How Many Nones Are There?
A comprehensive look at six different surveys about religion in the United States
One role that I take very seriously is the “numbers guy.” Reporters will email or set up a call with me and ask a lot of factual questions about the share of Americans who are evangelical or non-religious or attend services weekly and they expect that I have the answer. And, I do have AN answer. I don’t have THE answer. Because this is not hard science. We know that when water gets below 32 degrees, it freezes. We just don’t have objective metrics like that in my world.
When I was writing 20 Myths about Religion and Politics in America, I put together one chapter that was about this specific problem. The tentative title was, “Myth: We Know The Religious Composition of the United States.” Eventually, that chapter was cut from the volume. It showed up at Religion Unplugged under the headline, “Why Survey Data Of Religious Nones, Decline Of Christianity Differs” and I’m fairly sure no one read it.
Here’s the gist - there’s no objective way to measure religion. There just isn’t. The Census Bureau used to include a question about it on the Decennial Census, but that’s illegal now. Public Law 94-521 says that any religion question has to be asked on a purely voluntary basis. That’s a really big problem for people in my line of work.
If someone conducts a random sample survey and wants to know if it actually represents the population, you can benchmark it against the Census Bureau on factors like race, age, education, etc. Then you can weight your respondents up or down just slightly to get closer to that objective metric. We can’t do that with religion. There’s no objective measuring stick for us. So, everything is just a series of best guesses.
You want to see what that looks like in a really clear way? A simple question: what percentage of American adults indicate that they are non-religious. Seems straightforward, right? It’s not. Not even close. The graph below is the answer to that question using six different survey instruments from widely cited data sources on topics like this.
The last year in which I have a clear percentage from all six surveys is 2021. And here’s now many nones there are in those six instruments:
Cooperative Election Study - 36%
Pew - 31%
General Social Survey - 28%
Nationscape - 26%
Public Religion Research Institute - 25%
Gallup - 21%
Yeah, so the number of adult nones in the United States runs from 54 million (based on the Gallup numbers) to 93 million (based on those from the Cooperative Election Study). PRRI reported that the nones actually dropped from 2018 through 2020. Pew’s new 2023 estimate is 29%, down from the 31% they reported in the prior two years. Are the nones rising or are they falling? It’s literally impossible to know for certain unless someone wants to do a headcount survey of the United States and ask a religion question. It should only cost several billion dollars.
The Pew Research Center easily spends the most time on trying to get this answer right. The result of that is called the National Public Opinion Reference Survey (NPORS). One real innovation with this instrument is that it uses both an online component and a paper component. It does not use panels, like most survey firms use now. Pew contracts with Ipsos to send out a mailer to random addresses in the United States. A full breakdown of their methodology can be found here (PDF).
The survey that I typically use for analysis on this Substack is the Cooperative Election Study. It has a different approach to sampling. The CES contracts with YouGov and they use panels to solicit survey responses. These panels are used extensively in social science surveys now because they are, quite frankly, cheaper. And, there’s a heated debate about how much lower the quality is by using these panels instead of a random mailer. If you want to read a bit about how panels are conducted, check out this PDF from AAPOR. The other difference with the CES is that it’s entirely online - no paper surveys are filled out.