Guth, J., et al. (2006). "Religious Influences in the 2004 U.S. Presidential Election," Presidential Studies Quarterly 36(2): 223-242.
Religious identity is considered to come in two different flavors in American politics: ascriptive (assigned from birth as part of personal identity) and prescriptive (determined in its details via interaction with other areas of life, such as politics. The key difference between these two approaches is which direction the causal arrow points. Those who see religion as ascriptive see it as shaping attitudes; those who see is as prescriptive see attitudes as determining religion. Guth and his coauthors see both perspectives as key to understanding the role religion plays in presidential elections. However, they see the religious aspects of current survey data as limiting, in that it assumes that one and only one of the perspectives fits the actual world (which one depends on the survey writer). Using the Pew Survey on Religion and Politics, which has the benefit of capturing data that describes both axes of religious identity, they analyze the political preferences of various religious groups, including noting the breakdown of differences between "traditionalist" and "modernist" voters in certain groups. They then examine how the two major parties' political coalitions are built out of these religious groups, using a) the size of the group's voting-age population; b) its turnout rate; and c) the proportion of their vote they give to each party. Ultimately, their findings more or less confirm the conventional wisdom, and demonstrate that both theories have salience in determining presidential politics.
My take: I more or less knew this long before reading this paper, but it's nice to have confirmation that religious identities continue to play a role in presidential politics.