If I post a graph that indicates white evangelicals were more supportive of Donald Trump in 2020 compared to 2016, I know the inevitable comment: Yeah, but isn’t that because the composition of evangelicals changed during that four-year period of time?

The implication is that Trump had a galvanizing effect on the evangelical community. Specifically, he may have pushed some people out of evangelicalism who weren’t that conservative, while also drawing other people into the faith who may be more attracted to the movement because of its political beliefs as opposed to its theological convictions.

I get the impulse to believe that there’s a “sorting” process involving American religion and politics. In short, people move about the political and religious landscape in search of a suitable home on both dimensions. They want to feel comfortable in who they cast their ballot for as well as with the pew they sit on Sunday morning.

As a result, they are constantly on the hunt for welcoming spaces.

It’s rare to have a dataset that actually addresses this understanding. Beyond the “one off” surveys that happen all the time in social science, the most widely used instruments are longitudinal surveys. Instruments like the General Social Survey, the National Election Study, and the Cooperative Election Study all happen in regular intervals.

These types of surveys are amazing if one wants to track how groups have changed over time, but they are much less helpful in tracking movement at the individual level. They can tell us that white evangelicals were more supportive of Trump in 2016 compared to 2020, but they cannot as easily reveal whether the composition of the group changed along with their level of support.

There is a solution to this problem: the panel survey.