Sex, race, gender, and the presidential vote

Abstract Racial resentment has been shown to have a significant impact on voting by whites in recent presidential elections, and a much larger impact than the traditional gender-gap measure based on the male-female dichotomy. This analysis will use data from the American National Election Studies [ANES] to compare broader indicators of race and gender applicable to the Democratic and Republican parties as well as to respondents’ opinions of appropriate roles for women. Since the 1980s the parties have diverged considerably on abortion and women’s issues, and voters now view the Democrats as more supportive than Republicans of equality for women and reproductive rights. Perceptions of party differences on women’s issues strongly influenced vote choice, 1988–2008, and in 2008 had greater impact on whites’ votes than opinions on aid to blacks, abortion, gay marriage, or the economy. Although racial resentment was a strong predictor of the white vote in 2012 as in previous years, presidential voting was also significantly influenced by respondent sex as well as opinions on gender roles. Voters regarded the Democratic Party as “better for the interests of women,” and this proved to be a highly effective wedge issue for the Democrats in 2012.


Introduction
Relatively few whites now respond to overtly racist appeals or support demeaning stereotypes of African-Americans (Mendelberg, 2008). Yet a sizeable proportion of whites believe that African-Americans "don't try hard enough," are prone to crime, and benefit disproportionately from government policies such as welfare or affirmative action. Kinder and Dale-Riddle (2012) [hereafter KDR] show that such "racial resentment" has been a significant predictor of white votes for President since the 1980s. But KDR claim that "gender" (based only on the male/female dichotomy) had much less impact than racial resentment for voting by whites in presidential elections or in the 2008 Democratic primaries. They argue that "gender," unlike race, has not become linked to the political parties, and males and females do not exhibit the group solidarity that has characterized African-Americans' opinions and voting behavior.
However, other research, using much broader conceptions of gender, challenges the KDR conclusions. Both men and women have a range of opinions on appropriate gender roles at home, in politics, or in the workplace. Furthermore, since the 1970s, the Democratic and Republican parties have diverged considerably in their positions on policies concerning women, including reproductive rights, the Equal Rights Amendment, equal pay, child care, women in combat, and family leave (Freeman, 2000;Sanbonmatsu, 2000;Wolbrecht, 2000). Voters describe the Democratic and Republican parties in highly gendered terms (Winter, 2010), and the entry of women into party politics has influenced partisan rhetoric, agendas, and campaign strategy (Freeman, 2000;Lovenduski & Norris, 1993). Hansen (2014Hansen ( , 2016 found that perceptions of candidate differences in support for women's equality have been consistent and substantial predictors of the vote since the 1970s. In order to test the KDR claim that "race trumps gender," this paper will use American National Election Studies [ANES] data to analyze the electoral impact of different indicators of both race and gender. 1 It will show that voters are aware of increasing party divergence on issues, including aid to blacks, abortion, and equality for women. Perceptions of candidate differences on race and women's equality (but not abortion) have influenced presidential vote choice. This analysis will also include the 2012 election; Barack Obama's candidacy heightened the salience of race, and the prominence of gender issues in both 2008 and 2012 provides a test case for the KDR claims.
The next section of this paper will document trends over time in ANES respondents' own views of gender roles, abortion, and aid to blacks, and their perceptions of candidate positions on these issues. The second section will expand the KDR analysis to test for the relative impact of racial resentment, respondent sex, and gender roles on presidential voting by whites, 1988-2008. The third section will compare these factors with the impact of the economy and other issues in the 2008 and 2012 elections. Although the equal-roles questions were not included in the 2012 ANES surveys, a new question asked which party was "better for the interests of women." A majority of voters thought that the Democrats were, and the fourth section will show that this was a more effective wedge issue for the Democrats than opinions on the economy, foreign policy, Obamacare, gay marriage, or abortion. Race and racial resentment were evident in 2012, as in previous elections, but sex and gender also influenced the vote.

Perceptions of party positions on equal roles, abortion, and aid to blacks
The second women's movement that emerged in the 1960s challenged traditional conceptions of gender roles and the laws that had denied women equal status. Many more women entered the labor force, higher education, and the professions, and birth rates declined after legalization of the birth control pill and abortion. Support for equality for women has increased considerably since 1970. Gallup Poll data show that over 95% of Americans are now willing to vote for a woman as president. 2 In most elections since 1972, ANES respondents have been asked to place themselves on a seven-point scale, where 1 is "Men and women should be equal" and 7 is "women's place is in the home." Scores of 1 or 2 on this scale increased from 42% in 1972 to over 80% by 2012. Both males and females, and Democrats as well as Republicans, have moved considerably closer to the equalroles position since 1972.
