One of the most widely discussed phenomena during the 2016 election cycle was the growing educational fissure in vote choice among white Americans. The larger cracks relative to past years were beginning to show in pre-election polling leading up to November 8th, and the earliest data released about the 2016 election confirmed the result: while white voters without a college degree voted for Donald Trump by 39 points, whites with at least a degree voted Trump by a margin of only four points. Much attention has been focused on trying to explain this educational gap among whites, which according to exit poll data, was the largest in recorded history during the 2016 election.
Pre-2016 trends foreshadowed this outcome. Having measured changes in party identification—which closely mirrors vote choice—over time, Pew Research showed that some of the largest swings toward the Republican Party occurred among non-college whites, a trend detected with data from as early as 1992 but having accelerated since 2008. With the share of non-college whites in the electorate dropping rapidly over the last few decades, this intra-group shift in political attachment has proved crucial for the GOP.
But this should not be the extent of discussion on the 2016 white vote. As I partly showed in a previous post, considering several demographic dimensions in an analysis helps nuance an understanding of voting behavior patterns. One of those dimensions that has long assumed importance for the white vote is religiosity. This factor’s significance holds for whites only and the general electorate alike, and indicates that the more religious a person is, the more likely they vote and identify as Republican. Historical exit poll data indicates Democrats have won non-religious voters by at least a 2:1 margin since 1988; larger religious affiliation groups, such as Protestants, have consistently voted much more Republican. In 2016, for example, whites without religious ties identified as Democrat 28 points more than they did as Republican.
At the same time, the influence of religiosity in politics has not necessarily been stable. Both party bases have become less religious over the last few decades, but an important distinction comes in the fact that this decline has materialized a lot faster for the Democratic Party. In other words, the electorate as a whole has seen a rise in religiously unaffiliated voters, but that growth (since 1996 as reported by Pew) has largely affected the Democratic Party coalition. Moreover, some older research noted that Americans with high religious commitment—based on salience of religion and service attendance—were becoming increasingly more like to vote Republican toward the end of the 20th century.
Simply put, both education and religion are essential to analyzing the white vote, and especially so in recent times. That brings us to the 2016 campaign cycle. Before the election, Milo Beckman at FiveThirtyEight posited that education and religion helped best explain the white vote in 2016. Using a poll conducted through Survey Monkey, Beckman showed education (options spanning no college degree, some college experience, or a college degree) and specifically self-reported religious attendance rates were the strongest predictors of white vote choice between Trump and Hillary Clinton a few months out from the election.
With post-election data from the 2016 Congressional Cooperative Election Study in hand, I wanted to check how this same two-variable dynamic ultimately manifested in the election. For the below graphs and discussion, I calculated the vote choice—for Clinton, Trump, or another candidate—among whites only. I split that reported vote up by education: high school or less, some college, college degree, or postgraduate degree. Leveraging the CCES’s large sample size (about 53,000 Americans took the post-election survey), I was able to further break down white Americans’ vote choice by variables related to religion while still retaining large enough subgroup samples to make for reliable estimates. I’ll first do this process for the variable for self-described religious importance in one’s life, and then for the one for reported church attendance. (Before this, keep in mind the caveats to using this current data that you can read in the second and third paragraphs of my last DDHQ post here.)
In the first wave of the CCES, respondents were asked how important religion is in their lives, to which they could respond with any of the following: very important, somewhat important, not too important, or not at all important. Below I graph those responses split up by educational attainment among white respondents only for vote percentages.
There is very clearly a relationship between vote choice with both religious importance and with education among whites. Going from the top row to the bottom row moves from lowest educational level to highest, and going from the leftmost column to the rightmost one means lowest to highest religious importance—a good proxy for someone’s religiosity. Clinton vote increases/Trump vote declines consistently going from top to bottom (increasing education) while the opposite occurs going from left to right (increasing religiosity), as higher religious importance is associated with greater propensity for voting Trump.
Notably, the educational gap in Clinton vote is different across different levels of white religiosity. Let’s begin with respondents for whom religion is not important at all. Among these people, the gap in Clinton vote is 29.5 points between the lowest (HS or less) and highest (postgrad) educational levels. That same gap, however, is smaller for whites with higher religiosity. For example, among those for whom religion is very important, that same gap is nearly cut in half to 15.7 points. In other words, education matters a less in dividing the white vote for higher religiosity levels.
Looking down each column, there also appears to be a smaller religiosity gap in the vote—the difference in Clinton vote from lowest to highest religiosity—among whites with high school education or less; that gap grows nearly 20 points compared with every higher educational level.
The second important religious variable I’ll consider from the CCES is a respondent’s reported church attendance rate. When asked how often they attend religious services aside from weddings and funerals, respondents could choose from the following: more than once a week, once a week, once or twice a month, a few times a year, seldom, or never. I graph how vote choice broke down along those responses and educational level below.
Just like with religious importance, church attendance—in conjunction with education—has strong implications for vote choice among white Americans. As you move along from left to right for viewing this graph, you go from lowest church attendance—“never”—to highest attendance—“more than once a week.” Educational level still increases as you go from top to bottom on the graph. With greater church attendance comes a larger share of people voting Trump across all educational groups. Meanwhile, Clinton vote increases (and Trump’s decreases) as you move along higher educational attainment among all rates of church-going.
As you go up educational levels, the Clinton vote gap between whites with the highest and lowest church attendance grows a bit. For example, among those with a high school education or less, white Americans who never go to church voted Clinton 30.4 points more than those who attend religious services more than twice a week. College-educated whites saw a larger church attendance gap: those who never attended church voted Clinton 49.5 points more than those who attended more than once a week.
As you increase along church attendance rates, the educational gap in Clinton vote declines a bit. Let’s start with whites who never go to church. 75.8 percent of them with the highest educational level (postgrad) voted Clinton, while 42.6 percent of them with the lowest education (HS or less) voted Clinton—a 33.2 point gap. When we move to the opposite end of religiosity—whites who attend church more than once a week—that educational gap in Clinton vote shrinks to 19.3 points. This would suggest education starts to divide whites less once you move into higher religiosity levels (i.e. going by church attendance). Importantly, this matches the same finding when using religious importance as the religious variable of interest.
Taken together, this data on religious importance to oneself and church attendance rates speak to the importance of religiosity for explaining the white vote. That importance is shown to be clearly true in 2016, and past data supports the idea of this being a longstanding trend. Education is just as if not more important in revealing differences in voting behavior among whites. Combining education and religiosity exposes deep cleavages in the white vote, with education a strong negative association with the Trump vote and religiosity a strong positive associated with it. Finally, a more rigorous test confirms this: when I modeled white vote choice and controlled for gender, age, income, region, party, and ideology, education and religiosity (religious importance specifically) are the demographic variables most significantly associated with voting for Trump.