Research on Gender in Political Psychology Conference
Candidate Traits, Gender Roles, and the 2008 Vote
Susan B. Hansen, Kris Kanthak, and Jennifer N. Victor
University of Pittsburgh
Abstract: Candidates’ personal and leadership qualities are important predictors of voters’
choices. This paper uses new questions from the 2008 American National Election Studies to examine voters’ preferred candidate traits and their support of equality for women. Gender
differences in perceptions of candidate traits, especially morality, were more pronounced in 2008 than in earlier years. Barack Obama and John McCain were also perceived to differ considerably in their support for equal roles for women, more so by African-Americans and Hispanics than whites. These perceptions, along with respondents’ own support for women’s equality, had a significant impact on both vote choice and assessments of candidate morality in 2008 -- but much more so for males than females.
Research on voting behavior has consistently found that for most people, evaluations of candidates’ personal and leadership qualities have more impact on vote choice than do
candidates’ positions on issues (Miller and Shanks 1996; Hayes 2005, Saad 2006; Prysby and Holian 2007; Lewis-Beck, et al. 2008). Acquiring information about candidates’ stands on
specific issues can be costly and time-consuming. But it is less cognitively demanding for voters to form impressions of candidates’ characters based on a 30-second campaign ad, television
news, or personal appearances (Domke, Shah, and Wackman 2000). Voters may also rely on judgments about character and leadership qualities since these would presumably affect future, as well as past or current, policy decisions.
Since the 1980s American National Election Studies (ANES) surveys have asked respondents to evaluate the presidential candidates in terms of traits such as leadership, compassion (“cares about people like you”), honesty, and morality. These evaluations are of
course highly partisan, but even with controls for strength of partisanship, character traits have a
significant independent effect on the vote (Hayes 2005). Most previous ANES surveys have shown that men were more likely to perceive candidate differences in strong leadership, and women to perceive differences in compassion. But these perceptions were largely based on sex differences in partisanship; gender apparently had little independent impact on vote choice (Hansen and Wills-Otero 2006).
As we will show, however, in 2008 gender had a significant impact on candidate evaluations as well as vote choice. Women were considerably more likely than men to vote on the basis of their perceptions of candidate differences in compassion, honesty, optimism, and strength of leadership. Much more than in previous elections, perceived candidate differences in “morality” were important for women voters in 2008 – even more so than in 2004, when “moral
1values” were arguably a salient factor in the reelection of George Bush. What accounts for the
saliency of “morality” and the increased gender differences in trait perceptions in 2008?
We argue that in 2008, perceptions of the candidates were significantly affected by respondents’ views of the role of women in society. Moreover, we argue that there are
significant gender differences in the ways in which voters evaluated candidates in 2008 with respect to the ways in which voters evaluated Obama and McCain’s individual morality. Furthermore, voters’ evaluation of the candidates’ morality is strongly affected by voters’ views of gender roles. To be more specific, we find that men who strongly believe in the equality of
the sexes and who view McCain as more moral than Obama, vote for McCain; but the same men who see Obama as more moral vote for Obama. We use these findings to show that contrary to much of the existing literature there are significant gender effects in the 2008 election, when accounting for voters’ perceptions of candidates’ morality and voters’ personal views of gender
The next section of this paper will consider the impact of gender and candidate traits in 2008 compared with previous elections. We then analyze indicators of attitudes towards the role of women: time-series data on the 7-point equal roles scale, perceptions of candidates’ positions
on this scale, and new questions about equality for women added in 2008. In the third section we show that perceived candidate differences on the equal-roles scale had a significant impact on vote choice in 2008, even when partisanship, sex, race, and candidate differences on other traits (leadership, compassion, and morality) were considered. Finally, we compare how male and female respondents weighed perceived candidate differences in their voting decisions.