Trends in opinions on abortion vary considerably depending on the questions asked (Jelen & Wilcox, 2003;Rose, 2005). Slightly more Americans consider themselves to be pro-choice rather than pro-life (Saad, 2015). Yet support for the Roe decision increased from 55% in 1992 to over 70% in 2013 (Radnofsky & Ashby, 2013). General Social Survey [GSS] data show an increase in the proportion of Americans who would permit a woman to have an abortion "for any reason," from 32% in 1977 to 45% by 2010. ANES surveys include a four-point scale, ranging from 1, "By law, abortion should never be permitted," to 4, "By law, abortion should be a woman's choice." Pro-choice responses were 25% in 1980 but 47% in 2012.
Expressions of overt racism in surveys have declined considerably since the 1970s, although evidence of symbolic racism and racial resentment suggest that race is still very much an issue in the US. 3 ANES surveys since 1972 have asked respondents to place themselves on a seven-point scale where 1 is "Government should make every effort to improve blacks' social and economic position," and 7 is "Government should not make any special effort to help blacks; they should help themselves." Mean responses to this scale show that white Americans are highly skeptical of "aid to blacks," but African-Americans are more supportive. 4 The opinions of Democratic partisans have changed little since the 1970s, but Republican partisans have moved closer to the "blacks should help themselves" position. The KDR racial resentment scale is based on several ANES questions concerning African-Americans. 5 These questions have only been asked since 1988, and scale values were stable until a slight uptick in 2008.
Party platforms and campaign rhetoric show that the parties have diverged on race and social issues since 1980. Kinder and Dale-Riddle (2012) have documented the parties' evolving positions on welfare, affirmative action, crime, and voting rights, and the GOP "Southern strategy" proved to be highly successful in moving Southern Democrats into the Republican Party (Hillygus & Shields, 2008). African-Americans vote overwhelmingly for Democrats, and have fared much better in terms of income, poverty rates, and unemployment under recent Democratic rather than Republican presidents (Hajnal & Horowitz, 2014).
Since the 1980s Democratic platforms have increasingly favored gay rights and as of 2012 supported gay marriage. Republican platforms remain opposed to gay marriage, gays serving openly in the military, and to any "special rights" for gays and lesbians. In 1976, the Republican Party platform supported both the Equal Rights Amendment and Roe v. Wade. But these positions reversed after 1980; the Religious Right gained influence within the GOP, and by 2000 very few pro-choice Republicans could be found in Congress (Layman, 2001). Democratic platforms have emphasized support for reproductive rights, family leave, child care, health care, and equal pay for women (Hansen, 2014).
How aware is the electorate of these evolving party positions? The proportion of ANES respondents perceiving "no important party differences in what the Democratic and Republican parties stand for" declined from 44% in 1972 to only 23% in 2008 and 13% in 2012. Voters are more likely than nonvoters to say that the parties differ significantly. The polarized parties are thus providing much clearer cues to voters-especially partisans (Hetherington, 2001).
Beginning in 1972 most ANES surveys have asked respondents to place the parties and presidential candidates on the same seven-point scales used to assess respondents' own opinions on equal roles for women, aid to blacks, and several other issues. Although ANES has never assessed perceptions of party positions on gay rights or gay marriage, in several elections since 1980 ANES respondents have also been asked to place the Democratic and Republican parties and candidates on the same four-point abortion scale described above. Thus we can track trends over time in both voters' own views and their perceptions of party and candidate positions.
Although voters are more likely than non-voters to perceive party issue differences, only about half of voters meet all the criteria for issue voting: having an opinion on an issue, awareness of party positions, and perceptions of differences in party or candidate stances on that issue (Abramson, Aldrich, & Rhode, 2012, p. 156). However, as the parties have polarized in recent elections, fewer respondents have been unable to place one or both candidates on the seven-point scales ("DK" responses) or have made "wrong" responses (assigning the more liberal position to the Republican rather than the Democrat). 6 In 1980, as party positions on abortion were still evolving, over 40% of ANES respondents could not place either Carter or Reagan, and 31% gave the candidates equal placements on the 4-point scale. But by 2012, the don't know proportion was only 2-3%, only 7% of respondents saw no differences between Obama and Romney on abortion, and only 7% considered Romney to be more pro-choice than Obama. 7 Two issues linked to gender show evidence of both party polarization and divergence between partisans' views and party positions. Figure 1 shows the trend since 1980 in perceptions by all voters of the presidential candidates' positions on abortion (the ANES four-point scale), compared with the positions of Democratic and Republican party identifiers. The mean responses of these partisans have diverged considerably over this time period; the abortion issue may well have contributed to party realignment (Jelen & Wilcox, 2003;Layman, 2001;Sanbonmatsu, 2000). Democratic candidates are placed somewhat closer to Democratic voters, but the Republican and Democratic presidential candidates are perceived as even further apart since the 1980s.   Figure 2 displays the trends over time in voters' perceptions of the presidential candidates' positions on equality for women, as well as the views of Democratic and Republican partisans. Voters perceive considerable differences between the candidates, although recent Republican candidates are placed considerably closer to the "equal roles" position than in the Reagan era. Since 2000 both Democrat and Republican partisans strongly support equality for women.
