The earliest political behavior researchers documented the powerful effects of group attachments and other socioeconomic factors on vote choice and partisan identification in the 1940s and 1950s (Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Campbell et al. Reference Campbell, Converse, Miller and Stokes1960). Yet, research interest in the group-based origins of political behavior has waxed and waned in the intervening decades (Huddy Reference Huddy, Sears, Huddy and Jervis2003). The existence of both an African-American and female frontrunner for the Democratic presidential nomination in 2008 provides an opportunity to consider the contemporary electoral consequences of in-group loyalties and out-group antipathies. We take advantage of select survey and poll data collected during the 2008 Democratic primaries to evaluate the power of gender and race as both positive and negative influences on voter calculus in an election in which the two major candidates were differentiated less by their issue positions and beliefs than by their skin color and gender.
In analyzing the political power of group membership, it is instructive to contrast the effects of group solidarity in which members are motivated to support a candidate who belongs to their in-group with the effects of group aversion to a candidate from a disfavored out-group. Both in-group loyalties and out-group antipathies are apparent in American electoral politics. Recent political behavior research has typically paid greater attention to the electoral effects of out-group antipathy than in-group solidarity, with the lion's share of this work focused on the negative electoral effects of white racial animosity towards black candidates (Becker and Heaton, Reference Becker and Heaton1967; Kinder and Sears, Reference Kinder and Sears1981; Reeves Reference Reeves1997; Sears, Citrin, and Kosterman, Reference Sears, Citrin, Kosterman, Moreland, Steed and Baker1987; Terkildsen Reference Terkildsen1993). Nonetheless, the influence of in-group loyalties has also received research attention (Dawson Reference Dawson1994; Dolan Reference Dolan2008; Tate Reference Tate1994). Overall, race has a more powerful and less ambiguous effect than does gender on voting for an in-group and against an out-group candidate.
In-group Solidarity
Consider first the power of in-group loyalties. African Americans exhibit an impressive degree of in-group loyalty and are far more likely than whites to vote for black candidates (Philpot and Walton Reference Philpot and Walton2007; Reese and Brown Reference Reese and Brown1995; Sigelman and Welch Reference Sigelman and Welch1984). When examined more closely, stronger black support for African-American candidates is clearly linked to racial loyalties and common group interests. The electoral effect of group loyalties is most pronounced among African Americans who subjectively identify more strongly with their racial group (Dawson Reference Dawson1994; Tate Reference Tate1994). Both Katherine Tate (Reference Tate1994) and Michael Dawson (Reference Dawson1994) report that black support for Jesse Jackson in the 1984 and 1988 Democratic primaries was highest among African Americans with a strong sense of common fate. African-American support for black candidates can even occur at the expense of more relevant contextual information, such as the candidates' ideological predisposition (Kuklinski and Hurley Reference Kuklinski and Hurley1994), going beyond shared values and group interests to reflect a sense of group pride and solidarity with a fellow in-group member. From this vantage point, we would expect strong black solidarity for an African-American candidate; this is indeed what occurred in the presidential election of 2008 in which Barack Obama received stronger black support than did recent Democratic presidential candidates.
When it comes to women's support of women candidates, in-group favoritism is less universal and more conditional on electoral circumstances. Female voters who support antisexual harassment policy, take a pro-choice position on abortion, and favor parental leave policies typically drive the gender gap in support of female Democratic candidates. Moreover, this gap is most pronounced in elections in which such issues are salient (Dolan Reference Dolan2008; Paolino Reference Paolino1995).Footnote 1 Thus, shared group interests may play a similar role among both women and African Americans by fostering support for an in-group candidate. But as a general rule, women show considerably less political cohesion than do African Americans—they do not uniformly support the Democratic Party, are as supportive as men of most women's issues, and evince modest gender gaps across a range of policy issues (Huddy et al. Reference Huddy, Neely and Lafay2000; Huddy, Cassese, and Lizotte 2008a, 2008b; Kinder and Winter Reference Kinder and Winter2001). The modestly sized and variable gender gap in support of women candidates likely reflects women's typically fragmented political loyalties (Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte and Whittaker2008b).
