One of the most often repeated comments on social media when I post a graph about the rise of the “nones” or the continued growth of those without religious affiliation is something along the lines of this:

And, to be honest, I have been thinking about how to get to the bottom of that for the last couple weeks. Clearly, evangelicals have a large influence on the Republican party. At least three quarters of white evangelicals voted for Donald Trump. It’s incredibly hard to find another group that is so wed to a particular political party. (The only exception I can find is that black Protestants are ~90% Democratic voters in presidential elections). But, now that the nones are the same size statistically as evangelicals or Catholics, that means that those without religious faith should have some significant sway in electoral politics, right? Well, not really.

The clearest evidence of this growth failing to translate into elected officials came from Pew’s “Faith on the Hill” report which tracks the religious affiliations of all members of the 116th Congress. The findings are clear: the vast majority of members are Christian (88.2%). Over half of Congress identifies as Protestants, while three in ten affiliate with the Catholic church. Of those that remain, there are a few Mormons (1.9%), almost three dozen Jews (6.4%), and 2 or 3 Buddhists, Muslims, Hindus, and Unitarians. And then, near the bottom, it comes into full view: there is exactly one member of Congress who falls into the “unaffiliated” category. That person is Kyrsten Sinema, the new Senator from Arizona. Another member of Congress Jared Huffman, a Democrat from California, seems to identify as an agnostic. But he refused to answer the questionnaire for Pew’s report, however he has recently come out as a humanist. So, if the nones are nearly a quarter of the population, why are they less than 1% of Congress? Well, I wanted to try and answer that in a few posts.

The one theory that I really wanted to test comes from Twitter, as well:

That’s a solid theory that I think deserves a rigorous empirical test. To operationalize it, here’s what I wanted to test: are the nones more politically heterogeneous (meaning more diverse) than other religious groups? If they are, then that would indicate that they would have a hard time coalescing around a single candidate/policy/party because they can’t all agree on shared goals in policy making.

Before I get to the analysis I have to take a quick detour into statistics. I will be brief and not too jargony. If one wants to measure how diverse a group is on a variable the clearest way to do that is to calculate the mean, and then calculate how far from the mean the average member is (aka the standard deviation). That’s what I have done below.

I simulated three groups of 100,000 respondents. In each of the three groups, they have the same mean age: 45 years old. However, each of the groups has a vastly different standard deviation. The group in the left panel has a great deal of diversity in ages, with a standard deviation of 20. The middle group has a moderate standard deviation of 10, and the panel on the right has a great deal of homogeneity with a standard deviation of 2. You can see how they visualize completely differently. The high standard deviation group has a flatter, more even distribution, without a well defined peak. On the other hand, the group with a low standard deviation has a very clear peak that is well defined, with most respondents tightly clustered around that peak. In short: higher standard deviation = more diverse range of values, low standard deviation = more concentrated range of values. One standard deviation will encompass 68% of the sample, two standard deviations is 95%. If you want to dig into this more, I highly recommend this Khan Academy video.

So, I wanted to calculate the mean and standard deviation of party identification for a number of religious groups to determine which groups are more politically unified, and those that are diverse in their viewpoints. I used the largest and most recent sample I could find, the 2018 Cooperative Congressional Election Study, and then I simulated 100,000 members of each of the fifteen groups using the means and standard deviations calculated from the 2018 CCES sample. This helps to smooth out some of the smaller groups and helps to reduce disparities between the large groups and those with fewer respondents. The end result of that is the visualization you can see below. Note that for each group, the mean party identification is denoted by a vertical dashed line and the standard deviation is listed in the top right of each panel. Recall that a higher standard deviation represents a more diverse set of views, while a lower standard deviation represents more concentration around the mean.

First, note that atheists are the second most Democratic leaning group of the fifteen in the sample, only bested by black Protestants. Agnostics rank fourth from the left, just after Muslims. However, note the “nothing in particular” group, which is nearly 20% of all respondents and much further to the right. In fact, this group is just slightly to the left of the mean partisanship score for the entire sample. When the nones are classified, oftentimes these groups are added together which pulls the mean partisanship further to the right and therefore misrepresents the true political outlook of atheists and agnostics.

However, the standard deviation is really the item that should draw the most attention now. Which groups are more politically diverse in their viewpoints? It helps to look at the sample distributions that are flatter and don’t have a well-defined peak. Who fits the bill there? Non-white evangelicals, white mainline Protestants, and white Catholics. Which makes sense. The mainline is filled with old school, country club Republicans and younger social progressives. Non-white evangelicals are pulled in both directions, based on their social views as well as their opinions on immigration. Some Catholics are incredibly socially and politically conservative, while others are more moderate.

Which groups are more politically homogeneous? Well, this may come as a surprise to some but atheists are the second most politically unified of any of the fifteen groups with a standard deviation of 1.63. Following not too far behind are agnostics at 1.74. For comparison, these two groups are as politically concentrated as Muslims and Buddhists. Even the much more amorphous “nothing in particular” category is fairly homogeneous as well, with a standard deviation of 1.96. That score is very similar to the diversity among white evangelical Protestants.

So, what does all of this mean? Well, it gives absolutely no ammunition to the argument that the nones are too politically diverse to actually coordinate an effective voice to politicians and candidates. In fact, the nones as a group are more politically unified than many of the largest religious traditions in the United States (evangelicals and Catholics). One dimension that I thought about a lot was how much race plays a role in political partisanship. So, why didn’t I break the nones up into racial groups? Interestingly, I tried that and it actually made the standard deviations go up for white atheists and white agnostics compared to the more racially diverse groups, but not by much. So, there isn’t any evidence here that it’s a racial issue cleaving the nones, either.

I am working on some further analysis that will try and pinpoint why nones have a hard time getting elected, and if they were going to see success what parts of the country would be ideal locations. More coming soon.

Ryan P. Burge teaches at Eastern Illinois University in Charleston, Illinois. He can be contacted via Twitter or his personal website. The syntax for the post can be found in here.