The Exit Polls

The conclusion that citizens based their presidential vote largely on the basis of morals was first suggested by the results of a single question from the 2004 exit polls.²

The question asked 7,000 voters what they believed was the most important issue facing the nation. The now familiar results presented in Table 1 report 22% of exit poll respondents believed moral values was the most important issue, and 80% of these voters cast ballots for Bush. Comparatively, no other category received as many responses. With more voters choosing moral values over any other response category, observers quickly concluded that gay marriage and abortion must have been the prime mover of vote decisions in the 2004 presidential election.

Also contributing to the conclusion that “morality” was the most important consideration in voters' minds was the success of gay marriage bans on the ballot in 11 states.³

While there was substantial variation in the initiatives, there was popular support for prohibiting gay marriage in every state where citizens were able to vote on the issue.⁴ Further, each of the initiatives passed by significant margins—between 57% to 86% of the electorate—even in the so-called blue states of Oregon and Michigan. In the strategically important battleground states, especially Ohio, the gay marriage issue was thought to dominate nearly every other consideration. Anecdotal stories abound about the efforts of the Bush campaign and Christian conservatives to emphasize the gay marriage issue. Across America, church marquees read (and pastors preached) “On Tuesday, vote God.”⁵ The Baptist convention set out to counter the alleged impact of MTV's “Rock-the-Vote” with their own blend of music and politics they called “Redeem-the-Vote.” There were reports that some traditionally Democratic African Americans mobilized for Bush solely on the basis of the gay marriage issue (Shapiro 2004).

Although gay marriage was the moral value issue most prominently featured in the 2004 presidential election, we would be remiss to ignore the perennial hot button issue of abortion. The abortion issue received somewhat less campaign coverage, but it is decidedly an issue that most expect to be an important part of the values vote. Abramowitz (1995) concluded that abortion had a dramatic influence in the 1992 election—even though Governor Bill Clinton campaigned predominantly on the importance of the economy. The role of abortion in the 2004 campaign was highlighted when some Bishops in the Catholic Church were outspoken about their unwillingness to allow Senator John Kerry to participate in communion because of his pro-choice platform, and other church officials declared that it would be a sin to vote for a pro-choice candidate.⁶

Archbishop Charles J. Chaput of Colorado, among several others, argued that Catholics voting for pro-choice candidates were sinners who were “cooperating in evil.”⁷

Criticism of the Values-Voters Interpretation

Following the flurry of media reports concluding that moral values, and gay marriage and abortion in particular, were the primary explanation for Bush's re-election, a number of academics have pointed out several problems with the moral values interpretation (Burden 2004; Jacobs 2004; Fiorina 2004). First, there are certainly reasons to challenge the validity of the exit poll question itself. The question asked respondents about the most important problem facing the country today. In this context, what do moral values as a “national problem” mean? Is it some combination of gay marriage and abortion policy? Alternatively, do moral values also reflect opposition to Kerry's perceived flip-flopping or President Bush's decision to invade Iraq? Moral values as a response category is hardly comparable to the more specific policy and political issues listed as alternative answers to the question. In this context, moral values can be defined any number of ways, making it very difficult to draw meaningful conclusions about what voters meant when they selected it as an important issue (Fiorina 2004).⁸

Moreover, although moral values was the modal response, the proportion of voters selecting moral values is not statistically different from the proportion selecting the economy (20%), and is not substantially greater than the proportion selecting terrorism (19%) and the war in Iraq (15%). So, even if we were to define moral values by the policy issues of gay marriage and abortion, these numbers suggest that these may have been just one of many voter considerations in 2004.

Further, while gay marriage was generally cited in the media as the most likely issue behind the comparatively ambiguous category of “morals/values,” the same exit polls indicate that a majority of voters (59%) favored either civil unions (34%) or legal marriage (25%) for same-sex couples. The exit polls also indicate that the electorate generally cast ballots in line with their party identification: 89% of Democrats voted for Kerry and 93% of Republicans voted for Bush. While the importance of party identification for vote choice was established by political scientists nearly a half century ago (Campbell et al. 1960), and recent evidence suggests that the effect may have grown in recent elections (Bartels 2000), the mass media generally failed to acknowledge the importance of party identification in the 2004 presidential election. Jacobs (2004) concludes, “the preoccupation with moral values and then the postmortems on John Kerry's strategy and personal limitations as a candidate has distracted attention from the two most important and enduring influences on voting—the economy and partisanship.”⁹

Burden (2004) similarly argues that national security concerns, not moral values, appear to explain changes in Bush support between 2000 and 2004.

