Religion, Income, and Voting

In chapter 6 of Red State, Blue State we talked about some of the fascinating trends in religion and voting, in particular the big jump in the religious/secular voting gap starting in 1992, in parallel with the big jump in the correlation between voting and attitudes on social issues. For example, this plot showing the average position on abortion among Democrats, Independents, and Republicans:

Now that we have the Pew pre-election polls, we could look at what happened in 2008. To start with, we found that the more you go to church, the more likely you were to vote Republican:

No surprise but it's good to start with the basics.

Now let's break things down by religion and denomination:

(The size of each circle is proportional to the number of people represented in the survey. In particular, most of the people who attend church more than weekly are born-again Protestants. Also, some nonreligious people go to church; I assume this is for family reasons but I haven't examined the question in detail.) As in 2004, churchgoing is more strongly associated with Republican voting among Catholics and born-again Protestants than among non-born-again Protestants, and all three of these groups represent approximately equal proportions of the population. Whassup with those non-born-again Protestant regular churchgoers: didn't they get the memo? The patterns for Jews and Mormons are also interesting (and consistent with 2004). Finally, you can see that the "no religion" people continued to be a strong Democratic bloc.

What about income and voting? In Red State, Blue State, we talked about the pattern, consistent with the story of "post-materialism," that religious attendance is a more important predictor of vote choice for the rich than the poor. Here's what we see in 2008:

This is similar to what we saw in 2000 and 2004.

Finally, we can look at voting and income for different religious groupings:

Within the "no religion" group, income is associated very weakly with how you vote. This is consistent with the idea that social issues are more important for richer voters; thus, the richer people with no religion are remaining on the Democratic side because they don't like the Republican Party's socially conservative and religious orientation.

Discussion of methods

I think my style is off-putting to some readers because, rather then state my point right away, I often will lay out an argument and then look at it from different angles. (This isn't such a problem with my own blog because I've gradually built up a readership over several years, and the readers/participants are familiar with my style. But it is an issue when I drop these long graphics-filled posts into Nate's blog, whose readers are more used to his and Sean's more topical approach.) Also, I don't always have a strong conclusion; sometimes I just want to put the data out there and let you draw your own inferences. Synthesis is great, and we try to do some of it in our book, but I think it's also possible to make a contribution just by putting some data out there.

One final remark. After seeing my earlier graph-laden posts, several commenters thought I should be controlling for more variables--for example, in my graphs of political ideology and partisanship among sports fans and nonfans, people wanted me to control for sex. That would be fine--I have no objection to doing separate analyses for men and women, and in any case the GSS data I used are public and so anyone can feel free to do this analysis and send it to me--but for the purpose of answering the original question, the data combining both sexes were fine.

Recall that the original hypothesis of the conservative commentator was that sports fans are disproportionately conservative but not especially Republican, and thus he thought they represented a ripe low-hanging fruit for the Republican Party. Actually, though, the data showed the opposite: sports fans were more likely than the general public to be Republican but they were not particularly likely to be conservative.

So this simple data analysis revealed something. (And, yes, I agree that a serious study of this issue would require the use of more data than a single opinion poll from the 1990s.)

I do run regressions, and I'll report some of these on occasion. But it's a misconception to think that simple comparisons are meaningless or that regression analysis is always necessary. One thing sophistication can give you is an appreciation for the simple things in life.