Because of these party differences, a considerable "gender gap" in voting emerged in 1980 and has been evident in most subsequent elections. It should be noted, however, that male/female differences in attitudes and behavior are not based on innate or biological differences between the sexes, but reflect their very different situations with respect to the market economy, responsibility for child-rearing, the welfare state, and religiosity, as well as different attitudes concerning gay rights and violence (Deckman & McTague, 2014;Kaufmann, 2004). Although more women than men vote Democratic, males and females are divided along so many other dimensions (marital status, race, ethnicity, income, and religiosity) that the "gender gap" is much smaller than the racial gap in voting. Women are less likely than African-Americans to express group solidarity, and are not as physically isolated from men as African-Americans often are from whites.
KDR argue that race issues carry so much weight because they have become more closely linked to parties. Based on an ANES seven-point scale question concerning "aid to blacks," this may well be the case. Figure 3 shows the trend since 1972 in perceptions by all voters of the presidential candidates' positions on aid to blacks, compared with the self-placements of Democratic and Republican party identifiers. The opinions of Democratic partisans have changed little since the 1970s, but Republican partisans have moved closer to the "blacks should help themselves" position. The presidential candidates have been placed even further apart on this issue in recent years, especially in 2012. Here the Democratic Party may be out of step, since voters place Democratic candidates considerably closer to the "government should assist blacks" end of the seven-point scale than even Democratic identifiers position themselves.

Party differences, racial resentment, and the presidential vote
These ANES data suggest that much of the electorate perceives increased divergence between the Democratic and Republican parties on aid to blacks, abortion, and equality for women. As Winter (2010) has shown, ANES respondents describe "masculine" Republicans and "feminine" Democrats in very different gendered terms. But Winter's analysis did not link party perceptions to vote choice. Based on Hansen's (2014) analysis, perceived candidate differences on government assistance to blacks have significantly influenced presidential choice in several recent elections. Perceptions of candidate differences in support for women's equality have also been consistent and substantial predictors of the vote since the 1970s. However, perceptions of candidate differences on abortion never significantly impacted the vote in the years in which ANES included this questions. 8  Hansen's analysis also found race (white/non-white) to be a strong predictor of the presidential vote in elections since 1972, but she did not consider racial resentment. The KDR analysis of the presidential vote, 1988-2008, found racial resentment to be a highly significant predictor of whites' votes, but failed to include either respondent sex or any other indicators of gender (2012,105,232). (KDR did examine opinions on appropriate gender roles, but only for male voters in the 2008 Democratic primaries, where they apparently had little effect).
What happens when BOTH race and gender are considered? My initial analysis will add three indicators of gender to the basic KDR model predicting votes for the Republican candidate by non-Hispanic whites, 1988-2008. First is an indicator variable for sex (1 = male, 2 = female), the usual "gender gap" measure. A second indicator of gender will be respondents' attitudes toward how males and females should behave, based on their positions on the ANES seven-point equal-roles scale. Respondents with more traditional views of women should be more likely to vote for Republicans.
The third indicator will be perceptions of candidate differences on equality for women. ANES does not ask respondents to compare the candidates directly on specific issues, but the Republican and Democratic candidate placement questions are juxtaposed in the surveys (the order has been randomized in more recent surveys), thus encouraging implicit comparisons. The placement of the Democratic candidate on the equal-roles scale will be subtracted from the Republican placement. The scale thus ranges from −6 to +6; respondents who assign the candidates identical positions on the seven-point scale are coded 0, as perceiving "no difference." Higher values represent voters' perception of the Republican candidate as more supportive of traditional roles. Table 1 presents the results of a logistic regression of whites ' votes, 1988-2008 (1992 was omitted from this analysis because the equal-roles placements were not asked, and the 1996 ANES did not include the racial resentment questions). Racial attitudes are based on the KDR racial resentment scale. The KDR models also include party identification (1-strong Democrat … 7-strong Republican) and indicator variables for religious identity (Catholic, Jewish, other/none, with Protestants as the omitted category). Following KDR, indicator variables for Census regions and levels of education were also included in these regressions; the resulting coefficients were not significant and are not shown in Table 1.