Although there are growing doubts about the values voter hypothesis, existing empirical challenges have had to rely on aggregate data and can therefore not definitively test competing explanations of individual decision making.¹⁰

In the absence of such an analysis, the moral values conclusion continues to hold sway in the media. In fact, general discussion in the media and among many political observers continues to revolve around morality and values—even to the point where some evangelicals have called on President Bush to “pay them back” for the election victory that they provided.¹¹

Data

Ultimately, the explanations presented by the media, pundits, and even some scholars are, so far, incomplete. They fall short because they have not tested the values voters hypothesis with individual level data controlling for potential alternative explanations. In other words, do policy preferences about gay marriage and abortion have an effect on individual vote choice independent of party identification, ideology, and other prominent issues? To offer a more complete evaluation of the importance of values in the 2004 presidential election, we analyze data from a nationally representative post-election survey of more than 2,800 respondents. This post-election survey was designed to gauge public opinion on a wide variety of policy issues that tend to differentiate the political parties, and thus promises to provide an improved understanding of the role of political issues in campaigns, political evaluations, and political participation.

The Blair Center 2004 Presidential Election Survey is nationally representative and randomly sampled from the Knowledge Networks (KN) Internet panel.¹²

The KN panel consists of a national random sample of households recruited by random-digit dialing (RDD), who either have been provided Internet access through their own computer or are given a WebTV console in exchange for completing 3–4 surveys per month. Thus, although surveys are conducted over the Internet, respondents are representative of the U.S. population.¹³ By using a random probability sample for initial contact and installing web access for respondents without it, KN overcomes the most common shortfall of other Internet surveys; the KN method eliminates non-Internet coverage bias and allows researchers to accurately gauge the potential for self-selection and non-response bias.¹⁴

Findings

Given the importance that existing explanations of the 2004 presidential election placed on the issue of gay marriage, we begin by presenting attitudes toward gay marriage as measured by our post-election survey (question wording for this and other issues reported in the Appendix).¹⁵

Similar to other studies, we find that a majority of respondents in our survey opposed gay marriage (54%), while 31% supported marriage between same-sex couples, and 14% indicated they neither opposed nor supported homosexual marriage.¹⁶ Even though a majority of citizens expressed anti-gay marriage positions, when asked to rate the importance of various issues, respondents indicated that gay marriage was, on average, far less important than other prominent political issues. As shown in Figure 1, respondents were more likely to indicate that abortion, terrorism, the economy, and job security were more important than gay marriage, and a substantial number of respondents indicated that gay marriage was “not at all important.” In fact, our survey asked about the importance of some 15 different political issues, and only the issue of tort reform was considered less important on average.

To begin our investigation into the alternative explanations of the 2004 election, we look at vote choice by party identification and the issues noted above. Figure 2 reports the percentage of respondents voting for Bush (over Kerry) by party identification and issue preferences. For example, in the first set of bars we see that 86.9% of Republicans who supported gay marriage voted for Bush, as did 39.2% of Independents, and just 3.2% of Democrats. Stated differently, the majority of both Democrats and Independents who supported gay marriage voted for Kerry (96.8% and 60.8%, respectively). More interesting, of course, is that among Republicans who supported gay marriage, just 13.1% defected to Kerry. In other words, the vast majority of Republicans who disagreed with Bush on the issue of gay marriage still remained loyal to their party's candidate.

Comparing the relationship between issue preferences and vote choice, we find several interesting patterns. First, these data show that both Democrats and Republicans were exceedingly loyal to their party's candidate regardless of their position on the issues. In fact, we could not find a single issue where a majority of partisans voted against their party's candidate—even when they disagreed with the candidate on that issue. Looking specifically at attitudes about gay marriage, just 18.3% of Democrats opposed to gay marriage defected to Bush, while only 13.1% of Republicans supporting gay marriage defected to Kerry. It is also interesting to note that across all issues, Republicans were consistently more loyal to their party candidate than were Democrats, an asymmetry identified in previous elections as well (Hillygus and Jackman 2003). Democrats who held issue positions at odds with their candidate appear to have been somewhat more willing to defect from their party and vote for Bush.