Other factors unrelated to collective identity and interests may also play a role in driving women's greater support of women candidates. Most notably, the gender gap in support of female candidates rests heavily on women's stronger support than men of the government provision of social services, and their greater emphasis on social welfare issues in deciding for whom to vote (Dolan Reference Dolan1998; Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte, Wolbrecht, Beckwith and Baldez2008a; Shapiro and Mahajan Reference Shapiro and Majahan1986). Women's support of women candidates is, therefore, especially pronounced when a female candidate takes a liberal position on social welfare spending and is most prominent in elections that involve a woman Democratic candidate (Dolan Reference Dolan2008; Paolino Reference Paolino1995). Women's stronger support of social welfare spending might reflect group solidarity since women are more likely to benefit from certain social welfare policies. But the gender gap in support of women candidates extends well beyond women who hold a feminist identity or support explicit women's issues and is thus difficult to construe as a clear-cut instance of gender loyalties (Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte, Wolbrecht, Beckwith and Baldez2008a).
Women candidates need not support expanded social welfare spending, however, to elicit stronger support from female than from male voters. In low-stimulus elections, such as congressional House races, the gender gap in support of women candidates stems from gender stereotypes that portray women candidates as more liberal, supportive of women's issues, and pro-government than men, even when they are not (Koch Reference Koch2002; McDermott Reference McDermott1997; Sanbonmatsu Reference Sanbonmatsu2002). This gender gap can evaporate once women voters learn something about a female candidate's ideology and political views. Thus, a conservative woman such as Sarah Palin, the 2008 Republican nominee for vice president, was extremely unlikely to arouse strong female support because she did not attract liberal women voters who typically provide a boost to female Democrats.
Overall, women's support for a woman candidate has a somewhat different basis than blacks' support of African-American candidates. Black voters often express stronger support than whites for a black candidate, even after controlling for voter ideology and support for racial policies, which differ markedly between blacks and whites (Kinder and Winter Reference Kinder and Winter2001). In that sense, black racial loyalties more readily translate into unconditional support for black candidates than do gender loyalties (although this support does not extend to Republican black candidates, who are seen as unsupportive of black interests). Blacks are also more likely to acknowledge common group interests and perceive in group-based inequities than are women (Huddy, Neely, and Lafay Reference Huddy, Neely and Lafay2000). At present, it is fair to conclude that both blacks and women support in-group candidates in part because such candidates are seen to further the group cause. But blacks may additionally support African-American candidates because they share the same race; comparable gender solidarity is less evident among women.
Outgroup Antipathy
There is ample evidence that racially negative whites are reluctant to support black candidates (Becker and Heaton Reference Becker and Heaton1967; Kinder and Sears Reference Kinder and Sears1981; Sears, Citrin, and Kosterman Reference Sears, Citrin, Kosterman, Moreland, Steed and Baker1987). Donald Kinder and David Sears (1981) initiated this line of work, furnishing evidence that white symbolic racism reduced support for Tom Bradley, a black candidate running for mayor of Los Angeles in the late 1960s and early 1970s. Other studies demonstrate that white bias against black candidates may be especially pronounced for candidates perceived or known to be liberal ideologically or supportive of pro-black policies (Citrin, Green and Sears Reference Citrin, Green and Sears1990; McDermott Reference McDermott1997; Reeves Reference Reeves1997; Sigelman et al. Reference Sigelman, Sigelman, Walkosz and Nitz1995). Thus, liberal black candidates who are proponents of race-based policies likely elicit greater opposition from whites than do black candidates who hold a moderate record and avoid racially-charged issues. In the extreme, political figures such as Colin Powell are seen in completely race-neutral terms that elicit absolutely no racial animus from whites (Kinder and McConnaughy Reference Kinder and McConnaughy2006).
Is there comparable evidence of sexism among men that undercuts their support of women candidates? To date, there is insufficient research that focuses on male antipathy toward female candidates to fully answer this question. Men and women do not differ in their stated willingness to vote for a qualified woman nominated by their party for president (Dolan Reference Dolan2008). Women tend to think their country would be better governed if there were more women in office, but this may reflect some amount of female in-group solidarity, rather than male out-group antipathy (Dolan Reference Dolan2008). Men are less supportive than women of female candidates on occasion, but this also seems more a product of women's enthusiasm for liberal female candidates than male opposition to them. In reality, the issue of male opposition to women's political candidacies has received much less attention than it deserves.
2008 DEMOCRATIC PRESIDENTIAL PRIMARIES
This brings us to the 2008 Democratic primaries. To what extent did group factors influence support for Hillary Clinton and Barack Obama? As noted earlier, a primary campaign eradicates the typical ideological divide between Republicans and Democrats that can account for racial and gender gaps in candidate support. In general, women voters and candidates tend to be more Democratic than Republican; the same is true of black voters and black candidates (Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte, Wolbrecht, Beckwith and Baldez2008a). Thus in a typical election, support for an in-group candidate is often conflated with partisanship (Dolan Reference Dolan2008). A primary election provides a unique opportunity to examine group influences on vote choice when not confounded with ideology or partisanship.
