By Paul A. Djupe

In recent weeks, there have been anecdotal reports that a year out from Trump’s election people have returned to church for many of the usual reasons people turn to community – to celebrate and to cope. As the Washington Post’s Julie Zauzmer recently remarked:

The pews were packed in churches across the country on the Sunday after the election, filled with Trump supporters giving thanks for the turning of a page in America, with stunned liberals seeking solace, and with many who simply felt compelled to be together in community in an emotional moment.

Though she had no systematic data, her conclusions were measured: “There’s been no big turn in the tide…But there was a slight shift.” I’ll bet that’s right, but we need some data to make sure.

Could it be that partisan tides shape the pews of America’s congregations? While it seems easy to discount such an idea, there is a sensible logic underpinning it as Michele Margolis lays out in her 2016 article in Public Opinion Quarterly. In this view, religion is a source of compensatory control – “when government fails to provide order and structure, people look elsewhere to fill that void, including to religion” (2016: 718). Put in terms of the 2016 election, we would expect that those Clinton supporters who experienced loss would seek out some way of coping, which might entail a turn (or return) to religion in the aftermath. On the flipside, with the GOP in control, we might see some slackening of religious behaviors by conservative Christians.

Fortunately, I have access to data well-suited to testing this idea. In the fall of 2016, Anand Sokhey, Amanda Friesen, and I surveyed a sample of American adults in September and again in November. Using these data, we were able to sort out the degree to which disagreement over Trump drove people out of their churches. We also found that evangelicals could have been persuaded against Trump, but almost no evangelical clergy were talking about him.

Then, I collaborated with Andy Lewis, Ryan Claassen, Jake Neiheisel, and Anand, to resurvey a portion (~400) of the fall sample again in June 2017. In each wave, we asked detailed religious attachment and attendance questions so that we could follow the religious decision-making of a cohort across the election cycle and beyond. This is not a large sample, so we should take these results with some caution. However, it has the advantage of tracking the same people across just the period we’re interested in. In the run up to the election, Democrats were widely expected to win so that the post-election period was particular shocking to them. Therefore, we should see movement going in different directions in the pre- versus post-election periods.

This is over a rather short span of time, so we should not expect to see tremendous amounts of change. To give you some perspective, the mainline Protestant share of the population has been projected by PRRI to have dropped 5 percent (18 to 13%) from 2006 to 2016, which is a half percent a year.[1] Across the election, Mainline Protestants dipped 1.6 percentage points by our post-election survey in November and then regained some members by June. Given their historical pattern of decline, it is remarkable that the mainline did not lose members through this period. Catholic identifiers dipped several points by November and then fully rebounded by June. There was very little change in the percentage of evangelicals and perhaps a slight climb of about a point from September to June.

Registering the downs and ups of the more moderate to liberal religious groups, the religious nones increased their population share from September to November (by 2%) and then declined by .5% by June. Thus, the nones experienced a net gain from the election, but the pattern was not one of uniform, ‘secular’ gain.

Figure 1 – The Movement Among Religious Traditions Across Three Waves – September 2016 to June 2017

Let’s shift to look at church attendance – a reported behavior much less drastic than de-identifying with a faith tradition. Figure 2, below, shows the change from September to November. The plot tells the same basic story as Figure 1 if we make assumptions about the partisan affiliation of religious groups. That is, Democrats registered a sizeable decline (.2 on a 5 point scale = 4%) while independents and Republicans registered effectively no change (or very small declines). Remember that most everyone expected Clinton to win, so this pattern aligns with the theory.

Figure 2 – Change in Reported Church Attendance from September to November 2016 by Partisanship

Figure 3, then, shows the change from November to June. Again, consistent with Figure 1, this shows a rebound among Democrats (by about 2% of the scale), while Republicans show a decline of the same magnitude. If compensatory control is in effect, then we would expect Democrats to increasing access institutions that would help them cope, whereas Republicans would have less need for that.

Figure 3 – Change in Reported Church Attendance from November 2016 to June 2017 by Partisanship

But the Republican decline is especially telling. They showed no earlier increase from which to decline (which might be read as “regression toward the mean”) and Easter should have propped up those numbers (our survey, administered very early in June, might have picked up the tail end of the Easter bounce). This is where the compensatory control hypothesis factors in clearly – it is possible that united partisan control over the federal government weakened their need for religion.

With a few counter hypotheses lingering (like regression to the mean and yearly religious calendars), we can turn to another variable to nail down the mechanism: their emotional responses to politics. Those seeking compensatory control would show emotional stress from anxiety or anger. To capture this, in the June survey we asked, “When you think about politics in the United States in 2017, about how often would you say you feel angry?” It turns out that anger helps to maintain church attendance rates, while the lack of anger helps to reduce it as shown in Figure 4. The top line (those “often angry”) shows no change in church attendance (y axis) regardless of how often they attended church in November (x axis). However, the black line shows that more frequent November attenders posted substantial declines in June attendance if they reported not being angry.

Figure 4 – Feelings of Anger About American Politics Helps to Maintain Church Attendance Rates

We can also look at this from yet another direction by asking what sort of partisan environment Democrats and Republicans would be seeking out after the election. The sorting hypothesis is that they simply look for partisan-reinforcing social settings. The compensatory hypothesis is that social support would be particularly motivating for Democrats, and would be superfluous (or even demotivating) for Republicans. Figure 5 tests this by assessing whether a match between clergy and community support for Trump leads to different levels of church attendance for Democrats and Republicans. Community agreement (a difference of zero) leads Democrats to boost their church attendance, while Republicans reduce theirs. This is not sorting. This is compensation.

Figure 5 – Community Agreement Boosts Democrats’ Attendance, Reduces Republicans’

We rarely get a look at how people change their behavior and identifications over such a short period of time that overlaps with an election. Often religious measures are included in one wave of a panel (usually the first) under the assumption that those behaviors, beliefs, and identities are stable. This evidence does not completely undermine that claim because the change we see is not enormous, but the results do suggest that religious decisions are distinguishably fluid in a polarized political environment. People turn to religion for comfort and coping and turn away from it when those needs are not present. Layer on top the role of disagreement in driving marginal members out of churches and you begin to understand how risky politics is for modern clergy and churches.

Paul A. Djupe, Denison University Political Science, is an affiliated scholar with PRRI, the series editor of Religious Engagement in Democratic Politics (Temple), and co-creator of religioninpublic.blog (see his list of posts here). Further information about his work can be found at his website and on Twitter.

---

Notes

1. Of course the translation from denomination size to population statistic is complicated. Since the US population has been growing by about 0.7% a year, if the denomination membership remained flat, it would register as a loss relative to the population. In this case, the population has been steadily growing (though at the lowest rate since 1937, apparently) AND mainline denominations have been shrinking, which together result in sizable population-level losses. You can find a compilation of membership changes by select denominations here; The ARDA has a much more authoritative and comprehensive compilation here.