We next turned to the second key question concerning the role of religion in 2004—were highly religious voters responsible for Bush’s more decisive victory?—by examining more closely the trend, already noted by Pew researchers, toward a slightly diminished effect of religiosity as compared to 2000.

One of the strengths of the logistic regression technique is the ability to calculate the predicted probability at which different segments of the electorate (including very fine-grained segments) voted for Bush. We calculated the probability that men and women of different religious persuasions and levels of religiosity voted for Bush in 2000 and 2004. These probabilities were based on the underlying logistic regression model, and needed to be calculated for specific values for all variables in the model. For this purpose, we picked individuals of a specific gender, religion, and level of religiosity but otherwise average characteristics, choosing whites, ages thirty to sixty, in non-union households, with no college education and household incomes of between $30,000 and $100,000, who varied in gender, religious affiliation, and religiosity.

The predicted probabilities lend real insight into the ways in which Bush support shifted between 2000 and 2004, and they reveal some fascinating gender differences. Consider Protestant men and women. There is no question that religious Protestants were more likely than their secular coreligionists to support Bush in both 2000 and 2004. What is interesting, though, is that the difference between religious and secular Protestants was of the same magnitude among men in both elections, but actually decreased among women in 2004 compared to 2000. As shown in Figure 2, there was little change in the levels of Bush’s support among men.

But secular Protestant women were more likely to vote for Bush in 2004 than in 2000. Only 49 percent of female Protestants (white and with average characteristics on other variables) were predicted by the model to support Bush in 2000. This rose to 60 percent in 2004—a major increase over 2000, especially in light of an overall shift towards Bush of roughly 3 percent. We also found that women who attended religious services more than once a week were predicted to support Bush strongly in both 2000 (83 percent) and 2004 (87 percent), and the increased support for Bush among Protestant women was most pronounced among secular women. 

We can therefore conclude that, because Bush’s support remained fairly constant from 2000 to 2004 among religious Protestant women and increased among secular Protestant women, the difference in his support between these two groups actually diminished somewhat in 2004 when compared to 2000.

Similar gender differences emerged among both Catholics and Jews. Figure 3 depicts the predicted probabilities of Bush support among male and female Catholics, and Figure 4 shows them for Jews.

First, religious Catholics and Jews were more likely to vote for Bush than their secular counterparts, regardless of gender. This makes clear that Bush owed his support among religious voters in both elections to these voters in general, not just to religious Protestants. Second, Figure 3 highlights the diminished role of religiosity among Catholic women in 2004. Much like secular Protestant women, secular Catholic women stand out as being more pro-Bush in 2004 than 2000. Roughly 52 percent of secular Catholic women (white with otherwise average characteristics) voted for Bush in 2004, compared to 43 percent in 2000. In contrast, Bush support remained strong among highly religious Catholic women who attended religious services more than once a week: 73 percent were predicted to vote for Bush in 2004, compared to 71 percent in 2000. Once again, these findings underscore the diminished effects of religiosity among women in 2004.

Bush support among Jews presents a similar picture, although in this case the diminished effects of religiosity were observed among both men and women. Similar to Christian women, secular Jewish women were predicted to support Bush more strongly in 2004 than in 2000, whereas their religious counterparts were actually somewhat less likely to do so in 2004. In this instance, a similar trend was also observed among Jewish men. Twenty-one percent of secular Jewish men (who were white and average on other sample characteristics) were predicted to vote for Bush in 2000. This increased to 35 percent in 2004. Likewise, predicted support for Bush among secular Jewish women went from 10 percent in 2000 to 25 percent in 2004. Overall, secular Jews were much more supportive of Bush in 2004 than in 2000.