“Mainline” churches are emptying. The political effects could be huge.

In April, the New York Times’s Ross Douthat wrote a column arguing — appropriately for a Sunday opinion piece — that liberals need to go to church. More specifically, he argued: “The wider experience of American politics suggests that as liberalism de-churches it struggles to find a nontransactional organizing principle, a persuasive language of the common good.”

For a theologically confessional Lutheran and politically conservative Republican like me, this is an interesting suggestion. How much of the cultural and political change we have observed in the past 20 years can be explained by the quiet death of the “mainlines”?

For readers not familiar with Christian denominational classifications, “mainline” is the term that students of American religious life use to refer to more theologically or politically liberal white-majority churches. Since Sunday morning remains America’s most segregated hour, experts separate “historically black Protestant” churches into their own group, and then divide the white churches into various segments.

Mainline churches today remain diverse but tend to share a cluster of overlapping values: They may endorse same-sex marriage in the church, hold more lenient views on extramarital sex, confess less exclusivist views of eternal salvation, and focus more political efforts toward Christian social relief rather than Christian moral teachings; many mainlines have formal fellowship agreements with one another.

These churches are dying off at a very quick pace. Within the Presbyterian family, the “mainline” Presbyterian Church (USA) is shrinking while the “evangelical” Presbyterian Church of America is growing. Within the Lutheran family, the “mainline” Evangelical Lutheran Church in America (yes, the “Evangelical Lutherans” are mainline, not evangelical; one of many absurdities in religious verbiage) is shrinking rapidly while “evangelical” Wisconsin or Missouri Synod Lutherans are either holding steady or shrinking slowly. Within the Methodist family, “mainline” United Methodists are shrinking while “evangelical” Free Methodists are holding steady. Both sides of the Baptist family tree are shrinking, but the mainline side is shrinking faster.

Much of this is driven by natural population aging, given that the evangelical churches tend to attract families with more kids and have been somewhat more popular with immigrants than the mainline churches. But much of it has also been driven by conversion: In 2014, the Pew Research Center found that 1.7 peopled converted away from mainline Protestant groups for every one convert in, while evangelical Protestants had 1.2 people convert in for every leaver.

Many people may think these shifts are only of interest to church workers or the devout. But Douthat suggested, and I will elaborate on the view, that liberalism itself might be put on a stronger intellectual and social footing if it reembraced churchgoing. (Douthat also argued that going to church will ultimately help liberals and progressives attain what we might call eudaimonia; I won’t address that in depth here but, as a Lutheran I must take issue and suggest that, actually, while being a Christian is swell and all, “you will have troubles.” But I digress…)

Douthat kept the discussion on an abstract level, but could there be empirical ways to assess the political effects of the failure of the mainlines — evidence that people who leave such churches start to vote differently, for example? An admittedly crude but potentially telling question we might ask is: Are places dominated by these declining mainline denominations uniquely likely to turn toward a populist candidate such as President Trump? After all, if, when mainlines decline, they truly leave their old provinces bereft of “a persuasive language of the common good,” that does seem like it could potentially lead them to more cynical or extreme politics.

Examining the connection between mainline churches’ decline and voting

Spatial data about religion is hard to come by. Pew conducts semi-regular surveys about American religious life, but they have large margins of error for the subnational level and no county data at all. From the 1850s to the 1930s, the US Census Bureau conducted surveys of American religious bodies, assessing their property holdings, membership, attendance, and various high-level demographic and structural features, but those were discontinued during the Great Depression because of cost, privacy, and dwindling interest.

Since the 1930s, the Association of Religion Data Archives has collected as much data as it can to continue these national membership series, but they have a limited set of subnational data. Since the 1950s, however, the Association of Statisticians of American Religious Bodies has conducted periodical censuses of religious bodies in an attempt to get a complete picture of American religious life. In its 2010 census, it also tried to account for less denominationally organized religious membership, including nondenominational churches. This data is available by county, and is the data I will rely on when discussing the spatial distribution of American religious belief.

There are many ways to assess the tie between religion and voting. We can’t use this data to see who voted for which candidate exactly, but we can look at which places had high densities of certain religious groups, and how they voted.

My preferred method to do this is to look at the average change in the county-level Republican share of the vote per member of each denomination. To calculate it, I simply assign each adherent for each denomination the change in the GOP’s share of his or her county’s votes from 2012 to 2016, add these all up, then divide by the total national number of adherents for that denomination. The mathematical intuitions underlying this formula can be hard to wrap your head around, but the very simple upshot is an estimate of how the vote changed from Romney to Trump in the areas where each denomination’s members tends to live.*

The association between a given denomination and the swing toward Trump varies, but, broadly speaking, areas with a high density of mainline church members in 2010 saw a swing toward Trump versus Romney’s share in 2012, while areas with a high density of evangelical churches saw Trump underperform relative to Romney. If we weight together all the mainline and evangelical groups in my sample, we get a simplified estimate of how Trump performed versus Romney in mainline, evangelical, and Catholic areas:

As you can see, again, evangelical and Catholic areas saw underperformance by Trump relative to Romney, while mainline areas saw overperformance. This is notable because by far the largest evangelical group, Southern Baptists, does exhibit some swing toward Trump — but the swing away from Trump among other evangelicals more than offsets that move. In other words, Southern Baptists may not be typical of American evangelicals.