This overall pattern of party loyalty should not obscure the fact that there were notable variations in party defections by issue. In comparing defections and loyalty across the various issues, gay marriage and abortion were far less likely to discriminate vote choice than were opinions on the war in Iraq and the economy. In contrast to the rather small rates of defection on the issue of gay marriage reported above, some 33% of Republicans who believed the Iraq war was “not worth the costs” voted for Kerry, while 48.3% of Democrats who believed the Iraq war was “was worth the costs” voted for Bush. Likewise, 21.5% of Republicans who believed the economy had gotten worse in the last year voted for Kerry; while 32.8% of Democrats who believed the economy had improved voted for Bush. Thus, partisans who disagreed with their party on the issue of the Iraq war, terrorism, or the economy were more likely to defect than were partisans who disagreed with their party on the issue of gay marriage or abortion.

Similar to their partisan counterparts, the issue that best discriminated vote choice among Independents was evaluations of the Iraq war. A whopping 90.4% of Independents who thought the Iraq war was worth the costs supported Bush on Election Day, compared to just 22% who thought the Iraq war was not worth the costs—a 68.2% point gap. Among Independents who believed the economy had improved over the past year, 71.9% voted for Bush while only 30.8% who believed the economy had declined over the past year did—a 41.1% point gap. In contrast, the vote gap was far smaller for the issues of gay marriage (29.9% points) and abortion (23.1% points). In other words, opinions about the economy and the war in Iraq did a better job of differentiating vote choice (for both Independents and partisans) than either gay marriage or abortion attitudes.

Although the data presented in Figure 2 provide an initial test of the importance of gay marriage in shaping vote choice, a more rigorous examination simultaneously controls for alternative explanations. To do this, we estimate a multivariate logit model predicting a vote for Bush (over Kerry), allowing us to compare the effects of the economy, evaluations of the war in Iraq, abortion, approaches to terrorism, and gay marriage.¹⁷

For controls, we include standard demographic and political variables, including income, gender, race (black indicator variable), party identification (including leaners), and ideology.¹⁸

In addition to examining the effect of moral issues on vote choice in the general population, we also look at the importance of these issues among various groups where values-based voting may have been particularly influential in shaping election outcomes in 2004. The first model predicts vote choice among all respondents in the survey. The second model restricts the analysis to only those living in the South¹⁹

—the Bible Belt so heavily courted with moral values messages during the campaign. The third model is estimated for political independents, who might be expected to weight issue based considerations more heavily than party loyalties. The fourth model is estimated for respondents living in battleground states, with the rationale that the election outcomes were ultimately decided by votes in a relatively few number of swing states.²⁰ Given the intensity of the race in battleground states, voters there likely had more issue information than voters in other states, so we might expect issues to play a more prominent role in decision making in battleground states. In the final model, we estimate the effect of gay marriage and other issues for just those respondents living in states with gay marriage bans initiatives on the ballot. The results are presented in Table 2.

Examining the results for all respondents (Model 1), we see that several issues were significant predictors of voting for Bush (over Kerry). Specifically, respondents' evaluations of the war in Iraq and the economy, and opinions about a multilateral approach to terrorism were all significant predictors of vote choice—even controlling for demographics, party identification, and ideology. Notably, opinions on the Iraq war had the largest effect, with a coefficient more than four times that of gay marriage, and even larger than the effect of party identification.²¹

In other words, Republicans were indeed significantly more likely than Democrats to vote for Bush, but the difference was not as large as that between respondents who believed the Iraq war was worth it and those who did not. Likewise, self-identifying conservatives were more likely to vote for Bush than self-identifying liberals, but the size of the effect is again smaller than retrospective evaluations of the Iraq war or the economy. With respect to the effects of demographic characteristics, it is worth noting the absence of a gender gap—controlling for all else, women were no more likely to support Kerry than were men. The well-documented racial gap remains, with African Americans significantly more likely to vote for Kerry over Bush. This effect is strongest in battleground states, perhaps reflecting Kerry's concentrated campaign efforts at targeting blacks in those states.

Among those respondents living in the states of the Old Confederacy (Model 2), one of the most socially conservative regions of the country, the results are quite similar, except that the effect of attitudes toward abortion no longer reaches standard levels of statistical significance. The relationship between gay marriage and a vote for Bush is somewhat stronger among Southerners than for the nation as a whole, but the effect does not rival the impact of the Iraq war or the economy.

