypothesis 3, we predicted that subjects would rate the faces ofwomen candidates as less mature than those of men candidates. To deter-mine this, we modeled perceived candidate maturity as a function of thesex of the candidate. The maturity measure used was the mean score of allsubjects for each candidate. Table 2 displays the results of the regressionanalysis for Hypothesis 3.

TABLE 2 Regression Analysis of Maturity by Attractiveness,Candidate Sex, Flag, and Candidate Age

Trait Coefficient

Attractiveness 0.09(0.08)

Candidate sex 0.27∗∗∗

(0.05)Flag in picture 0.02

(0.05)Candidate age 0.02∗∗∗

(0.003)Constant 1.59

(0.28)R-squared 0.50N 127

∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05.Note: Standard errors are shown in parentheses.Maturity: Mean score of all subjects for each candidate based on 5-point Likert scale, 1 = “least mature”; 5 = “most mature.” Candidatesex: 1 = male candidate; 0 = female candidate. Flag in picture: 1 ifthe background in the picture has a flag; 0 all others. Candidate age:Candidate age at time of election. Attractiveness: Mean score of allsubjects for each candidate based on 5-point Likert scale, 1 = “leastattractive”; 5 = “most attractive.”

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


As Table 2 indicates, the results confirm the hypothesis. Men candidateswere perceived as more mature than women candidates, and the resultswere statistically significant. Specifically, men candidates had a predictedmaturity score of 3.38, and women candidates had a predicted score of3.11—a 0.27 difference. In the regression equation, only sex of the candidateand candidate age were significant predictors of maturity, and the effectsof age were quite small. Interestingly, the attractiveness variable, althoughfavoring men, was statistically insignificant as a predictor of maturity.

In sum, all three hypotheses pertaining to subjects’ perceptions of can-didates were confirmed by our data. Baby faces were perceived as lesscompetent than mature faces and, most central to this research study, thefaces of women candidates were perceived as both less mature and less com-petent than the faces of men candidates. This was true even when controllingfor another important facial feature, attractiveness.

The next set of hypotheses deals with the extent to which the sex ofthe subjects affects perceptions of candidate competence in terms of all can-didates, women candidates, and men candidates. Our models to test thesehypotheses used maturity ratings (the mean for all male subjects and themean for all female subjects), perceptions of attractiveness, and sex of thecandidate as the primary independent variables.

Table 3 presents the results of the multivariate competence models forboth female subjects and male subjects. Hypothesis 4, which predicted that,all else equal, female subjects would be more likely to judge baby faces andmature faces of the total group of candidates as equally competent, was notconfirmed. First, both female and male subjects saw mature faces as morecompetent, and the result were statistically significant.10 A beta test furtherindicates no significant difference between male and female evaluations.

The remaining two hypotheses in this set address the extent towhich sex of the subjects and sex of the candidates affect perceptions ofcompetence. Hypothesis 5 predicted that, all else equal, female subjectswould be more likely to judge women and men candidates as at least equallycompetent. As the results in Table 3 indicate, the data do not support thatexpectation. Among female subjects, the candidate sex coefficient in theequation without the maturity variables included is 0.30; with the maturityvariable (included in the results of Hypothesis 2, which showed that matu-rity is a mediator between candidate sex and competence) it is 0.21. Each ofthese coefficients is statistically significant. Hence, with respect to Hypothesis4 and 5, it appears that a similarity bias of the sort found by Lewis and Bierly(1990) was not present among our female subjects.

With respect to Hypothesis 6 (male subjects would be more likelyto judge women candidates as less competent than men candidates), ourresults suggest that, both before and after controlling for maturity, malesevaluated men candidates as more competent than women candidates,and the results were statistically significant.11 In specific, the candidate sex

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


TABLE 3 Regression Analysis of Competence by Female Subjects and Male Subjects

Females Males

Competence factor With maturity Without maturity With maturity Without maturity

Attractiveness 0.51∗∗∗ 0.56∗∗∗ 0.47∗∗∗ 0.48∗∗∗

(0.05) (0.05) (0.05) (0.06)Maturity 0.25∗∗∗ 0.26∗∗∗

(0.06) (0.06)Candidate sex 0.21∗∗∗ 0.30∗∗∗ 0.41∗∗∗ 0.47∗∗∗

(0.04) (0.04) (0.04) (0.04)Flag in picture 0.13∗∗∗ 0.15∗∗∗ 0.13∗∗∗ 0.12∗∗

(0.04) (0.04) (0.04) (0.04)Candidate age 0.002 0.01∗∗∗ −0.002 0.004∗∗∗

(0.02) (0.002) (0.002) (0.002)Constant 0.58 0.93 0.76 1.18

(0.21) (0.21) (0.22) (0.21)R-squared 0.60 0.55 0.65 0.59N 127 127 127 127

∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05.Note: Standard errors are shown in parentheses.Competence: Mean score of all subjects for each candidate based on 5-point Likert scale, 1 = “leastcompetent”; 5 = “most competent.” Maturity: Mean score of all subjects for each candidate based on5-point Likert scale, 1 = “least mature”; 5 = “most mature.” Candidate sex: 1 = male candidate; 0 =female candidate. Flag in picture: 1 if the background in the picture has a flag; 0 all others. Candidateage: Candidate age at time of election. Attractiveness: Mean score of all subjects for each candidate basedon 5-point Likert scale, 1 = “least attractive”; 5 = “most attractive.”

