People who report their religion as ‘none’ are rising in the US. Between 1990 and 2010 the size of this group rose 10% points. Concurrent with this decline is the rise of the internet. Are these phenomena related?
This is the claim in a recent paper that has been receiving some interest on the internet over the last few days. Sounds plausible, right?
Except that the relationship between internet use and religious affiliation may both be influenced by a series of other factors, such as age, sex, educational attainment, income, social class position, the religion you were raised in, and where you live. The authors try to remove the effect of these other factors and find that 20% of the observed increase in ‘None’ affiliation is due to the internet. To gauge the size of this effect, consider that the increase in college education accounts for only 5% of the rise. The internet has been 4x more influential in the rise of the religious ‘None’ than the expansion of higher education between 1980 and 2010.
Yet, the authors acknowledge that these results do not prove that the internet is the causal factor. So, what are the alternative explanations?
The most important are cohort effects. Religious affiliation is a classic example of a cohort effect and impact of replacement. If, on average, people born in the 1950s are more religious than those born in the 1970s then the cohort into which you were born affects your religiosity. As the proportion of those born in the 1950s in a particular population declines through death the relative importance of later generations (replacement) increases. The rise of the None is likely largely due to this replacement effect coinciding with the rise of the internet.
What does this mean for Mormons? Surely the availability of information regarding the Book of Abraham, the Kinderhook plates, polygamy, etc. (see any online survey regarding why Mormons leave) is somewhat influential. Maybe. One striking feature of these online communities is that they seem to be overwhelmingly dominated by people of a certain age, born in a particular period. That is, sad as it may be, their disaffection is likely part of this broader susceptibility to disaffiliation that is a feature of growing up in a particular cohort. In other words, communities like Mormon Stories merely provide a particular (and seemingly popular) narrative for disaffection among Mormons that is actually an idiosyncratic version of a broader population trend in the U.S.
The real question for debate is why are those born in the 1970s less likely to affiliate with a particular religion than those born in th 1950s. Although we have some some good ideas, I do not think we have a good handle on that question. Changes in the socio-cultural context is one plausible mechanism and secularization generally is another. While there is some ambiguity regarding what did cause the decline it is likely not the internet.
By Common Consent, a Mormon Blog 
Robert Putnam and David Campbell offer an explanation in chapters 3 and 4 of their 2010 book “American Grace: How Religion Divides and Unites Us” They argue that the recent rise in the religious “nones” is part of a wider cyclical pattern that’s been going on since the 1950s at least. The social upheavals of the 1960s led to a conservative backlash in the 1980s that included a stronger link between religion and partisan politics, including the rise of the Moral Majority and the Religious Right. The college-aged cohort was more politically and religiously conservative then their parents’ generation. This, in turn, led to a more secular backlash starting in the 1990s and 2000s among the next generation. Young people saw the Religious Right and thought “if being religious means conservative Republican politics, then being religious isn’t for me.” According to Putnam and Campbell, this is what has read to the recent rise in the number of Americans identifying as “no religion” which is heavily concentrated today among the Millennial Generation (nearly a third) who are more liberal on social issues than the previous generation. They ultimately argue that if churches want to more successfully retain the rising Millennial Generation they may need to loosen the ties between religion and conservative politics. (Interesting implications for Mormons especially.)
That’s their explanation anyways. It’s interesting to think how the presence of the internet may fit into that story.
In Mormonism, I think the cohort effect is amplified due to the coincidental fact that people born in the 70s were the first generation of LDS people raised entirely on the correlated curriculum.
Benjamin, Putnam and Campbell’s book describes one of the theories alluded to in the OP. Part of my scepticism is that a broadly similar cohort effect can be observed in the UK, where the political context and the culture wars have played out in a somewhat different way. There has not, for example, been a rise on conservative Christians, especially not in politics. Whatever was happening during that period it is likely something that can explain the patterns observed across these contexts. Their particular explanation is a little too US-centric.
Mark, that is plausible, however I have yet to see reliable data to suggest that the rise of the Nones among those raised as Mormon is distinct from this broader trend. The Correlation argument could be important in two directions. Correlation could set people up for a fall when they confront now widely available information via the internet. Or, it may be that correlation is somehow interacting with what created the cohort effect in the first place, e.g., liberal social issues in a culturally conservative church. The second is more likely in my view.
[Apologies if this has been discussed widely, elsewhere; this post is my introduction to Downey’s paper.] How do you get to “it is likely not the internet”? Wouldn’t it be better to say “it is almost certainly not only the internet”? It seems that a generational effect is widely recognized and one of the first to control for. Furthermore, as I read the paper, the key variable is not existence of the internet but use of the internet. In simple terms, it seems that more internet use is associated with lower religious affiliation, even within age groups. The three “tentative conclusions” (religious upbringing increases the chance of religious affiliation as an adult, college education decreases the chance of religious affiliation, and internet use decreases the chance of religious affiliation) all seem pretty reasonable.
It may be interesting to speculate on directionality. Religious upbringing has an obvious direction. College education as less obvious but likely direction. But internet use is not so clear to me. Just reading the bloggernacle, one could pose a reasonable hypothesis that disaffiliation leads to increased internet use.
In another of his books, “Bowling Alone,” Putnam presents data that shows that American involvement in social, community life has been in decline since the peak in the 1960’s. He shows that with each successive generation we are less likely to actively participate in our churches, the PTA, the Rotary Club, political campaigns, athletic clubs, etc.
