Skip to content

What parts of the country are most religious?

I saw this from Tyler Cowen:

The middle part of America is more religious than the South.

And I was like, Huh? So I followed the link which in turn linked to this article by J. D. Vance which said:

When Gallup ranked every U.S. state by its religiosity, states in the South took nine of the top ten spots, with South Carolina at fifth. These results were based on self-reported identification, meaning that a person’s faith was measured by what they told a stranger on the phone. Actual church-attendance rates tell a different story. Thanks to data compiled by the Association of Religious Data Archives, we have a very good sense of how many people in a given locale regularly attend church. By this attendance metric, the geographic heart of American religiosity isn’t the Southeast, but the middle part of the country — from Texas and Oklahoma through Iowa and the Dakotas.

Interesting. Here’s figure 6.3 from Red State Blue State, showing average religious attendance (as measured on a 1-5 scale from “never” to “more than once per week,” data from the 2000 and 2004 Annenberg surveys:

Comparing to what Vance wrote: Yes, Texas and Oklahoma have high reported rates of religious attendance, and Texas and Oklahoma are not in the Southeast, but they are often counted as part of the South. And Iowa and the Dakotas also have high rates of reported religious attendance. Not as high as Mississippi, Alabama, and Arkansas, but they’re up there. Reported religious attendance is not so high in some other states in the middle of the country, such as Ohio, Wisconsin, and Illinois, and it’s lowest in New England and the far west.

Following Vance’s recommendation, I googled Association of Religious Data Archives and I came across this map which I think is what he was talking about:

Screen Shot 2016-03-02 at 5.17.50 PM

This data source describes “adherents” as:

All members, including full members, their children and the estimated number of other participants who are not considered members; for example, the “baptized,” “those not confirmed,” “those not eligible for Communion,” “those regularly attending services,” and the like.

Of the 236 reporting groups, 49 reported members and adherents; 37 reported adherents only; 63 reported members only; four suggested a method for estimating adherents without reporting members; and 83 reported only congregation locations. Of the 63 that reported members only, four suggested their own adherent estimating processes, which we used to calculate adherents for them. For those 59 groups that reported members but did not report adherents nor suggest a method for computing them, we estimated total adherents for each county by dividing membership by the population at least 14 years of age and then applied this percentage to the Census 2010 100-percent count for the county.

This is a little different from what I graphed in that the data come from the congregations, not from survey respondents, also I was giving average attendance whereas here it’s a 0/1 variable. At this point I wouldn’t say that either chart is wrong; they are presenting two parts of the larger story.

I do wonder whether the statistics used in the above map are over-counting Catholics. I say this because of the dark colors of heavily Catholic states such as Massachusetts, New York, New Jersey, etc. Also this from the documentation:

How were Catholics counted?

For the 2010 study, each diocese was asked to provide, by parish or mission, their number of registered households, registered individuals, infant baptisms within the past year, deaths within the past year, and weekly attendance. This was part of an effort to provide a more congregational-based definition comparable to those of other religious bodies in the study. Put another way, the number of “adherents” is roughly equivalent to those who are known in some way to each parish or mission.

Altogether, this work enumerated nearly 59 million Catholics known in some way to each parish or mission in the United States. This number is substantially below two common measures of the Catholic population in the United States: survey data and The Official Catholic Directory (OCD). First, national surveys suggest that over 75 million Americans self-identify as Catholic. However, only 20 percent of respondents on the 2010 General Social Survey (GSS) were Catholics who attend at least once a year. When applied to the population as a whole, the estimated 62 million Catholics who attend at least once a year is reasonably close to the 59 million counted in the 2010 U.S. Religious Census. Second, The Official Catholic Directory (OCD) for 2010 indicated a Catholic population of 65,581,808 in the 50 states and District of Columbia. Yet there is reason to believe that the means used to generate the OCD number varies by diocese, with some dioceses relying on survey estimates of the Catholic population rather than parish-level statistics. For more information on the Catholic count, visit

At first this sounds like everything’s lining up—the total number in their count is comparable to the number of survey respondents who say they were Catholic and attend at least once a year. But then you remember that for other religions, these counts are much less than what was reported on surveys. So this suggests to me that, relative to other religions, Catholics are overcounted in the census that was used to create the above map. Given the shading of Utah, I wonder about Mormons too—they’re another highly centralized church.

In any case, if you divide the country into 4 regions: Northeast, Midwest, South, and West, then religious attendance rates—however measured—will be highest in the South (including Texas and Oklahoma). Yes, the rate in the middle of the country looks high—but remember that not many people live in North and South Dakota. Lots and lots of midwesterners live in Ohio, Michigan, Indiana, and Illinois, which have more moderate rates of church membership. (Illinois is on the high side but remember the Catholics.)

