By Paul A. Djupe, Denison University

Ryan P. Burge, Eastern Illinois University

Robert P. Jones, CEO of PRRI

American religion can be characterized in countless ways. There are 350,000+ congregations and hundreds of millions of adherents, identifiers, and believers in the US. Almost every combination of views and identities is out there somewhere. One of the primary tasks of science is to make sense of the chaos by proposing orderly taxonomies, combining and labeling like things so we can describe how much there is and whether and how it is changing (think of the fossil record). Rocks are categorized by how they are formed into igneous, metamorphic, and sedimentary. Animals are classified in 7 layers from kingdom to species. Stars are classified by their spectra and temperature. Generating these systems is not an orderly process and it takes considerable time and energy, not to mention some ingenuity, of a multitude of researchers. Because there is tremendous value in classification systems due to the research questions they enable, it is continually worth our while to examine their assumptions, fit, and coverage.

One of the most popular schemes for classifying the multitude of religious groups creates “religious traditions” and is called RELTRAD for short. Using General Social Survey data, the RELTRAD coding scheme as laid out in the Steensland et al. (2000) article appendix [ungated] simply drops some respondents from the survey sample who are arguably hard to classify. As shown below, the proportion that are “Unclassified” has been growing, so while only 2% were Unclassified in 1970, that number has grown to 6.5% of the sample in 2018. For context, this is four times the size of the Jewish population, half of the white evangelical population, or ⅔ of the mainline Protestant population. This is a problem.

Leaving Unclassified 6.5% of the sample distorts the portrait of American religion. Exactly how that picture changes by including the Unclassifieds depends on methodological choices that should be consistently applied across the population, but have not until this point. Here are two potential hot takes. If the Unclassifieds are unfit to be added to a religious category, then they should be added to the nones, serving to inflate their size by 28% (from 23.2% to 29.7%). Another take simply uses the existing RELTRAD scheme and adds the Unclassifieds as shown below. Once accounted for, all remaining groups drop in size, of course, given the expanded baseline. How should we think about the Unclassifieds? As their own group? Can we further classify them?

We support counting the Unclassifieds and favor parsing them into existing groups. In order to do this, we need to work through the genesis of the Unclassifieds in order to understand them. Here is how they were forgotten in the first place. These questions often start with larger “family units,” like Protestant, Catholic, or Jewish, before asking about specific organizations within that family unit: Episcopal Church, African Methodist Episcopal, Evangelical Lutheran Church in America, Southern Baptist Convention, etc.

People typically do not have a hard time picking a “family” – they suggest they are Christian or Protestant – but then at the organization level may select “Don’t know”, “Non-denominational”, or “interdenominational” when pressed for an organizational affiliation. In these cases, researchers are skeptical that they actually belong at all, and Steensland et al. decided to apply a worship attendance filter. Those who attend at least once a month were kept and classified as evangelicals; those who attend less often in these three categories are the Unclassifieds. There is no race filter. The instruction to do so comes in the last words of their article on p. 316, footnote d:

This was not a large slice of the population in 1970 when Protestantism was still largely denominational. But denominations have been shedding members and the conventional wisdom, at least, is that nondenominational churches have exploded onto the scene in the meantime. Obviously, getting an estimate of nondenominational adherence is complicated by this attendance filter that has been applied almost solely to them.

So, who are the Unclassifieds? They have to be from only a few groups, given the definition in footnote d, and most of them identify as Protestant. Only 69% (in 2018) are white, 24% are black – much higher than the black population percentage. They also answered a further religious identification question and 52% identify as “born again or evangelical” while 48 percent do not. Though they do not attend church at high levels, other elements of the survey appear to give us considerable information with which to classify them, a decision Woodberry et al. (2012: 68-69) support: “Ideally, coding schemes should attempt to assign these ambiguous responses based on doctrinal markers and/or religious movement identifications as well as other religious beliefs and practices.”

Approach 1 – Consistent Attendance Filtration

There are good reasons for using attendance filters, but it is unclear why they might be applied selectively. Is the seldom-attending person who is unclear about a denomination any more exposed to religious messages than the person who picks a denomination? Probably not. Instead, it is possible to use attendance filters consistently across the board. Here is an attempt to do that, keeping the basic structure of the RELTRAD scheme in place for the largest religious groups. Unclassifieds are folded in because the attendance filter that excluded many evangelicals from being included in that group has been removed at the sorting stage. This approach generates (1) tradition attenders (labeled “high” below – red bars) and (2) tradition nominal attenders or ‘nominals’ (labeled “low” below – blue).