The results indicate that party identification is the strongest predictor of the vote, as voting studies have consistently shown (Bartels, 2010, p. 11). Except in 1988, racial resentment is still significant for whites' votes. Religious identity had little impact, other than for Jews in 2008 (Jews had to be excluded from the 2000 analysis because of too few cases). Sex (male vs. female) is never significant; the negative coefficients indicate that except in 2004, women were more likely than men to vote for the Democratic candidate. 9 Respondents' own opinions on equal roles were significant only in 1988. But their perception of candidate differences in support for equal roles was highly significant in each of these years. The negative coefficients indicate that white voters who perceived the Republican as more supportive of traditional roles than the Democrat were considerably less likely to vote for him. While Republican candidates may have gained votes on the basis of racial resentment, they lost considerable support from voters' perceptions of them as less supportive of equality for women.

Race, sex, gender, and the economy in 2008 and 2012
Because of Obama's historic candidacy, the 2008 and 2012 elections provide a strong test for the KDR argument of the primacy of race. As Figure 3 showed, voters placed the candidates even further apart on the aid to blacks scale in both years. Mean values for whites on the Racial Resentment scale were also higher in both 2008 and 2012 than in previous years. In 2008 the campaigns of Hillary Clinton and Sarah Palin candidacies kept the spotlight on gender, and mass media coverage provided numerous examples of sexism and misogyny (Carroll, 2009;Lawless, 2009). Obama highlighted his opposition to the Ledbetter decision and his support of the Fair Pay Act. 10 Lilly Ledbetter herself was prominently featured on the campaign trail and at the Democratic national convention. But John McCain had voted against the Fair Pay Act in the Senate, and his campaign rhetoric avoided these issues altogether (Hansen, 2014, p. 75).
The 1988-2008 results shown in Table 1 suggested that indicators of both race and gender influenced votes for president. However, many other issues can impact vote choice, and the KDR analysis of 2008 included several of these, along with racial resentment, opinions on gay marriage, abortion, and assessment of President Bush's job performance. They also included basic demographics (age and education), frequency of church attendance, 11 and indicator variables for census regions, as well as sex (the male/female dichotomy). Table 2 displays the results of logistic regressions of the 2008 presidential vote by whites. The base model in the first three columns for 2008 replicates that of Kinder and Dale-Riddle (2012, p. 229) as described above. Racial resentment is again a highly significant predictor of the white vote, along with party identification and approval of President Bush. Age, education, and frequency of church attendance were likewise significant. But despite the sizeable gender gap in 2008, sex had no significant independent effect in this model (although based on the size of the regression coefficients shown in Table 1, it had a considerably larger impact than in earlier years).
The expanded model in Table 2 tests for the impact of sex, racial resentment, and the aid to blacks and gender roles questions (both respondents' own opinions and their perceptions of candidate differences), along with respondents' views of abortion and gay marriage. Since polls showed that the economy was the most important issue in 2008, questions on retrospective judgments of both personal finances and the national economy were added. 12 The expanded model is limited to 293 cases because ANES only asked subsets of respondents the traditional abortion and equal-roles questions. 13 Based on the expanded model in Table 2, retrospective assessments of neither the national economy nor personal finances had much independent impact on vote choice in 2008. 14 However, disapproval of the incumbent president was highly significant; as Fiorina (1981) suggested, this may represent a mediated retrospective assessment of the economy. Obama's race may well have trumped the economy; Kinder and Dale-Riddle argue that Obama would otherwise have won by a landslide in 2008 (2012, pp. 104-106).
In this expanded model, both racial resentment and perceived candidate differences on the equal-roles scale had considerable impact on the white vote, but respondents' own positions on the equal-roles and aid to blacks scales were not significant. Older voters were considerably more likely to vote Republican, but neither education nor church attendance had much independent effect. Despite KDR's argument that the impact of racial resentment is due in part to its link to the parties, perceived candidate differences in support for aid to blacks had little independent effect in 2008. As in previous years, the negative regression coefficients indicate that voters who perceived McCain as more supportive than Obama of traditional roles for women were significantly less likely to vote for him. Thus, the basic and expanded models both show that gender as well as race influenced vote choice in 2008.