1. Gender, Candidate Trait Perceptions, and Vote Choice, 2000-2008
ANES surveys since 1980 have included questions on perceptions of candidate traits: whether the candidate is a strong leader, moral, honest, intelligent, knowledgeable, or compassionate (“cares about people like you”). Perceived candidate differences on two of these
traits (strong leadership and compassion) have been consistently shown to be strong predictors of the Presidential vote, even with controls for party identification; morality had some impact only in 1992 and 2004 (Hayes 2005; Hansen and Wills-Otero 2006). How did Barack Obama and John McCain compare on these trait perceptions in 2008?
The 2008 survey differed in two respects from previous studies. First of all, perceptions of a new trait, candidate optimism, were assessed, and this proved to be significant to some voters. Second, two different indicators of candidate traits were used in the 2008 survey. Half the respondents were asked the trait questions using the same format of four responses found in earlier ANES election studies: whether a particular trait applied to a candidate “Extremely well, quite well, not too well, or not well at all.” The rest of the ANES respondents were asked to use
a five-point rating for each trait, also ranging from “extremely well” to “not well at all” but now including a middle category - “moderately well.” Note that respondents were not asked to
compare the candidates directly but to rate each separately on the individual traits.
Table 1 shows perceived differences in the seven candidate traits assessed in 2008. Each response for Obama was subtracted from the corresponding response for McCain, so that higher scores indicate a Democratic advantage on a particular trait. Using the five-point scales produced a slightly smaller proportion of “no difference” assessments on these traits (28 versus 33 percent) and a more limited range of responses. Both question formats gave Barack Obama a significant edge on most of the traits. McCain’s best showing was on the “strong leadership” trait, but even there he lagged significantly behind Obama based on the four-point scale and essentially tied him using the five-point scale.
It is striking that the four-point scale comparisons found only nine percent of respondents (overwhelmingly Republican partisans) more likely to feel that “John McCain cares about people like you,” and only 12 percent judged him to be more intelligent that Obama (Table 1). Sizeable differences in traits also appeared based on the question new in 2008: whether the candidate was optimistic. Based on the 5-point scales, 28 percent of respondents saw no difference, but 55 percent thought that Obama was more optimistic, compared with only 17 percent who regarded McCain as more optimistic. Only the “cares about you” trait gave Obama a stronger edge than
the optimism trait, regardless of the question format.
We might expect some gender differences in perceptions of candidate traits in 2008. Previous ANES surveys found that males tend to see larger differences between candidates in terms of strong leadership, and females in terms of compassion (Hansen and Wills-Otero 2006).
We would also expect women (who are more likely to be Democrats) to view the Democratic candidate’s traits more favorably in 2008 as they had in previous years.
The expected gender differences in trait perceptions do appear (Table 2), and are especially striking based on the 5-point scales used in 2008. Since 1980, perceived candidate differences on strong leadership have usually been negative, indicating a net Republican advantage especially among men, but in 2008 both males and females rated Obama as the stronger leader. The positive mean differences on “cares about you” show that this trait has consistently favored the Democratic candidate, considerably so in 2008 and particularly among women. The mean differences for candidate morality have historically favored the Republican candidate, but in 2008 Obama was rated as somewhat higher in morality by both men and women. Democratic candidates (especially Al Gore and John Kerry) were perceived as more intelligent and knowledgeable, but Obama had only a slight edge on these dimensions in 2008
2and gender differences were not significant. Women were significantly more likely than men to
perceive Obama as more honest and optimistic than McCain.
Which traits best predict the vote for President, and how did 2008 compare with previous years? Question format clearly makes a difference - and complicates any assessment of the relative importance of candidate traits. Table 3 shows the results of a logistic regression of 2008
3vote choice (0=Obama, 1=McCain) using both question formats and perceived candidate
differences on the three most important traits (compassion, strong leadership, and morality) identified in an earlier analysis (Hansen and Wills-Otero 2006). Party identification is included as well, as voting studies have consistently found it to be the single best predictor of voting behavior (Miller and Shanks 1996; Kaufmann, Petrocik, and Shaw 2008). Although assessment of candidate traits is heavily influenced by party, they have been found to have independent
effects (Hayes 2005). For purpose of comparison, similar models for 2000 and 2004 are also shown. Respondent sex was not significant in either of those years, but was expected to be more important in 2008 given the sizeable gender differences in trait perceptions noted above.