coefficient without the maturity variable is 0.47; with maturity it is 0.41.Further, a beta test indicates that the differences between male and female“voters” are statistically significant across equations.

To further investigate the effects of maturity, we performed Sobel tests.The results (not shown) confirm that maturity serves as a statistically signifi-cant mediator between sex of the candidate and competence among femalesubjects, but not among male subjects. On one hand, like male subjects,females judge men candidates’ faces as more competent than the faces ofwomen candidates, and their reactions are not based only on the sex ofthe candidate; maturity matters. On the other hand, male subjects judgemen candidates as more competent regardless of maturity levels of faces.12

These findings are consistent with the political science literature that sug-gests, although there are exceptions, females are generally no more likelythan males to vote for women candidates (see Dolan 2004).

The final of the three sets of hypotheses concerns the relationships ofcompetence, sex of the candidate, and sex of the subject to electoral victoryin the simulated and real elections. To determine the simulated electionwinners, we asked subjects for whom they would vote. This allowed us todetermine each candidate’s percentage of the vote. If this percentage wasover 50%, the candidate was said to have “won” the simulated election.Those under 50% were considered to have “lost” the election. Winners of

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


both simulated and real elections were coded dichotomously: 1 for winnersand 0 for losers.

Hypotheses 7 and 8 predicted, based on the Todorov team’s 2005 results,that candidates judged most competent would more likely to be the onessubjects voted for (or who won the simulated and true elections). To deter-mine the results for both elections, we performed two stages of analysis. First,following Todorov and colleagues (2005), we ran bivariate calculations. Andalthough the Todorov team did a limited multivariate analysis, we createdmultivariate equations that focused on both sex of the candidate and sex ofthe subject as well as on additional “image” variables discussed in detail inthis section.

The bivariate results are presented in the left-hand side of Table 4. Thesefigures are the percentage of time that candidates who scored higher onthe competence scale than their opponents were selected as winners in thesimulated and true elections. For the simulated election, those judged mostcompetent won 86.36% of the time among the total sample. Women can-didates who outscored their opponents on the competence scale won theirelections 76.47% of the time. Men candidates who outscored their oppo-nents won the simulated elections 89.80% of the time. Each chi-square resultfor these figures was statistically significant.13 For the total sample for thetrue election, candidates judged as most competent won 64.24% of the time.Among women candidates, those judged most competent won 76.47% ofthe time, and among men candidates, those judged most competent won73.47 of the time. Each chi-square result for these figures was statisticallysignificant, save the figure for men candidates.

To continue our focus on the ways in which maturity affects perceptionsof competence, we also present data in Table 5 that show the bivariaterelationships between evaluations of candidate maturity based on facialappearance and victory in the two types of elections. In the cases of judg-ments of the total group of candidates in both the simulated and true

TABLE 4 Percentage of Victory Based on Competence Comparisons between Competitorsand Candidate Sex

Competence

Candidates Sample election True election

All 86.36%(n = 57)

χ 2 = 69.82(p = 0.00)

64.24%(n = 49)

χ 2 = 16.33(p = 0.00)

Female 76.47%(n = 13)

χ 2 = 28.77(p = 0.00)

76.47%(n = 13)

χ 2 = 12.21(p = 0.00)

Male 89.80%(n = 44)

χ 2 = 32.70(p = 0.00)

73.47%(n = 36)

χ 2 = 1.40(p = 0.23)

Note: n represents number of cases where the candidate who scored higher than his or her opponent oncompetence was selected as the winner.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


TABLE 5 Percentage of Victory Based on Maturity Comparisons between Competitors andCandidate Sex

Maturity

Candidates Sample election True election

All 63.63%(n = 42)

χ 2 = 9.82(p = 0.00)

71.21%(n = 47)

χ 2 = 1.15(p = 0.01)

Female 38.10%(n = 8)

χ 2 = 1.82(p = 0.18)

57.14%(n = 12)

χ 2 = 3.37(p = 0.07)

Male 75.36%(n = 34)

χ 2 = 3.34(p = 0.07)

77.78%(n = 35)

χ 2 = 4.21(p = 0.04)

Note: n represents number of cases where the candidate who scored higher than his or her opponent onmaturity was selected as the winner.

elections, the more mature the candidate face, the more likely victory wasachieved. The effects are statistically significant and, in the true election,among all candidates, the percentage of victory is higher than is the casewhen competence is the independent variable. However, the relationshipsbetween maturity of facial appearance and victory among women candi-dates are not significant for either the simulated or true elections. For mencandidates, the effects are significant in the true election only and, interest-ingly, they are somewhat stronger than is the case when competence is theindependent variable examined. So far, the findings about elections indicatethat the results of the Todorov team’s (2005) study may be more nuancedthan their study demonstrated. In particular, sex of the subjects and of candi-date does appear to make a difference in how inferences about candidates’facial appearances affect voting. And it appears that men candidates areadvantaged in this regard.

