Why is there a significant difference between religious makeup of the legislative body and that of adult population in US?

Question

According to this article there seem to be quite a difference between religion of Congress members and that of general population:

The religious makeup of the legislative body is overwhelmingly and disproportionately Christian, according to Pew’s “Faith on the Hill” survey. Nearly 91 percent of members of Congress reported their religious affiliation as Christian, while 71 percent of U.S. adults identify as Christian.

A particular case is that of those with no religious affiliation:

Just one member of Congress, Rep. Kyrsten Sinema (D-Ariz.), identifies as religiously “unaffiliated.” She’s been the only “unaffiliated” member of Congress since she was elected in 2012, even as the nation’s ranks of religious “nones” — those who don’t subscribe to a particular religious creed or simply don’t believe in God — have grown to include nearly one-quarter of U.S. adults.

Question: Why is there a significant difference between religious makeup of the legislative body and that of adult population in US?

Answer 1

The main reason is game-theoretical asymmetry.

• If you "identify" (officially, as per self-label) as a Christian - or any other religion - non-religious people would most likely not decide to avoid voting for you based on that.

  There may be a small sliver of atheist zealots who would, but I seriously doubt it is enough to matter, especially in US First Past The Post two-real-parties system (only 5% of people would be more likely to vote for someone if they identify as "atheist" as per Pew poll in 2014)

• If you identify as a an "atheist" (and to many religious people, there's very little inclination to make fine nuanced distinction between 'nones', 'non-affiliated', 'agnostic', 'atheist' - they all get conceptually lumped together as "unbelieving atheist heathen"(tm) from the point of view of someone highly devout in a monotheistic Abrahamic tradition), there's plenty of strongly religious people who - according to polls - would refuse to vote for you.

  about half (53%) of Americans said they would be less likely to support an atheist. (Source: Pew poll 2014)

As such, there's absolutely no tangible electoral advantage to openly declaring yourself as not being affiliated with any religion; and practical tangible electoral risks/disadvantages in failing to do so.

As such, even if in practice you only attended Church 2 times in last 30 years and can't quote a holy book any better than last page of Hawking's doctoral dissertation, you're better off declaring yourself to be a Christian/Jew/Moslem than anything else. For every Shabbat-observing Joe Lieberman in Congress, there's probably 10 pork-eating "Jews" who haven't said a prayer in private since graduating college or kindergarten.

---

Additionally, as Alexei pointed out, Congress isn't statistically big - and there's even fewer valid 'none' seats if you start looking at Politically Wonkish details - Texas electorate outside Austin guarantees a Christian representation, so TX seat - or West Virgina - doesn't get to be counted if you study this phenomenon. As such, it may very well be probable that 'nones' simply randomly don't get represented as often, given the small sample size (you can ask on Cross-validated the probability of the exact outcome we see, especially adjusted for mimicry noted as the main reason).

---

I have explored both of these reasons specific to US Republicans in my answer here: Are there any atheist politicians in the Republican Party (GOP)?

A very useful stat from that answer:

Please note that there are at least 9 members of 114th Congress who didn't disclose their religious affiliation. Unfortunately, Pew didn't break down the party lines, so it's unknown how many of them were D vs R.

Answer 2

Looking at this purely mathematically, members of congress make up 0.00027% of the adult population. That's such an astonishingly small sample size (plus hardly taken at random), you shouldn't expect it to represent the general population in practically any measurement. The gender balance is 20/80. The average age is almost 60. Over 94% have college degrees, and 44% have law degrees. Median net worth is $456k.

They're supposed to represent the people politically. Mathematically speaking, there just aren't enough of them to also represent the people demographically. Some issues like religion might have complicating factors (like user4012 describes), but I'd argue the numbers would still be significantly skewed without them.

Answer 3

This is not at all surprising when statistics are used.

In "Athenian democracy", 538 individuals would be selected, the probability for of each individual falling into the class would be p, and the distributions for each of the 538 seats would be identical and independent. Thus the expected number would be 538 p.

But we don't have Athenian democracy, we have representation by district. In each district, you have some number of citizens voting (N). The distribution of voters in the class is Binomial with mean p N and standard deviation sqrt(N * p * (1-p)). Half the outcomes are above p, and half are below p, but we are interested in the fraction of outcomes above one-half (because the election is determined by the majority).

One of the figures on Wikipedia's page on Binomial distributions uses p = 0.7 which is a close match.

Notice that with N = 20 voters (green line), the chance to win fewer than 10 voters is far less than 30%, actually around 4%. And while with more voters the standard deviation becomes larger in absolute terms, it becomes smaller as a percentage of the vote.

The Central Limit Theorem provides that when N becomes "large" (more than a few dozen voters in the district), the Gaussian distribution becomes a very good approximation for the Binomial.

Now, the above results suggest that a class with 70% membership is absolutely dominant, such that 91% is actually unexpectedly low. The actual situation is more complex of course, because votes are not determined by a single class, but by a composite of many issues with less than perfect correlation. Also, like-minded individuals tend to cluster, so that the class membership in any given election district will vary from p. The election first samples the local prevalence of the class, then which of the registered voters choose to actually cast votes.

Answer 4

I think the trend you observed would come out purely mathematically, regardless of politics, even in a highly simplified model. For instance if you assume people are slightly to moderately more likely to vote for someone of their own religion, that the N% majority is present not just nationally but roughly repeated throughout all districts/units of representation, and that each district/unit of representation is winner-takes-all, then electing representatives just exaggerates the existing dominance of the majority.