Gender and Race-Based Loyalty
Exit Polls
We first examine the political effects of in-group loyalties in the 2008 Democratic primary exit poll data. Exit polls were conducted in 39 states during the primaries. To assess the impact of group loyalties on vote choice, we concentrate on data from the 31 state exit polls in which respondents were asked about the importance to their vote of the candidate's race and gender. These data reflect responses from more than 39,000 voters. Missing states include many that lacked exit polls because they held caucuses, not primaries, and several that held primaries and conducted exit polls but did not ask voters the importance of gender and race to their vote choice.Footnote 2
There is a clear gender gap in support of Clinton in these data. White women were more likely than white men to vote for Clinton (61% vs. 49%), and Latinas were more likely than Latino males to vote for her (66% vs. 56%). There was a much smaller gender gap in support for Clinton among blacks (12% of males vs. 15% of females). When considered across all racial groups, the gender gap is modest in size.
By comparison, the racial gap in support of Obama is enormous. Forty percent of whites and 35% of Latinos supported him, compared to a whopping 84% of African Americans. The racial gap persists if we consider males and females separately. Forty-six percent of white males and 39% of Latino males supported Obama, compared to 87% of black males. Among women, 36% of white women and 39% of Latinas voted for Obama, compared to 83% of black women. In essence, Clinton elicited a moderate gender gap of roughly 10 to 12 percentage points among whites and Latinos, whereas Obama aroused a large racial gap of 40 percentage points or more.
The exit polls provide another way to examine the impact of group loyalties on vote choice in the Democratic primaries. Respondents in all 31 states were asked whether race and then, separately, gender constituted the single most important factor, was one of several important factors, or was not important in his or her vote choice. This is admittedly a crude way to get at the influence of group loyalties because not everyone is aware of or willing to admit that their vote was affected by such considerations. Nonetheless, we can consider it a measure of blatant or explicit group-based voting that likely underestimates the role of group influence.
The data presented in Table 1 are instructive. Consider gender loyalties. A somewhat greater percentage of white women than men (21% vs. 13%) said that gender was important (most important or one of several factors) to their vote. And of those who said gender was important, almost 8 in 10 white women voted for Clinton, compared to just under 6 in 10 white males. This translates into roughly 17% of all white women who claimed overtly to have voted for Clinton for gender-based reasons, compared to 7.5% of white men. This reinforces findings of modest gender-based support for Clinton among women.
Table 1. Explicit race and gender-based voting in 2008 Democratic primary exit polls
Note: The table includes entries for 31 states in which exit polls were conducted in the 2008 Democratic primaries and that asked a question about the importance of race and gender. Missing states included a majority of the caucus states, plus the following states that held primaries but did not ask the race/gender questions: FL, MD, NH, MI, SC, VA. The specific question in the 2008 exit poll was as follows: “In deciding your vote for president today, was the race/gender of the candidate: the single most important factor, one of several important factors, not an important factor?” The percentage choosing the single most and one of several important factors are combined.
The racial gap in explicit group-based voting is much larger than the gender gap. Just under 30% of black men and women said that race was important to their vote, and they voted overwhelmingly for Obama. Fewer white men or women explicitly acknowledged race as a factor in their vote—15% and 14%, respectively—and on balance they did not vote for Obama. This results in roughly a quarter of black men and women who voted for Obama for explicit racial reasons, compared to 4% to 5% of white men and women and 8% to 11% of Latinos.
Time Pennsylvania Preelection Poll
As a further test of group influence on the 2008 Democratic primary vote, we turn to a Time magazine preelection poll of 676 likely registered Democratic Pennsylvania voters conducted several weeks before that state's primary (on April 22). The poll was conducted by Schulman, Ronca, and Bukuvalis, Inc (SRBI). The data include several questions on group-related attitudes. There is a gender gap of roughly 11 percentage points in the preelection Pennsylvania poll. Overall, 48% of females and 39% of males indicated that they would vote for Clinton if the election were held that day (with roughly 20% of both genders undecided). There was an even larger racial gap; 30% of whites and 80% of blacks indicated that they would vote for Obama. These findings mirror vote trends in the Democratic primary exit polls more generally.
Of greater interest, the Time poll asked respondents to indicate how worried they were (great deal, a fair amount, only a little, or not at all) about “Continuing discrimination against African-Americans,” and “The barriers that women face in jobs and promotions.” The two questions tap perceived group grievances. We analyze their impact on vote choice to assess whether they convey the influence of race and gender on support for an in-group candidate. If yes, it provides an indication of the degree to which women supported Clinton on the basis of a group-linked concern about gender barriers and blacks supported Obama on the basis of racial concerns about discrimination. The relevant findings from a multivariate probit analyses are presented in Table 2.