The logistic regression results show that, as expected, partisanship was the dominant predictor of vote choice. As in 2000 and 2004, sex had no significant independent impact using the four-point scales, but was close to significance based on the five-point scales. Use of the earlier (4-point) trait question format in 2008 produced results comparable to those for 2000: strong leadership as the dominant trait, but with morality slightly more important than compassion as a predictor of the vote in 2008. However, if the new format based on a 5-point scale is used, results parallel those for 2004, with a candidate’s perceived edge in compassion the
trait most important to voters. As in 2000, perceived differences in the morality trait were significant predictors of vote choice in 2008 – even more so than strong leadership, based on
analysis using the 5-point scales.
2. Gender Role Perceptions and Candidate Assessments
What accounts for the increased salience of gender and the “morality” candidate trait in
2008? Fortunately for our purposes, the 2008 ANES included a number of new questions concerning respondents’ views of gender roles. Respondents were also asked to place both
McCain and Obama on the 7-point ANES scale that ranges from 1, “Men and women should
4 have equal roles,” to 7, “Woman’s place is in the home.”
The equal-roles scale has been asked of ANES respondents since 1972. Figure 1 shows dramatic changes over time in the average responses by males and females, both white and African-American. In the 1970s white males were actually more likely than white females to
favor equal roles. All four groups have become more supportive of equal roles for women, although African-American males to a lesser degree than others.
As Table 4 shows, males were slightly less likely than females to select the equal-roles position on the 7-point scale (asked of only half the sample in 2008). But perceptions of the candidates’ positions differed considerably by race/ethnicity and gender (and no doubt by party as well). A sizeable proportion of white respondents (both male and female) gave Obama and McCain identical placements, but African-Americans (especially women) and Hispanics rated Obama as considerably more supportive than McCain of equal roles for women.
The 2008 survey included several new questions concerning the status of women: whether women were losing out on good jobs because of sex discrimination, whether sexual harassment complaints cause more trouble for women in the workplace, whether women are seeking equality or asking for “special favors.” Respondents were also asked whether they did or
did not hope for a woman as president. Table 4 shows negligible sex differences in responses to the questions as to whether women faced job discrimination. But men were considerably more likely to perceive women as seeking “special favors” rather than equality, and to agree that
claims of sex discrimination would “cause more problems than they solve.” Fewer males than
females agreed that they hoped for a woman as U. S. president.
Racial and ethnic differences in views of gender roles are also evident in Table 4. Hispanics are generally less supportive of gender equality, reflecting both Roman Catholic religious backgrounds and immigration from countries where machismo is a cultural norm. But gender differences are considerable for Hispanics as well, although the gender-role attitudes of younger Hispanics are converging on U. S. mainstream cultural norms (Su, Richardson, and Wang 2010). African-Americans are especially aware of job discrimination. Both African-
Americans and Hispanics are more likely than Whites to see sex-harassment claims as problematic, perhaps reflecting bitter personal experiences. African-American and Hispanic females are considerably more likely than their male counterparts to disagree that demands for equality involve requests for “special favors.”
Both gender and race/ethnicity influence whether or not respondents agree that “I hope
the U. S. has a woman president in my lifetime.” Only 31 percent of white males agree, compared with 50 percent of white females, 57 percent of African-American females, and 60 percent of Hispanic females. These responses are somewhat more pessimistic compared with the trend in a Gallup Poll question as to whether one would vote for a “qualified” woman candidate if she were nominated by one’s party. Over 90 percent of the public now says they would vote
2for a woman, with women and minorities only slightly more likely to say so. In 2008, of course,
the actual candidacies of Hillary Clinton and Barack Obama (and Clinton’s loss) no doubt
influenced the impact of party and race on responses to the ANES question.