Turning to multivariate analyses of correlates of victory in the simulatedand true elections, to establish a baseline, we modeled election outcomeswith and without the influence of competence and did so for the total sampleand separately for female and male subjects. We also controlled for anypossible effects of differences in backgrounds of photos, candidate attire,and the presence or absence of flags (which may be related to perceptions ofcompetence). We also controlled for the age of the candidates at the time ofthe election, perceptions of attractiveness of candidates, perceptions of facialmaturity, whether candidates were incumbents, and, following the Todorovteam (2005), whether they were racial minority or majority candidates (twominority candidates or mixed-race pairs). Logit analysis was the multivariatemodel used as vote choice in these analyses is dichotomous.

The results of multivariate analysis of the effects of competence on vic-tory in simulated and true elections are shown in Tables 6 and 7. Theyprovide further illumination of the complexity of the relationship betweenperceptions of competence and victory, and the effects of both candidateand subject sex on victory.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


TABLE 6 Logit Analysis of Winning the Simulated Election Based on Competence

Females Males

Competence factorWithout

competenceWith

competenceWithout

competenceWith

competence

Maturity 1.02 0.52 0.30 0.08(0.76) (0.82) (0.87) −1.27

Attractiveness 2.75∗∗∗ 1.84∗ 2.43∗∗ 1.98∗

(0.74) (0.90) (0.82) (1.04)Competence 1.97 0.86

(1.16) (1.27)Candidate sex 1.65∗∗ 1.27∗ 4.24∗∗∗ 3.86∗∗∗

(0.58) (0.62) (0.77) (0.93)Flag in picture 1.75∗∗∗ 1.52∗∗ 0.84 0.69

(0.53) (0.56) (0.61) (0.65)Candidate age −0.1 −0.02 −0.02 −0.02

(0.03) (0.03) (0.04) (0.04)Incumbency 1.20∗ 1.13∗ 0.36 0.36

(0.53) (0.53) (0.62) (0.62)Mixed race −0.18 −0.20 0.42 0.49

(0.70) (0.73) (0.79) (0.79)Formal setting in picture 0.07 0.17 0.63 0.68

(1.07) (1.14) (1.44) (1.49)Formal attire in picture −3.15 −3.45 −1.27 −1.21

(1.79) (1.88) (1.57) (1.55)Constant −8.43 −3.45 −8.05 −8.68

(3.24) (1.88) (3.62) (3.77)LR-chi-squared, df , p 57, 9, 0.00 60, 10, 0.00 77, 8, 0.00 77, 8, 0.00N 127 127 127 127Pseudo R-squared .33 .34 .44 .45

∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05.Note: Standard errors in parentheses.Simulated election: 1 if the candidate won the experimental sample election; 0 all others. Incumbent: 1 ifcandidate is an incumbent; 0 all others. Mixed race: 1 if candidates are both minority candidates and notof the same decent (e.g., an African American candidate and an Indian candidate); 0 all others. Formalsetting in picture: 1 if picture background is formal (e.g., formal office or Capitol); 0 all others. Formalattire in picture: 1 if candidate is dressed professionally (e.g., wearing a suit); 0 all others. Candidate sex:1 = male candidate; 0 = female candidate. Attractiveness: Mean score of all subjects for each candidatebased on 5-point Likert scale, 1 = “least attractive”; 5 = “most attractive.”

Table 6 demonstrates that adding controls to bivariate analyses rendersthe positive correlations between competence and victory nonsignificant.While the coefficients are in the correct direction, no statistically significanteffects appear among either female or male subjects (or among the totalpopulation, not shown here). Instead, sex of the candidate and attractivenessare the significant variables among both male and female subjects. Flags inpicture and incumbency are significant among female subjects.