Race was again salient in Obama's bid for a second term. Republicans tried in vain to keep the campaign focused on the sour economy and the unpopularity of Obamacare, but the social issues they tried to dismiss as distracting "shiny objects" captured considerable attention from both old and new media. 15 Mitt Romney's "binders full of women" comment went viral after the second presidential debate. Reproductive rights were also prominent because of the proposed birth control mandate in Obamacare, efforts by many states to restrict access to abortion and contraception, the Sandra Fluke controversy, and Republican candidates' impolitic remarks concerning rape and abortion.

Table 2. Logistic regression: 2008 vote for McCain (whites only)
Notes: Not shown: coefficients for regional dummies.

Base model (K-DR)
Expanded model Despite the Republicans' best efforts, "shiny objects" were highly visible in 2012. Proposals to ban abortion, require vaginal ultrasounds, further restrict gay marriage, or deny abortions to rape victims received extensive media coverage and were unpopular with much of the voting public. Several Congressional Tea Party supporters elected in 2010 were defeated, and Republicans also lost two winnable Senate seats because of their candidates' impolitic statements about rape and abortion (see Note 26).
Unfortunately, ANES did not include any of the equal-roles questions in its 2012 survey. However, respondents' attitudes toward appropriate gender roles can be assessed based on several other ANES questions concerning the status of women. As shown in Table 3, females of all ethnic backgrounds were more enthusiastic than males about a woman as president, more concerned with workplace equality, and less likely to discredit claims of discrimination as "causing more problems than they solve." African-Americans and Hispanics were more supportive of equal roles than whites. Responses to these questions were combined into a factor score where higher values indicate greater support for equality for women, and females scored considerably higher than males on this scale. 16 The 2012 ANES also included a new question: whether the Democratic or Republican party "does a better job for the interests of women, or whether the parties were equally good or bad." Unfortunately the ANES survey provides no further information to indicate what "the interests of women" meant to respondents. 17 Obama undoubtedly benefited from the perception of 51% of voters, 54% of women voters, and 81% of Democratic identifiers that the Democratic Party was better for women (Table 4). Although these assessments were highly partisan, only 37% of Republican identifiers thought the Republican party was better for women; most Republicans equivocated with the "equally good or bad" response. Since candidate placements on the equal-roles scale were not asked in 2012, this question will be used instead to indicate perception of party differences in support for "the interests of women." In Table 5 the basic logistic regression model for the 2012 vote includes the same variables as the KDR model for 2008. Disapproval of the incumbent president carried considerably more weight than in 2008; this could represent whites' dislike of Obama as well as their negative assessment of Obama's management of the economy. Racial resentment was statistically significant, but again, sex (male vs. female) was not. The expanded regression model for 2012 in Table 5 included the factor scores for women's roles described above; whites with lower scores were significantly more likely to vote for Romney. As in 2008, opinions on abortion, gay marriage, or retrospective assessments of the economy had little independent effect. Neither did respondents' perceptions of differences in candidates' positions on the seven-point aid-to-blacks scale or the four-point ANES abortion scale. 18 But sex (male vs. female) was statistically significant, reflecting the record gender gap in 2012 (Jones, 2012). Compared with the base model, racial resentment had much less impact, but assessment of presidential job performance was again significant. However, the question as to which party is better for women's interests was a highly significant predictor of the 2012 vote, independent of party identification, racial resentment, and other issues. Separate regressions showed this to be the case for both men and women. 19 As in 2008, indicators of both race and gender influenced whites' votes. 20

Wedge issues in 2012
The better-party-for-women question could well be endogenous, since party identification strongly influences voting behavior as well as perceptions of candidates' positions on issues. Thus partisans already committed to Romney or Obama would be highly likely to select the position favoring their party or candidate. 21 Panel data (interviews with the same voters over time) could help to disentangle the causal ordering, but this was not an option for the 2012 ANES. 22 However, a sizeable proportion of Americans are at odds with their party on at least one issue, and therefore "persuadable" to vote across party lines (Hillygus & Shields, 2008). Endogeneity cannot account for the behavior of such cross-pressured voters.
