Friday, January 24, 2014

Rates of atheism/agnosticism by ethnicity

Staffan wrote:
It would be interesting to look at the ethnic origin of White atheists. I'm betting they are of Northwestern European ancestry, with perhaps the Irish being an exception.
The following table shows the percentages, by ethnicity, of atheists or agnostics among GSS respondents from 2000 onward among groups with at least 50 respondents (and most samples are far larger than that):

1. Chinese24.3
2. Russian13.2
3. French12.8
4. Other Asian12.1
5. Scottish11.3
6. English/Welsh10.0
7. Polish9.8
8. Norwegian9.7
9. Swedish9.6
10. Italian9.4
11. German9.1
11. Puerto Rican9.1
12. Dutch8.9
13. Danish8.2
14. Czech7.1
15. Irish6.7
16. Canadian6.2
17. Filipino6.1
18. Native American5.8
18. Portuguese5.8
19. Non-Spanish West Indies5.7
20. Spanish (Iberian)5.6
21. Indian (dot)4.4
22. African3.5
23. Mexican3.0
23. "American"3.0

Perspicacious prediction regarding the Irish.

A few remarks: "Other Asian" does not include those of Japanese descent, a group for which the sample size was insufficient to be broken out with any reliability. In the American context, "Russian" also means heavily Jewish. The fathers of the Scottish Enlightenment need not spin in their graves. Echoes of Thomas Sowell's black rednecks. Should we prepare for a fifth NAM-led great awakening?

Fifteen of the 26 included ethnicities have contemporary data available from the WVS*. The percentages of respondents queried between 2005-2008 who said that God is "not at all important" in their lives:

1. China39.5
2. Sweden33.3
3. German31.2
4. Holland31.0
5. Norway27.9
6. France26.1
7. Great Britain19.7
8. Russia13.4
9. Canada10.3
10. India8.7
11. Italy2.0
12. Poland1.3
13. Trinidad and Tobago1.0
14. Mexico0.9
15. Ghana0.1

* Trinidad and Tobago serves as a proxy for non-Spanish West Indies ancestry and Ghana as a proxy for African ancestry.

The correlation between atheism and agnosticism among those in the US and their co-ethnics in their mother countries is an impressive 0.66 (p = .008). Excluding Russia, which is a bit of an aberration due to the heavy Jewish skew in the US, the relationship strengthens a bit more to 0.68 (p = .007). To some extent, we are indeed our father(land)'s children.

Sharp guys like Jayman and Greg Cochran have argued that ethnic self-identification among Americans should be taken with a grain of salt. Disputing a Cochran assertion instinctively feels like an act of folly, but the residual relation shown above suggests that self-reported ethnicity has utility. Census findings don't give the appearance of randomness. Of course there has been a lot of intermixing, but it seems plausible that in most cases the ethnicity that people self-identify as belonging to is the ethnicity they are most heavily descended from.

If self-described ethnicity isn't accurately derived from biology, then it is presumably a product of perceived cultural affinities. Yet among most whites in the US today, any awareness--let alone active sense--of ancestral heritage is virtually non-existent. Very few average joes think of themselves as fifth-generation German or tenth-generation English. Americans of Irish descent are a partial exception, what with St. Patty's Day and the t-shirts that accompany it, but in mid-March there are a lot of people who fancy themselves Irish for a day even though they lack any historical connection to the emerald isle. They answer "Italian" because at some point they were told by one parent that his parents' parents came over from there and from the other that her parents' parents' parents did, too.

WVS variables used (fifth wave): V192(1)

GSS variables used: YEAR(2000-2012), ETHNIC, GOD(1-2)


JayMan said...

"Of course there has been a lot of intermixing, but it seems plausible that in most cases the ethnicity that people self-identify as belonging to is the ethnicity they are most heavily descended from."

I wouldn't even go that far.

What does someone who is 3/4 WASP and 1/4 Italian answer on the Census? My money is on the latter.

That said, yes, self-reported ancestry does bear some correlation to genetic ancestry.

Curious, what would your correlation look like if you only considered Whites?

Fascinating finding nonetheless.

Audacious Epigone said...


It drops to a more modest .44 (p = .2). The high p-value isn't surprising though since we're only looking at eight data points.

Even among European ethnicity, you think there is a one-drop instinct at play? You may well be right, I don't know.

JayMan said...

Very interesting. I'd say .44 is nothing to dismiss.

Considering shifts in reported ancestry over the years (the considerable drop in British ancestry in favor of other identities, for example), I'd imagine fashion is at play. Claiming to be British (WASP) is probably too plain, and our climate of embracing all things diverse, I wouldn't be surprised that the more colorful one can make one's ancestry appear to be, the better.

Anonymous said...

Seven Northwesterners in the top. But less evident in the ancestral stats. This is probably partly due to the fact that many people left Europe for religious reasons.

Russia seems like an anomaly. They are not very intelligent, and not outbred and modern like NWs. And yet not very religious either.

Anonymous said...

(paradigmo is Staffan.)

Audacious Epigone said...


Indeed, Russia and the other former Soviet states are all outliers on a lot of religious/cultural/behavioral questions.

Anonymous said...

My guess is that surnames are a major determinant of European ethnic identity. For instance, if you're a typical white American, a mixture of many different European ancestries with no cultural ties to any of them, and your father's father's father's father was a German immigrant named Schmidt, your surname would be Schmidt and you'd probably identity as "German" entirely on that basis.


Anonymous said...

Re Russian Jews and Euro Russians, a while ago I recoded the variable Jew as rJew, where 0 is subject self classifies as not any sort of Jew, to try and sift out Jewish respondents on the Russians. That reduces the Russian sample size from 93 to 47.4. Another variable I recoded as rJewish (can't quite remember which) reduces the sample size to the same size, for what it's worth.

Anyway, in this selection filter (should you chose to accept it) the Russian membership in categories 1 and 2 is (3.2+7.5), so 10.7%, lower than the Russians here, but still higher than most White ethnics.

When I broaden the year range to 1980-2010 to get a bigger sample and look at wordsum means, the above "Russian, non-Jewish" by the above variables hits around the average White Wordsum for the sample, while the "Russian, Jewish" hits around the Jewish mean for the whole sample, if that helps confirm or deny.