One way to understand the differences in the effects of competencebetween our study and the Todorov team results is available by running ourequations without the candidate sex variable. Results of such an analysis

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


TABLE 7 Logit Analysis of Winning the True Election Based on Competence for Maleand Female Subjects

Female Males

Competence factorWith

competenceWithout

competenceWith

competenceWithout

competence

Maturity −0.65 1.08 1.09 1.53(1.39) (1.07) (1.51) (1.37)

Attractiveness −1.60 0.68 2.55 3.12∗

(1.59) (1.07) (1.52) (1.23)Competence 5.19∗ 1.35

(2.47) (2.24)Candidate sex −0.07 0.80 1.14 1.68

(1.05) (0.90) (1.30) (0.97)Flag in picture −0.40 0.19 0.01 0.20

(0.95) (0.86) (0.98) (0.90)Candidate age 0.07 0.05 0.09 0.09

(0.06) (0.05) (0.09) (0.06)Incumbency 5.59∗∗∗ 5.19∗∗∗ 5.93∗∗∗ 5.94∗∗∗

(0.89) (0.89) (1.10) (1.10)Mixed race −0.21 0.07 0.04 −0.01

(1.17) (1.13) (1.16) (1.15)Formal setting in picture 4.63∗ 3.35∗ 4.75 4.64∗

(2.06) (1.70) (1.96) (1.93)Formal attire in picture −2.92 −1.64 −2.25 −2.23

(1.95) (1.79) (1.93) (1.97)Constant −16.46 −11.90 −23.71 −22.63

(6.68) (1.79) (7.55) (7.30)LR-chi-sq, df , p 120, 10, 0.00 115, 9, 0.00 121, 10, 0.00 126, 9, 0.00N 127 127 126 126Pseudo R-squared .70 0.6 0.71 0.71

∗∗∗p < 0.001; ∗∗p < 0.01; ∗p < 0.05.Note: Standard errors are shown in parentheses.True election: 1 if the candidate won the true November election; 0 all others. Incumbent: 1 if candidateis an incumbent; 0 all others. Mixed race: 1 if candidates are both minority candidates and not of thesame decent (e.g., an African American candidate and an Indian candidate); 0 all others. Formal settingin picture: 1 if picture background is formal (e.g., formal office or Capitol); 0 all others. Formal attire inpicture: 1 if candidate is dressed professionally (e.g., wearing a suit); 0 all others. Candidate sex: 1 = malecandidate; 0 = female candidate. Attractiveness: Mean score of all respondents for each candidate basedon 5-point Likert scale, 1 = “least attractive”; 5 = “most attractive.”

(not shown) indicate that competence is significant in those equations.In addition, to continue to explore the effects of maturity in these anal-yses, we reran the regression equations without that variable. The results(not shown) indicate that the competence variable achieves statistical signif-icance (0.04) among evaluations by female subjects. Conversely, Sobel testswith maturity as a mediating variable between competence and victory showno significant mediating effect for male or female subjects.

The electoral results so far speak to the Todorov team’s (2005) results inseveral ways. First, several image variables that may be related to perceptions

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


of competence, such as formal candidate attire and flags in pictures.14 Suchvariables were not included in the Todorov team’s (2005) research. Second,the relationship among competence, maturity, and other similar traits iscomplex and inconstant. Most centrally, the Todorov team (2005) did notconsider the effects of subjects’ sex or candidates’ sex on perceptions ofcandidates’ competence or “image” and victory. Yet, in our study, thosevariables are key to understanding the results of both judgments of com-petence and victory in simulated elections. We find here, again, that gendergaps between female and male subjects exist, but that male candidates areadvantaged regardless.

The results of multivariate analysis in the true elections are presented inTable 7. In it, we display the analyses separately for female and male subjectswith and without the inclusion of the competence variable.

In the true election, the findings for competence largely but not com-pletely mirror the multivariate results of the simulated election. That is,evaluations of candidate competence inferred from facial appearance hadstatistically insignificant effects on electoral outcomes in all but one case.Among female subjects, competence was significant; that is, the more com-petent a candidate’s appearance, the more likely that candidate was toachieve electoral victory. Other relevant and significant associations in theseequations include incumbency and the formality of the setting.15 To bet-ter understand the differences in the effects of competence between ourstudy and the Todorov team’s (2005) results, we ran the same equation savefor the candidate sex variable. Results of that analysis (not shown) indicatethat, as was the case in the simulated election, competence is statisticallysignificant.

Here again, to understand the effects of maturity on victory, we ran theequations without that variable. The results mirror those in the equationswith maturity, but this time, among female subjects, competence is statis-tically significant. Finally, Sobel tests for possible mediation of maturity inthe relationship between competence and victory were significant amongfemales and males.

In sum, Hypotheses 7 and 8 are largely falsified. Once multivariatecontrols, especially the variable reflecting the sex of the candidate, areadded to tests of the relationship between competence and victory in sim-ulated and real elections, competence is rendered a statistically insignificantpredictor.

Moving from the competence/victory relationship to relationshipsbetween sex of candidates and victory, Hypothesis 9 predicted that femalesubjects would be more likely than male subjects to vote for women can-didates in both the simulated and true elections. As Table 6 indicates, inthe simulated election model with and without the competence variable,female subjects were more likely than male subjects to support women can-didates, and the gap between female and male subjects was large.16 Although

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


controlling for competence reduced this advantage, it remained strong andsignificant. Nevertheless, and critically, subjects of both sexes supported mencandidates. The gender gap is real, but it remains on the same side of thedividing line, a finding that echoes throughout a meaningful share of theissues-based gender-gap literature (Whitaker 2008).

