Pew still doesn’t understand the nones

Back in 2015, Pew Research Center released a massive report giving extensive projections on the growth of religions across the world between 2010 and 2050. Among the headline findings was that the Nones – those not part of any religion – would decline as a proportion of the global population from 16.4% to 13.2%. I wrote a rebuttal, explaining why their analysis was flawed, and they have now released a new report, defending their projections that makes all the same mistakes.

Before I even start, let me clarify that I am aware that “unaffiliated” does not equal “nonreligious”. But my argument is that the nature of the “nonreligious” is what is driving the odd results seen in the “unaffiliated” group in Pew’s report. And those odd results are the coal-mine canary warning of the failure of the model in general.

Now I have to make it clear that if you accept all the assumptions of Pew’s model, their conclusions are completely sound. And the model Pew uses is not ridiculous at all. If you are interested in tracking the growth rates of religions, their model’s assumptions are perfectly legitimate. The “if” clause in that sentence should give you pause, but bear with me for a bit.

With their model, they find that while the absolute number of nonbelievers will grow, it will be outpaced by the growth of believers, and thus the proportion will shrink. Here is the chart they use to illustrate their point (note that it shows the results up to 2060, not 2050 as the original report did):

Their model is primarily based on the birth and death rates of religious people, “births of religious people” being defined by the religion of the parents. The basic idea is not complicated. Muslims breed more prodigiously than Christians, on average, therefore the amount of Muslims is going to increase faster than the amount of Christians. If the death rates of both groups are the same, then the proportion of Muslims is going to increase. Of course, it’s a little more complicated than that, as Pew’s model also takes into account the death rates of each group, migration, and “switching” rates (more on that shortly).

This is a perfectly reasonable way to model the growth of religious populations. The problem – and this is the same point I made back in back in 2015 – is that it is not a reasonable way to model the growth of nonreligious populations. You cannot view nonreligion as “just another religious group”. And you cannot simply “ignore” nonreligion, chart the growth patterns of religions, and then say “everything left over is in the nonreligious group”. Nonreligion is fundamentally different from any and all religions, in some very key ways. And nonreligion, with its wildly different growth patterns, has real impacts on the growth patterns of all other religions.

First, religious people are almost entirely created at birth, using the parents’ religion. There is some adoption of religion later in life by people born to nonreligious families (ie, “switching” from unaffiliated to a religion), and some small amount of switching (from birth religion to another religion or no religion). But by and large if you see a baby, and you know what the religion of that baby’s parents is (because the two parents will almost always share the same religion), then you know what that baby’s religion is going to be when it grows into an adult. (I can just feel your atheist fingers twitching over the keyboard, ready to fire off an angry rebuttal. Bear with me.) Or you can look at it another way: If you see an adult member of X religion, it’s a pretty safe bet that their parents were also members of X religion. It’s not always true. But it’s bet you’re not really making a huge gamble on.

Nonreligious people, however, can be created at any point in a person’s life span. It’s probably true that nonreligious parents will tend to produce nonreligious offspring. But the inverse is not true: If you see a nonreligious person… you can’t even begin to guess what their parents’ religion was.

So while simply looking at the parents’ religion as a predictor of a person’s religion works well for religious people, it just doesn’t work for nonbelievers.

The second key difference has to do with how religion is transmitted. Allow me to use a hypothetical scenario to illustrate the problem: Imagine there is a country of a million people, 90% of whom are of the Foo religion, while the other 10% are all of the Bar religion. Assume there is no immigration, and virtually no communication with the world outside the country’s borders for most people. In that country, what is the probably of a random Foo worshipper switching to the Bar religion? Eh, it could happen occasionally. Same for a random Bar worshipper switching to Foo. But what is the probability of a Foo or Bar worshipper converting to the Qux religion? Answer: virtually 0%. In order to become a member of religion X, you need someone or something to proselytize X to you. You can’t just spontaneously become a member of a religion with absolutely no connection to other worshippers of that religion. You need existing members of a religion in order to create new members of that religion.