Catholic numbers have largely held steady, but have lost about 4 percentage points overall since the beginning of the series in 1972. However, frequent attending Catholics in 2018 are half of what they were in 1972 (19.4 down to 10.4) – nominal Catholics are the majority of all identifiers. For the purposes of this visual, we separated out non-denominationals from all other traditions. The result is rather stunning because denominational evangelicalism is only 10% of the population in 2018 and non-denominational Christians make for a LARGER group. It is also stunning because it reveals that there has been no explosion of non-denominationalism across this time period; instead they have been around all along, with their size rising and falling in tandem with denominational evangelicalism. Though still larger than denominational evangelicalism, it is important to note that non-denominationals are near their low point for this 46 year period.

Moreover, it seems clear that non-denominational Christianity is a more vital force than any denominational grouping. While 51% of denominational evangelicals are nominal attenders in 2018, only 30% of non-denominationals are nominal. For the record, a minority of mainliners are nominal (45.3%), though their numbers overall have been in freefall. Overall, in 2014 the proportion of nominals and nones totalled 56.5%; by 2018 that figure rose to 59.3%. Strong majorities of Americans are at best weakly tied to a church body.

Approach 2 – RELTRAD as Identity

An alternate way of going about this eschews the attendance filter altogether and thinks about classifying people based on their identification.[1] Even if people are not exposed to religious messages through attendance, they may still be disposed toward particular religious messages or hold particular beliefs and values because of their identification. There are (at least!) two ways to go about this. We could rely simply on individual identifications as “born again or evangelical.” Of those typically classified denominationally into mainline Protestantism, around ⅖ identify as evangelical (that’s true in 2018 – see below). At the same time, 20% of denominational evangelicals are not identifiers (26% in the 2018 CCES). So, there is definitely some play in the joints from using one approach or the other. But also note that Catholics, Nones, and Jews identify as born again as well. One really intriguing finding not in evidence here is the very high rate of born again identities among the transgender community.

When simply using the born-again/evangelical and race filters, RELTRAD as identification scheme looks like this.

There are alternate strategies, such as using the identification item for those having a hard time picking a specific denomination to adhere to. When people pick, “Methodist, don’t know which” researchers could use the “born-again or evangelical” item to sort them into either the non-evangelical, evangelical Protestantism, or Black Protestantism.

There are no easy answers to the classification of American religion, there are only choices, choices researchers need to make given the task to which they set themselves. At least one of the authors of this piece has expressed strong reservations about the degree to which a taxonomic model works with American religion – “American religion is not a matryoshka doll.” Still, as our work often finds such high level taxonomies convenient at a descriptive level, we need to apply a set of values in a neutral way:

1. Filters should be applied sparingly since compound categories are ambiguous leading to questions about just what is being classified (is it identification, exposure to organizational environments, or something else?).
2. Filters should be applied evenly to all groups. This would apply to behavior (e.g., worship attendance), race, and other identities.

It is clear that the existing RELTRAD scheme from Steensland et al. should not be used without alterations, something that their 2012 reflection in Social Forces makes clear, “ Further research efforts in these areas should take us beyond RELTRAD, while retaining RELTRAD, and to better explanations of the causal links between religion and social and political attitudes and behavior.”

Paul A. Djupe, Denison University, 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). Further information about his work can be found at his website and on Twitter.

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

Robert P. Jones is the CEO of PRRI and author of The End of White Christian America. He is on Twitter.

---

Notes

1. There’s a long debate wrapped up in this between those who adhere to a denominational approach and those favoring a shorter identification based approach. Both attempt to parse Protestants, but in different ways. The denominational approach prioritizes proper nouns – people need to choose a denomination that their church belongs to and then that denomination is classified as falling into evangelical, mainline, or Black Protestant camps. The identification approach simply asks people whether they are “born-again or evangelical” and classifies accordingly (factoring in race). As Burge and Lewis (2019) show, the two approaches involve very different portions of the population, but return surprisingly similar estimates of political phenomena of interest (see also the discussion here).