With regard to the true election, although patterns similar to the sim-ulated election are present in the candidate sex coefficients among maleand female subjects, including a slight edge to women candidates amongfemale subjects, none of the results achieves statistical significance. To under-stand those patterns more clearly, we excluded the maturity variable fromthe equations. The results indicated that among male subjects, the sex ofthe candidate was significant and they favored men candidates. Finally, aSobel test of whether maturity mediated the relationship between candidatesex and victory and found maturity was significant among both female andmale subjects.

Hypothesis 10 predicted that in both elections women candidates wouldbe less electorally viable than men candidates, although we also predictedthat the effects of gender on viability would diminish when competencewas controlled. Indeed, as Table 6 shows, in the simulated election, theeffects of candidate sex were statistically significant and in the expecteddirection regardless of whether the full sample, the female sample, or themale sample was analyzed. Among male subjects, the predicted probabilityof men candidates winning was 0.84 and for women candidates was 0.17.Among female subjects, men candidates had a 0.70 predicted probabilityof winning, and women candidates had a 0.31 predicted probability. Betatests indicated that these differences were statistically significant. To fur-ther test these relationships, we ran Sobel tests for possible mediation ofcompetence in the relationship between candidate sex and victory, and pos-sible mediation of maturity in the relationship between candidate sex andvictory. We found that while competence mediates victory, maturity doesnot.

A final set of tests of Hypothesis 10 demonstrates that, although thishypothesis was confirmed for the simulated election, it is disproven forthe true one. As Table 7 indicates, the candidate sex coefficients are sta-tistically insignificant. To learn more about this finding, we once againperformed mediation tests—this time, we tested whether maturity mediatedbetween candidate sex and victory. Once again, maturity proved to serveas a mediator of the relationship in question—to a statistically significantdegree.

Overall, our findings resulting from testing 10 hypotheses indicate twothings to us: judging candidates based on facial appearance is more com-plex than might be supposed from initial analysis, and sociocultural attitudesabout male models of political leadership are pervasive and, especiallyamong male subjects, continue to disadvantage women in politics.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


DISCUSSION AND CONCLUSION

Overall, the results of this study reflect important qualifications to the2005 Todorov and colleagues study and highlight the challenges facingwomen in politics. Our results indicate that the relationship betweenperceptions of competence (or maturity) based on facial features is morecomplex and nuanced than might be suggested in initial inquiry. And,with respect to gendered perceptions of candidates for congressional office,men candidates appear to be advantaged. In brief, we find that, based onphotos:

● Subjects rate mature faces as more competent than baby faces, but theyrate women candidates’ faces as less mature and less competent than men’sfaces.

● Female subjects do not appear to engage in much “positive discrimination”toward women candidates. Although to a lesser degree than their malecounterparts, they judge men candidates to be both more mature andmore competent.

● Not only do male subjects judge women candidates more harshly than theyjudge men candidates but they also judge women candidates more harshlythan do female subjects.

● In simulated and true elections, both male and female subjects were morelikely to prefer men candidates. Nevertheless, as in judgments of compe-tence, there was a considerable gender gap in all instances: female subjectswere more likely than male subjects to vote for women candidates.

● In general, and unlike the results in Todorov team’s (2005) study, withimportant controls, perceptions of competence do not significantly affectsimulated or true election outcomes.

Even though our findings regarding the relationships among “image” vari-ables, sex of subjects, and sex of the candidates reflect much less firmconclusions about their effects on election outcomes than the Todorov team’sresearch suggested, they may still be somewhat discouraging for those whoprefer strictly ratiocinative and unbiased electoral decision making. Theylend some credence to the conclusion that citizens judge candidates, tosome degree, on affective factors presumably unrelated to their experiences,skills, accomplishments, issue positions, fundraising ability, and other cen-tral aspects of politics and governing such as party identification. This isnot to argue that these other more meaningful factors do not matter orthat once information about candidates beyond images is acquired, it doesnot affect judgments about candidates. However, affective perceptions maybe the foundation upon which cognitive evaluations are created. And the

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


cognitive elements to the structure (voting choices) may augment, but noteradicate or change, the direction of initial affective determinations.

The findings of this study may also be discouraging for those hoping toimprove representational diversity and legitimacy by increasing the numberof women in Congress and other elective offices. Although female subjectswere less likely to judge women candidates harshly and less likely to awardvotes to men candidates, at least in simulated elections, men candidateshad a decided advantage among voters of both sexes. Perhaps one ray ofhope is that female citizens vote at greater rates than men and are a largershare of the population. Those differences arrayed across electoral races andelection cycles may help account for the widespread finding in women andpolitics literature that women candidates win in equal proportions to men.Nevertheless, the findings provide substantial support for the “higher cost”theory of women’s electoral and representational efforts discussed in theintroduction (Lawless and Fox 2005; Thomas 2005).