A scale of support for equality for women was constructed based on a factor analysis of four questions listed in Table 4: a woman for president, job discrimination, sexual harassment, and women’s demands for equality vs. “special favors (the 7-point Equal Roles question was not
included because it was asked of only half the sample). All items loaded on a factor accounting for 47 percent of the variance, with higher scores equal greater support for equal roles for women. The last entry in Table 4 indicate considerable differences by gender, race, and ethnicity in mean scores on this scale. Women have consistently higher scores than men. White women show the greatest support for equal roles, scoring significantly higher than white men. They also score higher than African-American females based on these ANES questions, although other studies have found stronger support for gender equity among African-American females (Carter, Corra,
and Carter 2009). Hispanics (both male and female) are least supportive of equality for women based on this indicator.
3. Assessments of Candidate Morality
Morality assessments appeared to be important to voters in 2008, and we expected those assessments of candidate’s morality to be based on two factors: assessment of the candidate’s
personal conduct and the candidates’ positions on “values” issues. With respect to the
candidates’ personal conduct, evidence of truly egregious misconduct can quickly drive a
politician out of public office; examples such as Elliot Spitzer, Mark Sanford, and John Edwards come to mind. But assessments of lesser faults are likely to be highly partisan and weighed against other factors (strength of leadership, policy outcomes). Thus Democrats were far more likely than Republicans to forgive Bill Clinton’s sexual dalliances. In both 1992 and 1996 they
5considered him to be less “moral” than his opponent, but voted for him anyway. In 2008 John
McCain’s divorce and remarriage were minor issues, but Republicans still rated him higher on
“morality” than Barack Obama. In fact both candidates were viewed favorably; 76 percent
thought that the “morality” trait described Obama well or extremely well, and 62 percent said the same about McCain.
Another likely source of morality assessments is a candidate’s stance on the “values”
issues that have been so highly salient in American politics since the 1970s, specifically abortion, gay rights, and the status of women (Hunter 1991). Of course these assessments are likely to be partisan as well, since supporters of the Religious Right and traditional values have become more closely identified with the Republican Party, while seculars, gays, and feminists are strongly identified with the Democrats (Layman 2001).
We would also expected several other groups to be especially attentive to values issues and to rely on them to assess candidate morality: frequent church-goers, people with evangelical or fundamentalist religious views (based on those who claimed to be “born again”),
the less-educated, older Americans, opponents of gay rights or abortion rights, and respondents with lower scores on the women’s-equality factor scale described above. As a preliminary test of these conjectures, Pearson correlations were calculated between these factors and the 4- and 5-point scales assessing differences in perceptions of candidate morality.
Table 5 shows strong correlations between candidate morality perceptions and both partisanship and ideology, but surprisingly, Democrats and liberals perceive sharper differences. However, correlations were negligible between the 4- or 5-point morality-difference measures and education, church attendance, or being born-again. Women, younger respondents, and pro-choice respondents tended to perceive slightly greater differences. But strong correlations with perceived morality differences emerge for perceptions of greater candidate differences in support for women’s equality. Correlations are also significantly higher with higher scores on the
gender-equality scale and stronger support of gays/lesbians (based on thermometer scores, gay adoption, gay anti-discrimination laws, or gay marriage). Thus contrary to our expectations, younger people and those with socially liberal rather than conservative views of gender roles, reproductive rights, and homosexuality saw larger differences in candidate morality; religion or religiosity had very little effect.
Furthermore, perceived candidate differences on this equal-roles scale had a significant impact on vote choice in 2008, even when partisanship, sex, race, and candidate differences on other traits (leadership, compassion, and morality) were considered (Table 6). Logistic regression was again used since the dependent variable was a dichotomy (1=Obama, 2=McCain). The