Bud was Right: “Measurement error is sin.”

Featured image credit: WNYC

By Paul A. Djupe and Ryan P. Burge [1]

Since the birth of the Religion and Politics section of the APSA, a more or less genial war has been simmering over the classification of religious people in the United States (for now, let’s leave Darren Sherkat out of this). Drawing on the hard work of early APSA members to change the available religion measures in the ANES was born the RELTRAD approach that allowed classification of individuals into different religious traditions based on their denominational affiliations (see the Leege and Kellstedt 1993 volume as the first fruits of these labors). Of course, the credit has since been given by dint of citations to Steensland et al. (2000) who wrote up the approach from what they learned at a Wheaton seminar put on by “The Gang of Four” (Corwin Smidt, John Green, Bud Kellstedt, and Jim Guth).

There are multiple objections to this scheme that can be discussed in another post, but one important objection has been: Does it need to be so complicated? The RELTRAD approach based on denominational affiliation takes up 3 pages in the ANES codebook and has changed three times since 2000. In fact, one of the defectors on this question was an ‘enemy within’: John Green. He pushed Pew to adopt a simpler scheme based on evangelical self-identification to break up those who identify as Protestant (and later as “Other Christian”). PRRI is a consistent user of this approach as are exit pollsters, among others. This begs the question that has generated outsized reactions: which one should we use? While one of these authors suggests neither, we both agree that empirics should be the judge.

One of the articles of faith of either approach, really, is that Baptists are evangelicals (with a few exceptions). In most studies of reasonable size, this has had to remain an article of faith. But with its enormous sample and its inclusion of measures appropriate to test both approaches, the 2016 CCES affords us the potential to examine such an assumption with a great deal of certainty – there are 7,613 (!) Baptist identifiers in the CCES.

The figure below shows us what proportion of each Baptist denomination in the CCES identifies as “born again or evangelical Christian.” In summary, 28 percent of Baptists do not identify as evangelical and the range within the denominational family group is considerable. Just under a fifth of Southern Baptists do not identify as evangelical, while 41 percent of Free Will Baptists are non-identifiers. Most importantly, 66 percent of American Baptist Church members ARE identifiers though they are typically classified as “mainline Protestant” (i.e., non-evangelical Protestants).

born again baptists.png

There are a number of denominations that are traditionally considered “mainline” using the Steensland et al. typology. The United Methodists, Disciples of Christ, and Episcopalians, among a few others, are considered non-evangelical Protestants. The figure below indicates what percentage of each of these denominations self identifies as born again or evangelical. As previously described, over ⅔ of the ABCUSA respondents responded in the affirmative, but these were not the only mainliners to describe themselves as evangelicals. 46.1% of members of the Disciples of Christ, 38.8% of United Methodists, and about 28% of the PCUSA and ELCA say that they are born again or evangelical. And to muddy the waters even more, the RELTRAD denominational scheme puts non-denominational Christians in the evangelical camp, but the CCES indicates that only slightly over half (51.7%) see themselves that way.


This discussion is more than a mere academic exercise as these classifications can have dramatic effects on the perception of religion and politics. The figure below shows how tweaking the definition of evangelical can significantly alter the percentage of individuals who voted for Donald Trump. Using the strictest definition – white denominational evangelicals who attend church regularly – the data indicate that 81% were in Trump’s camp. But if the definition is relaxed to include anyone who says that they identify as an evangelical, regardless of their attendance or race, the number drops to 61%.


None of the evidence here suggests which approach should be used. In fact, they suggest the weakness of the other and therefore raise troubling questions about just what we purport to be measuring. The critical concerns are accurate placement and how a RELTRAD scheme captures “exposure.” That is, do individuals know what kind of church they are attending?[2] Do they identify with what scholars think is the dominant identity and belief system attached to that denomination?

The mis-match that we have identified here either suggests that some congregations within these denominations are distinctive, transmitting different normative religious beliefs, identities, and political information, or that some individuals do not fit coherently in their denominations. Either way, these results suggest it is problematic using either the denominational or self-identification approach to assume some coherent exposure to either religious or political norms, especially if avoiding the sin of measurement error is a goal.

Paul A. Djupe is an affiliated scholar with PRRI, the former editor of Politics & Religion, and the series editor of Religious Engagement in Democratic Politics (Temple). Further information about his work can be found at his website and occasionally on Twitter.

Ryan P. Burge teaches at Eastern Illinois University in Charleston, Illinois. He can be contacted via Twitter or his personal website.


1. The title quote from Lyman “Bud” Kellstedt comes from the preface of the Gang’s 1996 book Religion and the Culture Wars: Dispatches from the Front. How the contributors thought about measurement error was meant to highlight their different religious traditions. It is telling, perhaps, that Green’s is “What’s a little measurement error among friends?”

2. In the ANES, there has traditionally been a category for those with “no further specifics” – that is, they identify as “Lutheran,” for example, but cannot place themselves in a specific denomination. This, of course, suggests that they do not really know what kind of church they attend.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s