Our findings also cast a different light on the experimental researchthat uses texts of speeches or other written materials to judge hypotheticalcandidates (Leeper 1991; Sapiro 1981–1982) and for issues of race or sexualorientation, see Sigelman et al. 1995; Herrick and Thomas 1999). The lim-itation of this type of research is that it ignores variations in appearances.The research presented here, along with the Todorov team’s (2005) study,both evaluate the effects of visual representations of candidates, and theyboth use real rather than imagined candidates. It may be that future experi-mental research could be improved by incorporating both types of candidaterepresentations.

The findings of this study are, on one level, illuminating; on anotherlevel, they leave many important questions unanswered. First, both our workand the Todorov team’s (2005) research on which it is based, assume thatvoters are, at least to some degree, aware of the appearances of the candi-dates they vote for in true elections. This is a reasonable assumption basednot only on the results of these studies but also on the fact that races forthe federal legislature often receive more media attention than do lowerlevel races. Even in congressional races that are not the most competitivecan expose voters to ads on local television, yard signs, direct mail litera-ture, Web sites, and Internet advertising, all with photos. Even if voters donot retain precise images of candidates, photos may create retained impres-sions. Nevertheless, research that directly measures the extent to which thisassumption is accurate is necessary.

Second, investigating the extent to which facial appearances and all theattendant variables that may reflect traits such as competence (such as theformality of the picture background and attire of candidates, and whetherthere is a flag in the picture) in primary elections is critical. It may be that,all else equal, only those women whose facial appearance is perceived ashighly competent tend to win primary contests.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


And, third, each of these studies was conducted at a single point in timeduring a political era in which foreign policy concerns dominated. Hence,the results may not be the same in a different political era—especially onein which economic concerns dominate. As always, additional research canaddress these related and important issues.

NOTES

1. These included competence, intelligence, leadership, honesty, trustworthiness, charisma, andlikability.

2. In an interesting variation of the facial appearance research, voters appear to prefer can-didates who look like them. Bailenson and colleagues (2006) found that a candidate’s photo thathas been morphed to look like the subject received more favorable ratings than an unmorphedphoto.

3. “Image” information means photos. Evaluatively “mixed” versus “consistent” refers to whetherboth evaluations are in the same direction (either positive or negative) as compared to one evaluationbeing positive and another negative. It is important to note that in this study, the authors speak ofappearance but do not comment upon facial structure per se.

4. Manipulations related to clothing, jewelry, and poses were also studied.5. The choice of using the baby-face/mature face designation follows the work of Berry and

McArthur (1985) and Zebrowitz and Montepare (2005).6. This is not to say that similarity bias is based solely on sex, only that sex may be one component

of similarity bias.7. Using college students in political experiments spurs the usual array of concerns regarding

representativeness. We did take care, however, to concentrate our search for subjects in Introduction toAmerican Government courses. Since all students in this large Midwestern university are required to takethe class, not just those who major in political science, we were able to appeal to a cross-section ofthe undergraduate population. Those who participated in our surveys were offered extra credit, and allparticipation was completed using a computerized program outside of the classroom.

8. One indicated lack of competence and five indicated a high degree of competence.9. The alpha scores for maturity, competence, and attractiveness are 0.94, 0.97, and 0.92,

respectively.10. Another way to analyze these relationships is as follows: for every 1 point increase in matu-

rity on the 1 to 5 scale, perceived competence of candidates increased 0.26 among male subjects and0.54 among female subjects. Among male subjects, candidates rated at the low end of the maturityscale (2.00) had a predicted competence score of 2.55. This competence score increased to 2.82 forcandidates rated midlevel on the maturity scale (3.08), and 3.09 for candidates rated at the high endof the maturity scale (4.14). For female subjects (column 1 of Table 3), candidates at the low end ofthe maturity scale (2.38) had a predicted competence score of 2.73. This increased to 2.94 for candi-dates in the middle of the maturity scale (3.23) and 3.15 for candidates at the high end of the maturityscale (4.07).

11. The mean candidate competence scores as rated by male and female subjects were 2.86 and2.93, respectively. The mean competence scores of men candidates and women candidates as ratedby male subjects were 3.03 and 2.64, respectively. The mean competence scores of men candidatesand women candidates as rated by female subjects were 3.02 for men candidates and 2.82 for womencandidates.

12. One other interesting relationship in the Table 3 equations is that among both female and malesubjects there was a positive and significant relationship between attractiveness and competence—and itwas present regardless of the presence of the maturity control.

