The Pew Research Center has released a very interesting analysis of voting patterns in the last Presidential election. In addition to the well known differences between religious groups, the analysis notes that even among adherents to the same faith, support for the Republican candidate grew with the frequency of worship attendance:
An analysis of national exit polls from 2004 shows there is not one but two religion gaps -- one based on religious affiliation and the other based on frequency of attendance at worship services. Recent surveys by the Pew Research Center provide evidence that both of these religion gaps are at work as the public evaluates the candidates for the 2008 presidential race. The surveys also indicate that the Democrats may be doing better than they did in 2004 among some religious groups.
. . .
Voters who report attending religious services at least once a week -- regardless of religious affiliation -- tend to vote more Republican. Those who say they attend religious services less often (termed "less observant" for the purpose of this analysis) tend to vote more Democratic.
In the 2004 presidential election, exit polling by the National Election Pool found that religious affiliation and frequency of attendance at worship services had a larger impact than many other, better-known factors, including the "gender gap" between men and women and the "class gap" between the most and least affluent voters.
The difference in the votes of evangelical Protestants and black Protestants is an example of the affiliation gap; 79% of evangelical Protestants voted for President Bush, compared with 14% of black Protestants -- a difference of 65 percentage points. . . . Members of some religious groups, including mainline Protestants, divided their votes more equally between Bush and his Democratic opponent, Sen. John Kerry. But large differences separated mainline Protestants from evangelical Protestants in their support for Bush (a 25-percentage-point gap) and mainline Protestants from black Protestants in their support for Kerry (a 40-point gap).
The religion gap based on attendance, although not as stark as the affiliation gap, is also significant. Within all the major religious traditions surveyed, people who attended religious services at least once a week voted more Republican than did their less-observant counterparts within the same religious affiliation.
Among evangelical Protestants, for example, Bush received support from 82% of those who attended services at least weekly, compared with support from 72% of those who attended services less frequently (a gap of 10 percentage points). Comparing evangelical Protestants' support for Bush with that for Kerry, Bush held a 44-point advantage over Kerry among evangelical Protestants who attended services less than weekly. However, the advantage for Bush increased to 64 points among evangelical Protestants who attended services at least weekly.
The pattern also held true within religious traditions that generally supported Kerry. Among black Protestants, Kerry received greater support from those who attended services less than once a week than from those who attended services weekly or more often (92% vs. 83%, a nine-percentage-point gap). Comparing black Protestants' support for Kerry with that for Bush, Kerry held an 84-point advantage among those who attended services less than weekly. However, the advantage decreased to 66 points among black Protestants who attended services at least once a week.
The gap based on frequency of attendance at worship services also is apparent even within religious traditions that split their votes somewhat more equally between the two candidates. For example, among non-Latino Catholics, Bush received much greater support than did Kerry from those who attended services at least weekly (a 24-point gap), while Bush's support among those who attended services less than once a week was not as overwhelming (a six-point gap).
Interestingly, the analysis also notes that the 2008 Democratic candidates appear to be already attracting more religious support than John Kerry:
There is some preliminary evidence that the affiliation gap again is at work in the lead-up to the 2008 presidential election.
A January 2007 Pew survey, for example, asked people if they would most like to vote for a conservative Republican candidate, a moderate Republican candidate, a moderate Democratic candidate or a liberal Democratic candidate for president in 2008. As in 2004, religious affiliation appears to be an important factor in the responses: Evangelical Protestants were the most likely to back a Republican presidential candidate. Conversely, the lowest level of support for a Republican came from unaffiliated voters and a composite of religious groups -- including black Protestants -- that were too small to break out individually in the Pew survey. Falling in between were mainline Protestants and non-Latino Catholics; their support for a Republican was lower than that of evangelical Protestants but higher than that of the unaffiliated.
The gap based on frequency of attendance at worship services also may be at work. Regardless of affiliation, weekly attenders were more likely to back a Republican candidate than were the less observant. The survey also showed, however, that many of the constituencies that backed Bush in 2004, including less-observant evangelical Protestants, are more likely to support a Democratic candidate in 2008. In addition, the Democrats are attracting even stronger support from religious constituencies that backed Kerry in 2004. For example, 38% of weekly attending non-Latino Catholics voted for Kerry in 2004, but 52% say they would "like to vote" for a Democrat in 2008.
Read it all here. It is interesting that frequency of worship was positively correlated with voting Republican even in denominations that supported Democrats. This may reflect a perceived hostility to faith by the Democrats--a perception that the current crop of Democratic candidates are fighting quite hard to change. this may explain the better numbers for the Democrats in 2008.