Plus ça Change: Race, Gender, and Issue Retrospections in the 2008 US Presidential Election

Abstract

What role did race and gender play in vote choice for US president in 2008? With an African‐American Democratic presidential nominee and a female Republican vice presidential nominee, were the black–white racial gap and the gender gap different in 2008 than in previous years? We find that the racial gap between black and white voters was larger in 2008 than it was in three of the four previous presidential elections, but not significantly larger than the gap in 1996 when white Democrat Bill Clinton won reelection. This gap is still present after we include control variables. On the other hand, the apparent gender gap whereby women are more likely than men to support Democratic candidates goes away once we include controls. Other factors – namely Iraq retrospections, ideological voting, and partisan identification – played a role as large or larger in citizen vote choice as did race or gender in the 2008 general election. Even though a voter’s race was a key predictor in 2008 and even though there may be a widening racial gap, it is important not to overstate the role of race relative to other well‐known predictors of the vote.

Acknowledgments

We thank participants at the Geurin‐Pettus Conference on the 2008 US Elections at SMU‐in‐Taos as well as two anonymous reviewers for reactions and suggestions to an earlier version of this paper.

Notes

1. We use the terms race and gender throughout the manuscript (instead of, e.g., sex), consistent with past literature examining vote choice.

2. It is possible that racial or gender bias may have exacerbated the black–white racial gap or the gender gap in 2008, but this paper focuses on a specific question: Did a citizen’s race or gender affect vote choice, even after controlling for other factors that predict the vote? We do not directly test for the presence of racial or gender bias or resentment in this paper, but instead compare the size of the racial and gender gaps in 2008 to those of previous elections – see Jackman and Vavreck (2010) for a direct examination of racial resentment in the 2008 primary. In other words, we cannot ascertain whether gaps are due to bias emanating from racism or sexism (or even preferential treatment toward candidates of a certain race or gender), or rather to some other considerations that are correlated with race or gender. What we are able to do, however, is to demonstrate whether the gap is still present and whether its size has changed over time. An implicit assumption may be that there is no a priori reason – other than the race (gender) of the candidates – to expect that the racial (gender) gap should be larger (smaller) in 2008 than in previous years. Thus, if the gap in support between black and white voters for Obama is larger than the gap in support for a white Democratic candidate, an implication is that racial biases may have affected the vote (though we cannot say that with certainty).

3. While it is true that this is not the first time that a black Democrat faced a female Republican in any context, the fact that this occurred in the context of a presidential election is important on three grounds. First, it means that we have a national sample of respondents who faced this choice, rather than the smaller samples usually available for congressional or local contests. Second, congressional, state, and local elections are difficult to compare given the large variation in local contexts, candidacies, and other state‐ and local‐level considerations that vary across these settings. In contrast, the 2008 presidential election offers the opportunity to compare the effect of race and gender when the same candidates face each other across the entire country. Third, a presidential campaign is the most highly visible, well‐covered contest, and as such this analysis offers an examination of race and gender issues in the context of a relatively high‐turnout, high‐salient contest.

4. Obama identifies as biracial. As Obama stated himself, “I’ve always been clear that I’m rooted in the African‐American community but not limited to it” (quoted from Leibovich in Bobo & Dawson, 2009). However, given the media coverage of the general election that focused on his history‐making campaign to become the nation’s first African‐American president, we consider black voters as sharing Obama’s racial identity, but white, Latino, Asian‐American, and other voters as not sharing his racial identity; this perspective is echoed by Smith and King (2009), who declared that “Americans elected their first black president”. Of course, early on in the 2008 primary campaign readers will recall that Obama’s racial background was questioned. Some asked whether Obama was “black enough” to garner the support of African‐American voters (Coates, 2007). By the general election, though, the popular press was much more likely to identify Obama as a black candidate. Nevertheless, we should also remember that race and gender are cultural constructions, and not rigid categories – even if we and others who examine the effect of race and gender on vote choice are sometimes forced to measure them as such.

5. All percentages in Table 1 refer to the two‐party vote.

6. We exclude from the analyses respondents who cast ballots for third‐party or independent candidates.

7. We include the southern dummy variable because many white southerners are critical components of a Republican candidate’s coalition; and because race and vote choice are both highly correlated with that region. Nevertheless, we estimated our models without the southern dummy and the results were generally consistent (see the last column in Appendix Table A1).

8. The Religious attendance variable is measured as 0 if the respondent never attends; 1 if attends a few times a year; 2 if attends once/twice a month; 3 if attends almost every week; 4 if attends once a week or more than once a week.

9. The Education variable is measured as 1 if the respondent did not graduate from high school; 2 if a high school graduate; 3 if some college/associate degree; 4 if the respondent received a bachelor’s degree; and 5 if the respondent has a graduate degree.

10. For instance, to calculate the predicted probability of voting Democrat in 2008 for African Americans, the White and non‐white, non‐black variables were set at 0 and all other variables were held at their 2008 means.

11. This predicted probability of 0.70 for African Americans is a little lower than may be expected. This low value occurs because of the loss of respondent cases when we include the ideological distance variable in our models. When we include those cases by estimating a model without the ideological distance variable, the point estimate for the predicted probability of voting Democratic for African Americans is higher (greater than 0.80). But importantly for our analysis, if we drop the ideological distance variable, the finding with respect to the statistical significance of a racial gap does not change: it is lower in 2004 than in 2008 (see the second‐to‐last column in Appendix Table A1).

12. Again, the relatively low predicted probability of African Americans voting Democratic in 2000 (0.63) is due to the exclusion of cases where respondents did not answer the ideological distance question. As can be seen in Appendix A, the racial gap is lower in 2000 than in 2008, with or without the inclusion of the ideological distance variable. While the predicted probability point estimate for black voters differs somewhat with the inclusion of this variable, the presence of a statistically significant relationship between the race variables and vote choice is generally consistent in both models.