The diminishing impact of religiosity between 2000 and 2004 among Protestant, Catholic, and Jewish women is summarized in Figure 5, with trends for men provided as a comparison.

It is clear that among men, Jews were the only group for whom religiosity played a reduced role in 2004. In contrast, religiosity had diminished impact in 2004 among Protestant, Catholic, and Jewish women. For both men and women, the largest jump in Bush support came from secular Jewish men and women (who still supported Kerry over Bush, on average). Predicted Bush support was up by roughly 15 percent in 2004 among Jews who never attended religious services, but it also increased by roughly 10 percent among secular Protestant and Catholic women. Increased support for Bush among secular women may have helped Bush improve his electoral performance in 2004 over 2000, and may prove to be the true but largely unsung story of the 2004 election.

One other point should be made about the logistic regression results: They underscore the stability of vote choice in the electorate between the two elections. For the most part, demographic factors had roughly the same influence on Bush support in 2000 and 2004, with a few interesting exceptions. Blacks and older voters were equally unsupportive of Bush in 2000 and 2004, and members of wealthy households earning $100,000 or more were equally supportive. As has been noted by several analysts, however, the gender gap was somewhat muted in 2004, consistent with findings reported here of increased support for Bush among secular women. Voters under the age of thirty switched from Bush support in 2000 to slight opposition in 2004. Hispanics, members of union households, and the college-educated were even more opposed to Bush in 2004 than in 2000. The logistic regression techniques thus shed further interesting light on differences in vote preferences between the two elections which can be obscured in bivariate analyses because some of these factors—such as education and income or religiosity and race—are confounded.

Religion played a major role in the outcome of the 2004 election, just as it did in 2000. Voters who said they regularly attended religious services were much more likely than their secular coreligionists to support President Bush in both elections. Bush’s support was strongest among religious Protestants. But it was also pronounced among religious Catholics and, to a lesser extent, religious Jews. These findings can be seen in simple bivariate relationships. The exit poll results show that more than six in ten voters who attended religious services at least once a week voted for Bush, while over six in ten voters who never attended services chose Kerry. 

The use of logistic regression in both the 2000 and 2004 exit poll data allowed us to go beyond the simple conclusion that religion played an influential role in both elections, to address how Bush support shifted between 2000 and 2004. Our findings accentuate the general importance of religiosity to both election outcomes but contravene the notion that religious voters accounted for Bush’s slight increase in electoral support in 2004. In our analyses, the influence of religious-service attendance on vote choice in fact diminished from 2000 to 2004 among women, with Bush attracting the votes of more secular women in 2004 than he did in 2000.

We cannot say with certainty what factors were responsible for increased Bush support among secular women and Jews more generally in 2004, but one possible factor was support for the administration’s actions on terrorism and its handling of events in the Middle East. In that sense, our findings lend tangential support to the findings of other researchers such as Abramowitz, Barry Burden, Gregory Lewis, Sunshine Hillygus and Todd Shields, and Merrill Shanks and colleagues. Could these findings indicate that religion is diminishing as a factor in presidential vote choice? That assumption might be premature, but the degree to which religiosity affects vote choice has declined from 2000 to 2004, and bears watching in 2008. A comparison to pre-2000 presidential elections would also be instructive.

Leonie Huddy is an associate professor of political science and director of the Center for Survey Research at Stony Brook University; Stanley Feldman is a professor of political science at Stony Brook University; and Sarah Dutton is deputy director of surveys at CBS News.