Neither party can win national elections with votes only from its own supporters. Party strategists need to find "wedge" issues that will appeal to Independents and attract at least some supporters of the opposing party. Was "the interests of women" an effective wedge issue in 2012? If so, how did it compare with other potential wedge issues such as the economy, abortion, gay marriage, or Obamacare? To help answer these questions, I constructed a "wedge weight" variable. This represents the proportion of cross-pressured voters among Democratic or Republican partisans, multiplied by the percent of defectors among cross-pressured voters. Following Hillygus and Shields (2008, p. 55) Republican and Democratic "leaners" are excluded from the coding of partisans. Thus Obama received 44% of the votes of the cross-pressured Republicans who thought the Democrats were better for women. Since these voters constituted 12% of Republican identifiers, the "wedge weight" is 528. Obama received only 50% of the vote from the cross-pressured Democrats who thought the Republican Party was better for women, but since these voters were less than one percent of Democratic identifiers, the wedge weight for Democratic partisans was only 50. Table 6 compares the wedge weights for Democratic and Republican identifiers with cross-pressured positions on abortion, gay marriage, aid to blacks, evaluation of the Democratic or Republican Party as better for women or the economy, or other issues prominent in 2012: the budget deficit, the Troubled Asset Relief Program [TARP], the Affordable Care Act [Obamacare], defense spending, LGBT discrimination, affirmative action, and gun control. Table 6 also indicates the percentage of crosspressured partisans at odds with their own party's position on each issue. 23 In 2012 a substantial proportion of partisans were cross-pressured, especially Democrats on aid to blacks, TARP, gay marriage, and defense spending, and Republicans on abortion, gun control, and LGBT discrimination. 24 But overall, Republicans were more likely to defect. Based on these issues, perceived party differences in support for women was the most effective wedge issue for the Democrats (wedge weight of 528), followed by LGBT discrimination. Obamacare ranked third for the Democrats (444); Republicans who favored it were slightly more likely to defect, although a substantial proportion (36%) of Democrats who opposed Obamacare voted for Romney (wedge weight of 360).) Abortion was fourth for the Democrats; while 88% of pro-life Democrats voted for Obama, 15% of pro-choice Republicans supported him. Only 7% of Democrats, but 28% of Republicans, were at odds with their party on this issue.
Based on the wedge weights, the economy in 2012 was a net plus for Obama. He received 88% of the votes from the three percent of Republicans who thought the Democratic Party was better for the economy. Most Americans favored reducing the budget deficit and few voted across party lines on that basis. But the Targeted Asset Relief Program [TARP] was more effective as a wedge issue for the Republicans; 25% of Democrats opposed it, while only nine percent of Republicans supported it. Government aid to blacks and affirmative action were also effective wedge issues for Republicans, since a third of Democrats opposed them (although most of them still voted for Obama).
In 2012, gay marriage was a more effective wedge issue for Democrats than for Republicans. Although 21% of Democrats opposed gay marriage, only 11% of them defected. Nineteen percent of Republican identifiers supported gay marriage, and 17% of them defected. Thus, Obama gained more votes from Republican supporters of gay marriage and reproductive choice than he lost among Democrats at odds with their party's positions. However, more Democrats than Republicans oppose their party's positions on gay marriage, aid to blacks, defense spending, and affirmative action, all potential wedge issues for Republicans. 25

Conclusion
In 2012, 84% of respondents to an ANES question claimed that Obama's race had no influence on their vote. But Kinder and Dale-Riddle argue that racial resentment remains a key issue in American voting behavior, in part because the political parties differ so sharply on race-linked issues such as affirmative action, crime, and welfare. As Figure 3 showed, voters indeed perceive increasing differences between the parties based on the ANES aid to blacks questions, and Democratic and Republican voters have also moved further apart on this issue.