We find that among Independents (Model 3), respondents in battleground states (Model 4), and respondents in states with an anti-gay marriage initiative was on the ballot (Model 5), that abortion attitudes and attitudes toward gay marriage had no significant effect on presidential vote choice when party identification, ideology, and demographics are taken into account. In contrast, the effect of the Iraq war, the economy, and terrorism remain critically important to voter decision making among all of these groups.

To better illustrate and compare the substantive effects of these various issues, we estimate the change in the predicted probability of voting for Bush across the range of issue preferences.²²

Given the similarities of the various estimates across the models, we provide graphs for just three groups: all voters, Southerners, and Independents.

Figure 3 illustrates the predicted probability of voting for Bush across issue preferences for all voters, holding all other variables at their means and indicator variables to zero.²³

Not surprisingly, respondents expressing pro-administration positions on the five issues studied in this investigation were significantly more likely to vote for Bush, while respondents holding anti-administration positions on these issues were more likely to vote for Kerry, even controlling for party identification, ideology, and demographics. Although all of the issues are related to presidential vote choice, the relationship is much stronger (i.e., the line is steeper) for the war in Iraq, support for a multilateral approach to terrorism, and evaluations of the economy than for either gay marriage or abortion. The first difference, or change in predicted probability, between those with the most anti-administration views compared to those with the most pro-administration views was a whopping 74% points on the issue of Iraq (92.4% minus 18.4%), 68.5% points on evaluations of the economy and 55.1% points on support for a multilateral approach to terrorism, compared to just 23.6% points on the issue of abortion, and 22.2% points on the issue of gay marriage.

Although these findings suggest that Bush did not benefit nearly as much from moral issues as initially suspected, it is interesting to note that Bush generally found support among those with neutral positions on the various issues. Looking at the predictions for those without strong beliefs one way or the other on the issues, we find that they were always predicted to vote for Bush even if they took a moderate position on the issue. The only exception is gay marriage, with those who were neutral on the issue of gay marriage more likely to support Kerry.²⁴

It appears that when a voter was ambiguous or in doubt about an issue, they ultimately put their support behind the incumbent. In other words, those torn on an issue appeared to revert to the status quo. Consequently, it may be time to modify interpretations of the election outcome: rather than concluding that Bush received a moral mandate, we should instead recognize that Bush simply received a second chance among those who were not ideologically polarized on the prominent issues of the campaign.

Graphing the predicted probabilities for Model 2, we find a similar pattern as indicated in Figure 4. Even in the conservative South, the issues of the war in Iraq, the economy, and terrorism did a better job of differentiating vote choice than did gay marriage or abortion preferences. Our model predicts that respondents with the most pro-administration positions on the issue of the Iraq war had a 93.7% chance of voting for Bush, compared to a 21.3% chance among those with the most anti-administration viewpoints—a difference of 72.4% points. That difference was 66.6% points for evaluations of the economy, 45.7% points on support for a multilateral approach to terrorism, and 31.4% points on the issue of gay marriage (the difference was not statistically different from zero for the issue of abortion). Overall, then, the influence of moral issues pale in comparison to the decisive influence of the Iraq war, the economy, and terrorism even in the area of the country thought to be the most receptive to the moral values campaign messages.

Finally, in figure 5, we graph the predicted probability of voting for Bush across the range of issue preferences for Independents. It might be hypothesized that Independents weigh the issues of the campaign more heavily in their decision making since they are likely to be less loyal to a particular party candidate. The relative impact of the various issues is particularly dramatic among Independents, with the graph illustrating a clean separation of values and non-values issues. As in the previous figures, respondents' attitudes regarding the war in Iraq, the economy, and terrorism have much steeper slopes than the issues of abortion and gay marriage (which did not reach statistical significance). Overall, those who held anti-administration positions on the issues of the Iraq war, the economy, and terrorism were almost sure bets to vote for Kerry, while those holding pro-administration positions on these issues were almost sure bets to vote for Bush. The effects of the so called value issues of abortion and gay marriage show little change across the range of responses to these issues. Put simply, to correctly predict who an Independent voted for in the 2004 presidential election, we need to know their evaluation of the Iraq war, not their opinions about abortion or gay marriage. To the extent that moral values played a role in 2004, it was to reinforce support among partisans not to attract most Americans to a particular candidate.