13. Although not presented in Table 4, female subjects chose the most competent candidate 86.36%of the time, the most competent woman candidate 88.24% of the time, and the most competent mancandidate 85.17% of the time. Male subjects in our sample selected the most competent candidate 78.46%of the time, the most competent man candidate 93.75% of the time, and the most competent woman

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


candidate only 35.29% of the time. Each of chi-square results for these figures was statistically significant.The huge gap between female subjects’ choice of the most competent woman candidate 88.24% of thetime and male subjects’ choice of the most competent woman candidate 35.29% of the time is worthnoting.

14. Although the Todorov team (2005) performed multivariate analysis on their data, they limitedcontrols to judgments of age, attractiveness, face familiarity, and race.

15. We performed the same analyses for the total population, but those results are not shown.Relevant effects are discussed in our narrative.

16. This is true with and without the maturity variables included in the equations.

REFERENCES

Bailenson, Jeremy, Philip Garland, Shanto Iyengar, and Nick Yee. 2006. “TransformedFacial Similarity as a Political Cue: A Preliminary Investigation.” PoliticalPsychology 27(3): 373–85.

Berry, Diane S. 1991. “Attractive Faces Are Not All Created Equal: Joint Effects ofFacial Babyishness and Attractiveness on Social Perception.” Personality andSocial Psychology Bulletin 17(5): 523–31.

Berry, Diane S., and Leslie Z. McArthur. 1985. “Some Components and Consequencesof a Babyface.” Journal of Personality and Social Psychology 48(2): 312–23.

Berry, Diane S., and Leslie Zebrowitz-McArthur. 1988. “What’s in a Face: FacialMaturity and the Attribution of Legal Responsibility.” Personality and SocialPsychology Bulletin 14(1): 23–33.

Brownlow, Sheila. 1992. “Seeing is Believing—Facial Appearance, Credibility andAttitude Change.” Journal of Nonverbal Behavior 16(2): 101–115.

Brownlow, Sheila, and Leslie A. Zebrowitz. 1990. “Facial Appearance, Gender, andCredibility in Television Commercials.” Journal of Nonverbal Behavior 14(1):51–60.

Budesheim, Thomas L., and Stephen J. DePaola. 1994. “Beauty or the Beast? TheEffects of Appearance, Personality, and Issue Information on Evaluations ofPolitical Candidates.” Personality and Social Psychology Bulletin 20(4): 339–48.

Clark, Cal, and Janet Clark. 1999. “The Gender Gap in 1996: More Meaning Than aRevenge of the Soccer Moms.” In Women in Politics: Insiders or Outsiders?, ed.Lois Duke Whitaker. Upper Saddle River, NJ: Prentice-Hall, 68–84.

Davidson, Roger H., Walter J. Oleszek, and Frances Lee. 2008. Congress and ItsMembers. Washington, DC: CQ Press.

Dolan, Kathleen. 2004a. “The Impact of Candidate Sex on Evaluations of Candidatesfor the U.S. House of Representatives.” Social Science Quarterly 85(1): 206–17.

———. 2004b. Voting for Women: How the Public Evaluates Women Candidates.Boulder, CO: Westview Press.

Eagly, Alice H., and Linda L. Carli. 2003. “The Female Leadership Advantage: AnEvaluation of Evidence.” Leadership Quarterly 14(6): 807–34.

Fenno, Richard F., Jr. 1978. Home Style: House Members in Their Districts. Boston:Little, Brown.

Friedman, Heidi, and Leslie A. Zebrowitz. 1992. “The Contribution of Typical SexDifferences in Facial Maturity to Sex Role Stereotypes.” Personality and SocialPsychology Bulletin 18(4): 430–38.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


Herrick, Rebekah. 2001. “Gender Effects on Job Satisfaction in the House ofRepresentatives.” Women and Politics 23(1): 85–98.

Herrick, Rebekah, and Sue Thomas. 1999. “The Effects of Sexual Orientationon Citizen Perceptions of Candidate Viability.” In Gays and Lesbians in theDemocratic Process: Public Policy, Public Opinion, and Political Representation,eds. E. Riggle and B. Tadlock. New York: Columbia University Press, 170–91.

Huddy, Leonie, and Nayda Terkildsen. 1993a. “The Consequences of GenderStereotypes for Women Candidates at Different Levels and Types of Office.”Political Research Quarterly 46(3): 503–25.

———. 1993b. “Gender Stereotypes and Perceptions of Male and FemaleCandidates.” American Journal of Political Science 55(1): 119–47.

Jackson, Linda A., John E. Hunter, and Carole N. Hodge. 1995. “PhysicalAttractiveness and Intellectual Competence: A Meta-Analytic Review.” SocialPsychology Quarterly 58(2): 108–22.

Jacobson, Gary C. 2008. The Politics of Congressional Elections, 7th ed. New York:Longman.