This is interesting, because it suggests that there’s some extra complexity to the story that evangelicals came out in support of Trump. It’s true, a large share of evangelicals who voted cast a vote for Trump, but it seems to be the case that areas with lots of evangelicals saw worse Republican performance overall, suggesting it’s possible that a larger share of evangelicals stayed home. (Evangelicals represented the same share of the electorate in 2016 as in 2012, but we don’t have any more refined data than that about turnout.)

Speaking anecdotally, I know many evangelicals who fit exactly this profile: They couldn’t vote for a candidate who perpetuates what they see as a campaign of mass murder of the innocent (namely abortion, the right to which Clinton supports), couldn’t vote for a candidate who they felt threatened essential norms and institutions of the republic (Trump), and, for whatever reason, didn’t like the idea of casting a “wasted vote” for a minor candidate.

The experience of my hometown might help illuminate the phenomenon I’m describing. Wilmore, Kentucky, which at 5,000 people houses about 2,000 students at a religious college and seminary, and thus is extremely religious, gave Evan McMullin one of his highest vote shares of any place in that state. The precincts that include the oldest parts of town — that is, not the new deep-red suburbs of Lexington that house commuters but the parts of town where “Old Wilmore” families live — gave McMullin more than 10 percent of the total vote. I suspect that particularly religious enclaves around the country had similar experiences of either abstention or third-party voting.

We can see two interesting stories: Areas with many Christians, without specifying the type of Christian, seem to show greater favorability toward Trump than the US on the whole. Religiosity is correlated with partisanship, and partisanship predicts voting behavior. But if we want to understand political fissures within partisan tribes, we have to look deeper. Among Christian groups, there is a great deal of inconsistency, with evangelical- and Catholic-heavy areas supporting Trump less than they supported Romney, while mainline-heavy areas saw the opposite trend.

We can also look at this spatially. Here’s a map of adherence for the Evangelical Lutheran Church of America, Disciples of Christ, United Methodist Church, the American Baptist Church, the Episcopal Church, and Presbyterian Church (USA) combined, representing the plurality of the mainline denominational adherence.

Notice the dark spots in Pennsylvania, which Trump flipped, and Wisconsin, which Trump flipped, and Iowa, which Trump flipped, and rural Minnesota, which Trump came astonishingly close to flipping. Here’s a map showing each county by Trump’s improvement over Romney in 2012:

It’s easy to look at these maps and see race, and undoubtedly race is an important factor. The question, however, is why states that voted for non-racist candidates for decades flipped; why did voters who in the past supported centrists cast votes for Trump? Did something related to race suddenly change? Yes it did: Doubtless, the election of President Obama raised the salience of race to many voters.

But race is not the only important social dividing line in America. Religion has always been, and remains to this day, consequential. It’s trivially easy to spot the deeply conservative Mormon areas on these maps. Is it so preposterous that other denominations may also be associated with different political preferences too, which may not map neatly onto right-left lines? And if mainline denominations are hemorrhaging members, does it not seem at least plausible those people’s political allegiances may be plastic? That is to say, without the common moral language of liberal Protestantism to steer these voters away from demagogues on the left or the right, might they not drift into more extreme political positions?

This is, to be sure, a first cut at the data, and far from definitive. Experts will be looking for explanations of the 2016 election for years to come. But many institutions that have traditionally directed Americans away from extremism have declined in recent years, from labor unions, to quality local newspapers, to churches. All of these bodies in the past helped create a shared set of pre-political biases about fairness, decency, agreed upon facts, and other important norms.

While progressives are keen to see in the decline of labor unions an important component in the rise of conservative political power, they rarely consider the impact of losing their movement’s soul. Despite mainline denominations commanding as much or more popular support and membership as labor unions, their decline seems to be unmourned within the progressive movement they birthed; the consequences of that decline likewise go unconsidered.

For those who see in the decline of American religion a progressive, liberalizing force, I must offer a word of caution. Some meaningful share of the rise of populism in the Midwest is likely due to the decline of the moral and political organizing force of mainline Protestant denominations. When moderate swing states lose their religious restraint, the right drifts to Trump, the left to Sanders (note that Sanders performed well in the kind of rural areas where mainlines were traditionally vital community institutions).

But if you think the current wave of populism is a rough ride, wait until you see what happens when the South is freed of the moral restraint of the Southern Baptists — the Southwest of Catholicism, or the West of Mormonism. The social and political disorder unleashed by those approaching changes could truly be something to behold.

*I also tested this limiting to just counties with high shares of given denominations, where county-level results more closely approximate denominationally specific views. Results were essentially unchanged. At the 10 percent, 20 percent, 25 percent, 35 percent, and 50 percent thresholds, the measured denomination-level vote swings never change their sign, although they do tend to become more extreme. Just as an example, in areas with 35 percent Evangelical Lutheran Church of America adherence, the pro-Trump effect direction is consistent, but the coefficient is even larger. Ten percent ELCA counties had a 4.7 percent swing for Trump; 25 percent ELCA counties had an 8 percent swing for Trump; 35 percent ELCA counties had an 11 percent swing for Trump. This broad pattern is true for most of the denominations I assessed.

Lyman Stone is a regional population economics researcher who blogs at In a State of Migration. He is also an agricultural economist at USDA. Find him on Twitter @lymanstoneky.