After reviewing these results, we chose to explore further the basis of evangelical support for Bush, this time by examining motives less grounded in policy. To this end, we analyzed the results of a second poll of Tennessee residents, conducted in February 2005.

The poll included a series of questions asking respondents what they expected the Bush administration to do during its upcoming term. Borrowing their structure from a Gallup battery, each of these nominal measures described an outcome and asked whether the respondent thought the Bush administration would be able to make that outcome occur. The outcomes considered included improving moral values in the United States; increasing belief in God; amending the Constitution to ban gay marriage; outlawing abortion; ending restrictions on prayer in public schools; ending restrictions on displaying the Ten Commandments in public buildings; and securing God's blessing on the United States. The questions were presented in random order during the interview.

Responses to these items were crosstabulated with whether or not respondents were evangelical Bush voters, as well as with each of the other expectation items. Among the sample's evangelical Bush voters, 70 percent expected the Bush administration to improve the country's moral values, 69 percent to secure God's blessing on the country, and 66 percent to increase belief in God. Generally fewer expected the administration to ban gay marriage (62 percent), end limits on displaying the Ten Commandments (51 percent), end limits on school prayer (39 percent), and ban abortion (36 percent). The association between these expectations and the respondents' being evangelical Bush supporters was significant for every item except "ban abortion" and "improve moral values."

The results summarized in Figure 1 support an affirmative, though qualified, answer to the question of whether voting preferences in the 2004 presidential election could be predicted by a single measure of self-identification as an evangelical or born-again Christian. Among a random sample of Tennessee adults, such self-identification did, indeed, strongly predict presidential voting preferences--but only among whites, and most strongly among better-educated whites. A significantly higher proportion of people who self-identified as evangelicals said they had engaged in witnessing behavior. The finding suggests some validity for the evangelicalism measure, given the importance evangelical theology assigns to witnessing.

Additionally, the evangelical measure served as a better predictor of voting behavior than frequency of worship service attendance, indicating a capacity to single out evangelical Christians from devout Christians in general. The relationship between self-identifying as an evangelical and taking a literalist view of the Bible, though, suggests that the evangelicals involved in the study tended to be of the fundamentalist or neofundamentalist variety. This result serves as a caution against using the study's findings to characterize all evangelicals. Overall, though, the analysis indicates self-identification as an evangelical is a reasonably valid measure that pollsters would be wise to include in polls and surveys aimed at predicting voting behavior in presidential elections.