Except that’s not the case for nonreligion. Even in a country that has zero nonbelievers and zero interaction with the outside world, do you know what the probability of someone spontaneously becoming a nonbeliever is? It’s almost 100%.

So any model that tries to model the growth of religious populations will fail when nonreligion is introduced into the equation because of the massive differences in how they’re transmitted. Religions spread with an epidemiology consistent with a virus: you need a carrier to spread the disease. Nonreligion doesn’t “spread” so much as it pops into existence in people, sometimes without an external cause. Sure, the existence of other nonreligious people and nonreligious ideas in an area increase the chance of a new case of nonreligion, but even in the complete absence of those things it will eventually happen.

Pew tries to defend its model by saying it includes “switching”. Sure, yeah, but look at how that actually looks in their own results, as published in the original report:

A chart like that is not a sign that all’s well and good with your model. That is a flashing warning sign screaming at you that something very different is going on with the unaffiliated.

The extremely high rate of abandonment of Christianity should also be troubling the defenders of Pew’s model. What, do they think there’s something special about the nature of Christianity that, after dominating most of the world for 1,500 years, now people are suddenly hemorrhaging out of it? Maybe they also think it’s just a coincidence that ~100 million people switched out of Christianity and ~100 million people “switched in” to unaffiliated.

The evidence is staring Pew right in the face: Their model breaks down with the nonreligious. There is something strange happening in predominantly Christian populations… which, not coincidentally, also tend to be more affluent and educated than most other religious populations. Their model is just not accounting for something.

Without knowing the precise nature of their model, I can’t comment on the precise problems. However, I can point out some important factors they’re probably not taking into account.

One is that the “switching rate” into unaffiliated populations cannot be a constant or even a simple factor of any kind. Canada’s nonreligious population almost doubled in 20 years between 1991 and 2001, both in real numbers, from ~3.4 million to ~7.8 million, and in proportion, from 13% to 24%. In the same period, roughly, the US went from ~15 million to ~35 million, or 8% to 15%. Obviously that can’t be by birth and death or migration. It all has to be “switching”. But how can we make any predictions from that? Surely we wouldn’t expect Canada’s nonreligious population to be 400% of Canada’s population in 2060. But what Pew predicts is that despite the nonreligious population literally doubling in North America in 20 years between 1991 and 2011, it’s merely going to do the same is 40 years, between 2010 and 2050 – with the proportion changing from 17% to 26%.

Clearly their model is missing something about how nonreligious people are “created”.

The final set of arguments they make in their defence is that they’re just not seeing a lot of nones in majority Muslim countries. They even try to deflect the point that nones flourish in well-developed areas by using – I shit you not – India as a counterexample.

That anomaly is – I would have thought obviously – explained by the lack of freedom and/or strong social stigmas against being unaffiliated in the vast majority of the world, and particularly in those places. The most charitable thing that could be said of Pew’s model, perhaps, is that it might predict the number of people who say they’re Muslim or Christian or whatever, assuming the same social stigmas that exist today survive into the mid-21^st century. But of course, those social stigmas are already being strained today. It is not inconceivable that something could snap, and a slight lightening of the social repression against nonreligion could result in a flood of closeted nonbelievers suddenly coming out… exactly as we’ve seen in many majority-Christian countries.

Someone needs to hit the people at Pew over the head with a history book. If you go back a hundred years or two hundred years, all the points they make against the growth of the nones would have perfectly described the situation then, too. If you went back to 1970 – 40 years before the start of their 2010-2050 window – and used their model, there is no way in hell you could have predicted Canada would be almost a quarter unaffiliated by 2010.

I don’t doubt that their model is an excellent tool for projecting the future of religious populations. I just assert that nonreligious populations are fundamentally different, in many ways. The nonreligious are an X factor in religious demographics that cannot be accounted for by birth/death rates or migration, or even by any consistent “switching” factor. I assert that any model that tries to treat the nonreligious as just another congregation, or that just considers them only as an afterthought or the remainder after applying the religious population model, is flawed.

We’ll have to wait a few decades, but I’m confident my prediction will be borne out, and Pew’s projections will be wrong: the proportion of people not affiliated with religion will be higher in 2050 than it is today.