As this analysis has shown, racial resentment, based on the KDR scale, was alive and well among white voters in 2012, as it had been in 2008 and previous years. But the evidence presented here has challenged the KDR claim that "race trumps gender." Although the "gender gap" is indeed much smaller than the racial gap in voting, the simple male/female dichotomy is a very limited test of the broader impact of gender on presidential elections. Based on the ANES seven-point scale, the American public has grown considerably more supportive of equal roles for women, but opinions on appropriate gender roles have seldom had much impact on vote choice. As party platforms and campaign rhetoric show, the Democratic and Republican parties have diverged considerably on abortion and other issues affecting women, and the electorate is increasingly aware of these differences (Figures 1 and 2). As shown in Table 1, both racial resentment and perceptions of candidate differences on equality for women have significantly influenced the white vote in presidential elections, 1988-2008. Despite Republican efforts to keep the focus on the economy and Obamacare, women's issues figured prominently in the 2012 campaign. Unfortunately for research purposes, the 2012 ANES did not include any of the equal roles questions, but three other indicators of gender proved to be significant predictors of the 2012 vote. Women were more likely than men to vote for Obama, resulting in the largest gender gap ever recorded. And respondents with lower scores on a women's roles scale favored Romney. Perception of the Democrats as "better for the interests of women" was highly significant, and proved to be the most effective wedge issue for the Democrats in 2012 (Table 6). These results substantiate the conclusion of Deckman and McTague (2014) that the "war on women" in 2012 aided the Democrats. 26 Abortion has often been used to assess the influence of gender issues in elections, but this analysis has supported evidence by Hansen (2014), Deckman and McTague (2014), and Sides and Vavreck (2014) of its minimal independent effect on the vote in recent years. In 2012, neither respondents' personal views of abortion nor their perceptions of candidate differences (based on the ANES fourpoint abortion scale) had much impact. Abortion may have contributed to partisan realignment, especially among pro-life Democrats (Killian & Wilcox, 2008). Yet a substantial proportion of prochoice Republicans remain within the GOP, as Table 6 showed, and abortion was a successful wedge issue for the Democrats in 2012. The Republican strategy to emphasize abortion in future elections may well backfire.
In addition to KDR's work on racial resentment, considerable scholarly effort has been devoted to uncover evidence of implicit or unconscious racism, including Mendelberg (2008), Peffley and Hurwitz (2010), Bonilla-Silva (2013), and a lengthy series of questions based on images of African-Americans and Hispanics in the 2012 ANES. The ANES equal-roles scale suggests that most Americans claim to support equality for women, and the lack of variance on this indicator was one reason it was dropped from the 2012 ANES. 27 But other ANES questions (Table 3) show that men are considerably more skeptical than women about workplace equality, discrimination, or a woman as president, and a scale based on opinions about these issues did influence the vote in 2012. Perhaps more attention is needed to alternative experimental or neurological analysis of gender attitudes and implicit sexism, such as recent work by Cicara, Eberhardt, and Fiske (2010).
More attention should also be devoted to the highly gendered political environment confronting voters. The Republican Party thus faces a significant dilemma; the GOP is further away from voters' positions than the Democrats on both abortion and equality for women, and its opposition to gay marriage is increasingly out of step with Americans' liberalizing opinions. Demographic trends favor the Democrats, with growing numbers of seculars, younger voters, those with advanced degrees, divorced or unmarried individuals, and minorities. Republican opposition to contraception and abortion, and their demonization of Sandra Fluke, may have cost them a significant number of votes from single women, who turned out in large numbers in 2012 (Center for American Women in Politics, 2012). The public does support many restrictions on access to abortion, particularly late-term abortions. But repeal of Roe v. Wade, and the outright bans on abortion (even in cases of rape) advocated in Republican party platforms are seriously at odds with public opinion. Proposed "personhood" amendments to ban all abortions have been defeated even in conservative states such as Mississippi and South Dakota. 28 At least some Republicans recognize the party's "women problem." Party leaders and some major donors have urged candidates to clean up their rhetoric, and have actively discouraged conservative firebrands and Tea Party supporters from challenging incumbents in primaries (Peters, 2014). By contrast, the Democrats are well aware of their strategic advantages on women's issues, and their candidates have emphasized their support for equal pay, reproductive rights, and child care. However, Democratic candidates are perceived as far more supportive of "aid to blacks" than even their own partisans, and this proved to be a significant wedge issue for the Republicans in 2012. Thus, both race and gender influence presidential voting, and both parties must confront difficult strategic choices as they look for votes.