Kahneman, Daniel. 2003. “A Perspective on Judgment and Choice: MappingBounded Rationality.” American Psychologist (September) 697–720.

Keating, Caroline F., David Randall, and Timothy Kendrick. 1999. “PresidentialPhysiognomies: Altered Images, Altered Perceptions.” Political Psychology 20(3):593–609.

Lawless, Jennifer, and Richard L. Fox. 2005. It Takes a Candidate: Why Women Don’tRun for Office. New York: Cambridge University Press.

Leeper, Mark S. 1991. “The Impact of Prejudice on Female Candidates: AnExperimental Look at Voter Inferences.” American Politics Quarterly 19(2):248–61.

Lewis, Kathryn E., and Margaret Bierly. 1990. “Toward a Profile of the Female Voter:Sex Differences in Perceived Physical Attractiveness and Competence of PoliticalCandidates.” Sex Roles 22(1): 1–12.

Mattes, Kyle, Michael Spezio, Hackjin Kim, Alexander Todorov, Ralph Adolphs, andR. Michael Alvarez. 2009. “Predicting Election Outcomes from Positive andNegative Trait Assessments of Candidate Images.” Political Psychology 31(1):41–58.

Poutvaara, Panu, Henrik Jordahl, and Niclas Berggren. 2009. “Faces of Politicians:Babyfacedness Predicts Inferred Competence But Not Electoral Success.”Journal of Experimental Social Psychology 45(6): 1132–35.

Rosenberg, Shawn W., Lisa Bohan, Patrick McCafferty, and Kevin Harris. 1986. “TheImage and the Vote: The Effect of Candidate Presentation on Voter Preference.”American Journal of Political Science 30(1): 108–27.

Rosenberg, Shawn W., Shalmit Kahn, and Thuy Tran. 1991. “Creating a PoliticalImage: Shaping Appearance and Manipulating the Vote.” Political Behavior13(4): 345–67.

Rosenberg, Shawn. W., and Patrick McCafferty. 1987. “The Image and the Vote:Manipulating Voters’ Preferences.” Public Opinion Quarterly 51(1): 31–47.

Rosenwasser, Shirley M., and Norma G. Dean. 1989. “Gender Role and PoliticalOffice: Effects of Perceived Masculinity/Femininity of Candidate and PoliticalOffice.” Psychology of Women Quarterly 13(1): 77–85.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2


Sanbonmatsu, Kira. 2002. “Gender Stereotypes and Vote Choice.” American Journalof Political Science 64(1): 20–34.

Sapiro, Virginia. 1981–1982. “If U.S. Senator Baker Were a Woman: An Experimentalstudy of Candidate Image.” Political Psychology 3(1): 61–83.

Sczesny, Sabine, and Ulrich Kuhnen. 2004. “Meta-Cognition About Biological andGender-Stereotypic Physical Appearance: Consequences for the Assessmentof Leadership Competence.” Personality and Social Psychology Bulletin 30(1):13–21.

Sigelman, Carol K., Lee Sigelman, Barbara J. Walkosz, and Michael Nitz. 1995. “BlackCandidates, White Voters: Understanding Racial Bias in Political Perceptions.”American Journal of Political Science 39(1): 243–265.

Sigelman, Lee, Carol K. Sigelman, and Christopher Fowler. 1987. “A Bird of aDifferent Feather? An Experimental Investigation of Physical Attractiveness andthe Electability of Female Candidates.” Social Psychology Quarterly 50(1): 32–43.

Thomas, Sue. 2002. “The Personal Is the Political: Antecedents of Gendered Choicesof Elected Representatives.” Sex Roles: A Journal of Research 47(7–8): 343–53.

———. 2005. “Cracking the Glass Ceiling: The Status, Significance, and Prospects ofWomen in Legislative Office.” In Gender and American Politics: Women, Men,and the Political Process, 2nd ed., eds. S. Tolleson-Rinehart and J. Josephson.Armonk, NY: M. E. Sharpe, 242–63.

Todorov, Alexander, Anesu N. Mandisodza, Amir Goren, and Crystal C. Hall. 2005.“Inferences of Competence from Faces Predict Election Outcomes.” Science308(10): 1623–26.

Whitaker, Lois D. 2008. Voting the Gender Gap. Urbana: University of Illinois Press.Zebrowitz, Leslie A. 1997. Reading Faces: Window to the Soul? Boulder, CO:

Westview Press.Zebrowitz, Leslie A., and Susan M. McDonald. 1991. “The Impact of Litigants’ Baby-

Facedness and Attractiveness on Adjudications in Small Claims Courts.” Law andHuman Behavior 15: 603–23.

Zebrowitz, Leslie A., and Joann M. Montepare. 2005. “Appearance DOES Matter.”Science 308(6): 1565–66.

Dow

nloa

ded

by [

sue

thom

as]

at 1

7:12

15

May

201

2

Gender and Perceptions of Candidate Competency

Documents