Table 2. Vote choice in 2008 Democratic Pennsylvania primary: TIME poll, April 2–6, 2008
Note: Entries are unstandardized regression coefficients; standard errors are in parentheses. Estimates in bold are statistically significant at p < .05. Likely registered voters were respondents who said they were likely to vote when asked the following question: “Thinking about the 2008 Democratic Presidential Primary in Pennsylvania, many people tell us they're not planning to vote, Would you say that you are definitely not going to vote, probably not likely to vote, might or might not vote, probably will vote, or are you 100% certain to vote in the upcoming Democratic Primary?” Vote intention was coded as 0 (Clinton), 0.25 (lean Clinton), 0.5 (undecided), 0.75 (lean Obama), 1 (Obama). Income is collapsed into the following dummy variables: low (<$50,000), medium ($50,000–$75,000), high (>$75,000), and no income information. All variables are coded 0 to 1 to help interpret unstandardized coefficients.
The first two columns of Table 2 contain probit estimates and standard errors indicating the impact of basic demographic factors and political ideology on vote choice. As seen in this table, African Americans supported Obama far more strongly than did whites, and women supported Clinton to a greater degree than did men. Group-linked attitudes are added to this analysis in a second equation presented in columns 3 and 4. In this second equation, a concern about either gender or racial discrimination influences vote choice. Individuals who worry about racial discrimination are far more inclined to support Obama, and those who worry about gender discrimination are more likely to vote for Clinton. More importantly, the inclusion of both discrimination items does nothing to diminish the large coefficient for race, and it slightly reduces the impact of gender. The coefficient for gender declines from –.31 in column 1 to –.23 in column 3.Footnote 3 Overall, women are somewhat more concerned than men about gender barriers, and this concern about gender issues partly accounts for their greater support of Clinton (a Sobel-Goodman mediation test indicates that gender discrimination accounts for 12% of the gender effect).
A similar pattern of findings is observed when comparable analyses are run for evaluations of Clinton and Obama in the Pennsylvania primary preelection poll. Respondents were asked “Is your opinion of [Barack Obama/Hillary Clinton] very positive, somewhat positive, somewhat negative, or very negative?” Attitudes toward Obama and Clinton were regressed onto the same independent variables included in Table 2. Once again, analyses were conducted in two stages, with group discrimination variables omitted in the first equation and included in the second. These analyses are presented in Table 3. Blacks rated Obama more favorably than did whites, and whites rated Clinton more favorably than did blacks. Women viewed Clinton more favorably than did men, although there was no gender gap in evaluations of Obama. These findings mirror the racial and gender gaps observed in vote choice analyses.
Table 3. Evaluations of Obama and Clinton among registered voters likely to vote in Democratic Pennsylvania primary: TIME poll, April 2–6, 2008
Note: Entries are unstandardized regression coefficients; standard errors are in parentheses. Estimates in bold are statistically significant at p < .05. Feelings toward each candidate were assessed with the following question: “Is your opinion of [Barack Obama/Hillary Clinton] very positive, somewhat positive, somewhat negative, or very negative?” See note to Table 2 for wording of the likely voter question, and coding of income. All variables are coded 0 to 1 to help interpret unstandardized coefficients.
A concern about racial discrimination boosted Obama's evaluations, but this only partly explained why blacks evaluated him more positively. The coefficient for race was 1.16 without the racial discrimination concern variables and 1.00 with it included; a Sobel-Goodman mediation test indicates that racial discrimination accounted for 30% of the race effect. In contrast, a concern about gender discrimination boosted Clinton evaluations and accounted for almost a half (45%) of women's higher ratings of Clinton. The coefficient for gender goes from a statistically significant .27 to a nonsignificant .15 with gender concerns included. Thus, women liked Clinton to a large degree because they were concerned about gender discrimination and may have seen her as likely to act on such issues. In contrast, strong black support for Obama was predicated more modestly on a concern about racial discrimination.Footnote 4
Overall, women's support for Clinton in the Pennsylvania preelection poll depends to some degree on concerns about gender discrimination, indicating that her candidacy elicited support from women on the basis of shared gender interests. Clinton did not overly stress women's issues or the historic nature of her candidacy in her campaign, but nonetheless garnered support from women who either saw her candidacy as a coup for women or believed she would address women's issues if elected president. In contrast, African-American support for Obama was less dependent on blacks' concern about racial discrimination. Worrying about racial discrimination only modestly mediated the effects of race; moreover, a concern about discrimination did not have a greater impact among blacks than whites on support of Obama (as tested by the inclusion of a nonsignificant interaction term between race and concerns about racial discrimination in analyses not shown here).