So . . . when comparing states or regions, I’d stick with my scatterplot above. I agree that survey responses on religious attendance are problematic, and I like the idea of bringing in data from other sources, but I think it’s too tricky to believe the comparisons between states and regions given the non-uniformity of reporting. And, even by that measure, “the middle part of America” (if it includes the Midwestern states of Ohio, Michigan, etc.” is not more religious than the South.

Just to be clear: I think the religious attendance data collected by the Association of Religious Data Archives are great; you just have to be careful with the interpretation.

P.S. And, yes, I could make my scatterplot as a map and make that other organization’s map into a scatterplot. I just didn’t feel like putting the time into grabbing the data and creating those new graphs. My bad for my file system being such a mess.

P.P.S. Seth Stephens-Davidowitz sends along this map of Google searches for “church”:

Screen Shot 2016-03-02 at 5.52.28 PM

Not the world’s most evocative color scheme, but you get the picture. Seth reports that maps of “Bible” and “God” are almost identical.

And, from my work with Yotam Margalit, a map showing the relative proportion of survey respondents, by state, who say they know anyone who attends church regularly:

Screen Shot 2016-03-02 at 5.54.34 PM

Blue is high, orange is low, and of course you’ll see a lot of variation in the small states.

Overall, though, the message is that church attendance rates are high in the South-except-for-Florida. I think the Tyler Cowen quote that led off this post was mistaken, a product of the misreading of a map plus artifacts in data collection.


  1. Elin says:

    I don’t know if it is true or not, but I have heard that one year the GSS ended up with an unusually high number of clusters in Utah and as a result there was a temporary over representation of Mormons and also a conservative shift on a number of variables.

    That said I think you are getting at the always interesting distinction between self-definition and behavior and between those and institutional definitions. So you may self-identify as Catholic (and when you die you want to have a mass or if you have children you want them christened) but never actually attend any service. Further the Catholic Church is centralized, as you mention, and has infant christening and other rites that happen at young ages. And they make sure everywhere has a parish that someone can join. In contrast, many Protestants are decentralized, have adult Baptism, and so on. You don’t have to be born again to be Catholic. In sociology of religion folks would talk about the difference between a church and a denomination. I don’t know enough about LDS to say for sure, but I’d guess that living in Utah it’s probably easier to self identify as LDS than to not. So I think if you want to compare levels of religiosity or levels of attendance or levels of self-identification as an adherent or official membership you have to keep in mind that each of those is very different from the others.

  2. Dale Lehman says:

    I’m surprised that Andrew has not yet realized that economists always start with the story and then torture the data until it supports it. This is nothing new. And, if pushed, I’m sure we can find a new intervening variable that will once again show that the midWest is more religious than the South (how about charitable contributions?).

    By the way, the Association of Religious Data Archives has loads of interesting data. One data set, calling it a “cult,” has data on people who have been abducted by aliens – all 50 something of them (this was the early 1990s). At that time, aliens seem to have had a preference for highly educated humans. Or maybe it was their power poses….

    • Rahul says:

      Only Economists?

    • Andrew says:


      In this case I’m guessing that Cowen and the tweeter he linked to were just misled by data problems, not that they had a pre-existing goal of showing that the midwest is more religious than the south. To the extent there’s a bias here, I think it’s a bias toward the counterintuitive, perhaps too much readiness to believe a fishy statistic if it confounds our expectations.

      • Dale says:

        Yes, it is not a particular bias towards the theory that the MidWest is more religious than the South, but towards the counter-intuitive. Economists love to find something counter-intuitive – it is one way we show that economics really is “the Queen of the Social Sciences.”

  3. paul alper says:

    This discussion of “church attendance” as a proxy for religious belief is misleading. “House of worship” is an awkward phrase but at least it includes synagogues and mosques. Furthermore, other spiritual gatherings take place in a home or even a field. In the current U.S. political scene there is the unfortunate and incessant eliding of religious (theological?, superstitious?) beliefs with worthiness and sincerity.

  4. Bill Harris says:

    Hmm. Didn’t you and Phil write a paper about maps that you recently publicized? :-)

  5. Coach says:

    The Association of Religious Data Archives map was only from the 236 respondents. That is a problem for Appalachia as many churches do not belong to any sort of large denominations. Many West Virginia residents (mostly in the southern half of the state) regularly attend churches that wouldn’t respond to any survey request from anyone. This would also be true in eastern Kentucky and southeast Ohio.

  6. J1 says:

    A little OT, but Texas is really part of three of the regions mentioned (south, west and Midwest), depending on where you are in the state.

  7. BenjaminL says:

    Would the scatterplot look different if it were just white people?

    I ask because: 1) In the context of the GOP primary, we are mostly concerned with white people, and 2) I would expect that in MS for example, high religiosity of African-Americans affects the overall number…

Leave a Reply