Funding
The author received no direct funding for this research. 205-207) on alternative measures of racial resentment, racial stereotyping, and symbolic racism; each of these was a significant predictor of white votes. Bonilla-Silva (2013) analyzes "color-blind" racism. 4. The "aid to blacks" wording is obviously problematic, but will be used in this analysis because it permits comparisons across party positions and elections since the 1970s. Specific policies (welfare, affirmative action, and criminal justice) show similar sharp differences in support by whites and African-Americans (Gilens, 2005). 5. The KDR racial resentment scale combined several ANES questions concerning "blacks" (see Appendix A for question wording). The scale was constructed by recoding the 1-5 responses to these questions to range between 0 and 1, summing them, then dividing by 4. 6. "Wrong" assessments may be problematic because the ANES issue scales do not correspond to specific policies, nor do they capture nuances in the candidates' positions. In 2004, for example, John Kerry was personally opposed to abortion but supported Roe; George Bush opposed Roe but would have permitted abortions in cases of rape or incest. 7. See Hansen (2014, pp. 17-19) and Abramson et al. (2012, pp. 152-68) on trends over time in don't-know and no-difference responses to the ANES seven-point scales. 8. See Abramowitz (1995) on the impact of abortion in 1992. 9. The "gender gap" is usually defined as the difference in the percent of women's and men's votes for a candidate. In most elections since 1980 more women than men have voted for the Democrat. The usual pattern reversed in 2004; George Bush's "compassionate conservatism" and anti-terrorist policies convinced more women than usual to vote Republican (Carroll, 2008 proven to be a far better predictor of attitudes and voting behavior than religious identity or denomination, although its impact has diminished since the 1990s (Hansen, 2011;Putnam & Campbell, 2012). 12. These retrospective judgment have been found to be better predictors of the vote than respondent's income, employment status, or predictions of future economic trends (Fiorina, 1981). Vavreck (2008) found that assessments of the economy were also influenced by campaign strategy. 13. The questions on perceptions of candidate positions on abortion had to be omitted from the 2008 analysis because of too few cases. Although techniques are available to interpolate missing data, they are not appropriate when so many cases are excluded by survey design. 14. Assessments of the economy have become highly partisan (Evans & Anderson, 2006). In 2008, Democrats were more likely to see themselves as "worse off" and to blame President Bush, while in 2012 Republicans claimed to be worse off and blamed Obama (Hansen, 2014, pp. 143-144). 15. Romney's campaign adviser Eric Ferhnstrom claimed that the Democrats would use social issues as "shiny objects" to distract voters from economic concerns (Edwards, 2012). See Halperin and Heilemann (2013) for further analysis of the impact of social issues in 2012. 16. The Women's Roles scale was based on a factor analysis of the questions in Table 3 (principal components analysis, Eigenvalue of 1.677, 42% of the variance explained). Comparable questions asked in 2004 and 2008 showed strong positive correlations with the seven-point equal-roles scale, but considerably more variance across ANES respondents. 17. The better-party-for women question showed modest positive correlations with education, liberal ideology, infrequent church attendance, pro-choice views, and scores on the women's roles scale. But a considerable number of conservative, church-going, pro-life respondents preferred the Democrats over the Republicans on "the interests of women." 18. Candidate differences on abortion were calculated by subtracting the perceived Democratic candidate's position from the perceived Republican position. The scale ranged from −3 to +3, with zero indicating no difference and higher values indicating the Republican as more pro-life. 19. On male/female differences in voting in 2012, see Deckman and McTague (2014) and Hansen (2014, pp. 147-48). 20. Following KDR, Hispanic whites were excluded from this analysis. Hansen (2014, p. 180, 185) found that Hispanic self-identification had no independent impact on the presidential vote in either 2008 or 2012, although a majority of Hispanics in both years supported Obama. See Sides and Vavreck (2014) for further analysis (based on much larger surveys) of the Latino/Hispanic vote in 2012. 21. See Redlawsk (2002) for analysis of how such "motivated reasoning" affects people's assessment of party and candidate positions. 22. For further discussion of the endogeneity problem and use of panel data to estimate the impact of issues on the Obama vote in 2008, see Hillygus and Henderson (2010). 23. Republicans classified as cross-pressured were those who supported TARP, opposed reducing the federal budget deficit, or responded 1, 2, or 3 on the aid to blacks scale (supporting government assistance to blacks); cross-pressured Democrats gave responses of 5, 6, or 7 and thus agreed that "Blacks should get ahead on their own." TARP had been proposed by President George Bush and passed Congress with bipartisan support in the fall of 2008, but attracted widespread Republican opposition after the Obama administration expanded it to cover bailouts of the auto industry (Hensarling, 2009). 24. Only about 25% of partisans agreed with their party positions on all of these issues, comparable to results reported by Hillygus and Shields (2008, p. 71) for previous elections. 25. Any Republican advantage on gay marriage may be eroding; since 2013, polls have shown that a majority of Americans support gay marriage, and in 2015 the Supreme Court's Obergefell v. Hodges decision made it legal in all states. 26. Their analysis used an American Values survey conducted by the Public Religion Research Institute, which included questions on abortion and support for the birth control mandate but no other indicators of gender. Sides and Vavreck (2014) argue, based on panel data from YouGov polls, that neither the Republican "War on Women" nor abortion led many voters to switch to Obama during the campaign. However, their analysis does not account for the record gender gap in 2012 or the surge in turnout by single women. Based on exit polls, women's votes were responsible for the Republican Senate losses in Indiana and Missouri (Hansen, 2014, p. 139