In reality, Obama's campaign did not stress traditional racial issues or civil rights, and he appears to have garnered support from blacks somewhat independently of such concerns. Ultimately, blacks may have supported him out of sheer group pride in his candidacy. We should add, however, that blacks were more uniformly concerned about racial discrimination than women were about gender barriers in the Pennsylvania preelection poll. Seventy-three percent of blacks indicated that they worried a great deal about racial discrimination, compared to 30% of women who worried a great deal about gender barriers. It is thus possible that black support for Obama is linked to a widely shared concern about racial discrimination among African Americans.
Women are more fragmented politically than are African Americans, undercutting their general political cohesion. But the existence of both a female and a black candidate in the 2008 Democratic primaries underscores the political power of an African-American identity to foster black support for an in-group candidate to a far greater extent than a woman candidate mobilizes female support.
Racism and Sexism
Group influences can help or hurt a political candidate. We finally consider the negative influence of gender and race on the outcome of the 2008 Democratic primaries. The Democratic primaries elicited heated discussion about sexist treatment of Clinton by some members of the news media. But there is little evidence in the exit poll data of such overt opposition to her candidacy. In Table 1, almost all white and Latino men and women who mentioned gender as important in their vote choice voted for Clinton.Footnote 5
In contrast, the exit polls provide ample evidence of outright racial negativity. Table 4 depicts the percentage of white voters who said race was important and either voted for or against Obama. States are included in the table if they were above average in the negative or positive impact of race on vote choice. The top panel lists states in which race explicitly undermined support for Obama among whites. Across all 31 states in which the question was asked, roughly 10% of whites voted against Obama in part because of race. Overt racist opposition to Obama was most pronounced in southern states. Over 20% of white Democrats in Mississippi and West Virginia voted against Obama on the basis of racism. It is interesting to note that potential swing states, such as Ohio and Pennsylvania, also appear in the list of states that were above average in the percentage of white voters who explicitly voted against Obama for race-related reasons.Footnote 6
Table 4. Explicit positive and negative impact of race on vote choice in the 2008 Democratic primary exit polls: Whites only by select states
Note: The table includes entries for all states that were at or above average in the negative or positive influence of race on vote choice. The race question was asked of voters in Democratic primaries in a total of 31 states. See the note to Table 1 for more information on wording of the race question and states excluded from the data.
CONCLUSION
The 2008 Democratic primaries were historic. Democratic voters across the country had the opportunity to decide between a woman or an African American as their party's presidential candidate. This choice prompted a series of questions from journalists, pundits, and others about whether Americans were ready for such a choice. Could whites really support a black candidate? Would men support a female? And on the flip side, the primaries fueled rampant speculation about “tribal” voting among women and blacks that would lead to bitterness and divisiveness as one group or the other ultimately lost the party's nomination.
As often happens, this portrayal of group influence on electoral outcomes was overly simplistic. In line with past research findings, race had a powerful influence on vote choice. Blacks across the country overwhelmingly supported Obama, possibly because of group interests but also perhaps because he is black and his candidacy offered an opportunity to express group pride in the first African-American major-party presidential nominee. Sizable minorities of whites in the South and elsewhere refused to vote for Obama on explicitly racial grounds. In contrast, the effects of gender were more modest. Women were more supportive than men of Clinton. And women who supported Clinton were more concerned about women's issues. But as we know, not all women share such concerns, thus muting the impact of gender on vote choice even in the 2008 Democratic primary in which Clinton's gender was a key issue (Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte, Wolbrecht, Beckwith and Baldez2008a).
If one lesson can be extracted from the 2008 Democratic primaries it is that what we have known for some time about group politics played out yet again in this fascinating series of primary elections. On the basis of current and past research, it is clear that blacks and women strongly supported Barack Obama in the November election. Blacks supported Obama in the Democratic primaries on the basis of group pride and a possible defense of racial interests. Women supported Clinton partly out of a concern for gender discrimination, a concern that did not translate into support for the McCain/Palin ticket. Our findings also make clear that a minority of whites could not support Obama because he is black. Journalists and pundits would have been well-advised to consult with students of group politics during the 2008 election to avoid a number of sensational, wrong-headed predictions about the defection of female Clinton supporters to McCain, a possible standoff at the Democratic convention, or concerns about implicit racial bias in the context of explicit racial bias. Group politics redux.