He also correlates community involvement with community involvement… For example, people who actively participate in the PTA or the Rotary Club are more likely to show up to church on Sunday. It was a very interesting read and, at least for me, had some implications for how our community involvement could impact our missionary efforts.
So I think the decline in religious affiliation is part of a broader trend.
I think the answer lies in a variety of reasons. Certainly, the fact that most young people know–and are friends with–someone who is openly gay causes a rift. Many young people don’t want to belong to a church that openly preaches against gay marriage.
I think other factors mentioned above also play big roles.
Christian, the problem is that internet use and cohort is largely confounded. Further, the internet is largely a period effect (an affect linked with a particular period in time). The challenges age, period and cohort effects is well documented. Hence, this type of modelling cannot give us a clear sense of the magnitude nor the direction of the effects. The internet might play a minor role but the evidence suggests it is not a primary factor.
Confounding and period effect are problems, but aren’t they dealt with? For example, if your hypothetical is that for persons born in the 1970s religious affiliation is not correlated with internet use, isn’t that proposition rejected?
Direction and causation are, as usual, very difficult. Magnitude also. But on magnitude, I thought Downey’s article fairly useful (about half unexplained, meaning generational displacement but without a deeper reason yet, about 25% religious upbringing, about 20% internet use, about 5% college education). That says two things to me: (a) internet use is significant and relevant, and (b) there’s a lot more than internet use to talk about.
Downey’s article is certainly intelligible wrt to magnitude but that assumes he has the magnitude right, and I do not think he does.
The problem is more fundamental. It is mathematically impossible to identify the age-period-cohort effect using this type of modelling strategy. Attempting to do so can lead to wildly spurious associations. Sure, you can adjust for various interactions and test whether they are significant but as soon as you try to adjust for all three – as you would need to do here – the model is not identifiable.
Yeah, the social science statistics world has been blasting this paper (and the reporting about it): http://andrewgelman.com/2014/04/04/going-criticize-blog-im-just-tired-things-like-whats-really-horrible-news-article-takes-seriously-problem/
Andrew, to be in Gelman’s company is a privilege.
I think Mark raises a very important point. Those born in the 1970s and thereafter don’t know what it’s like to grow up in a non-correlated church.
The church’s emphasis on correlation, combined with the wealth of information (not just anti-Mormon information, but real scholarly historical information) available thanks to the internet in the last 10-15 years, makes it understandable that members of this generation would start to become disaffected when the narrative the church has built for itself over the last 40 years doesn’t quite match up to the narrative you start to see when you examine the history.
I think church members have every right to be irritated or even angry that the church has whitewashed its history through correlation. For some, that frustration turns into disaffection. For others, it manifests in snarky blog posts, but no more.
So I think you are saying that there is an identified relationship but it is not possible to estimate the magnitude with the modeling strategy used. If that’s a fair recap, then I will look further into the modeling strategy and mathematically problems (out of personal interest). But also repeat my criticism of your “likely not the internet” conclusion. I would think a “not only” or “not proven” or “personal opinion” would be more appropriate.
Aaron, thanks for this post. I still remember my undergraduate psychology professors inculcating the fact that ‘correlations do not prove causation!’ I think it would also be interesting to look at the internet in terms of pluralism. In some ways, if their was a causal link between internet usage as exposure to multiple worldviews and disaffiliation in America, then it would be something of a blow to Rational Choice Theory. Your description of cohorts and replacement sounds a lot like simple issues of plausibility, a concept that would necessarily entail all of the mentioned variables (i.e., cohorts, cultural shifts, church curriculum, internet usage/availability, globalization, etc.). Love it! Keep this sort of thing coming, Aaron.
Aaron R,
Doesn’t the author of this paper control for the variable of year born? I thought the whole point of regression analysis is to control for variables so that you can see causation.
FZL, while in principle regression analysis can approximate the average causal effect, in practice that is very rarely the case, as the authors note. For example, you could take any variable with a similar linear trend and you would observe a similar effect to the one documented here.
Adam, thanks. The implications for RAT of religion is interesting, particularly because a forthcoming paper has problematised the critics of RAT by observing that the security argument (wealthier countries have stronger social safety nets) does not hold when you examine countries over time. The challenge with the pluralism argument is that people tend to filter new information according to pre-existing beliefs fairly consistently plus the internet allows people to filter what they encounter to a far greater degree. Both of these suggest that the influence of the internet is simply very difficult to identify.
I would think the broader rejection of “spiritual evidence” and obedience to authority that rose out of the 60’s and 70’s would be a large factor, and I agree that generic indoctrination over personal, individualized spiritual growth/understanding within “the ring generation” has to have an impact.
Relying on “the (generalized) answer only can take someone so far (“borrowed light”), and that includes relying solely on academic answers.
Interestingly, I have seen more difficulties within my generation and the one right below me (age 30-50) than within my children’s generations (12-30) – perhaps because the younger generations assume things will change and/or see the current changes and particularly with regard to gay rights issues. Th expectations of infallibility are very different for them than they were for lots of people my age.
“rising generation”
FZL and Aaron: As Aaron says, I don’t see any claim of causation analysis, but only qualitative “it makes sense” or “can’t think of a reason” sorts of statements.
And just to make the fine point, the numerical work is a logistic regression reported in probabilities. Somewhat different from a classic regression. There are interesting questions about what the probabilities mean and how robust they are. For now I wouldn’t (cannot, with any confidence) say more than (a) it isn’t simple and (b) the analysis probably doesn’t say what the popularizing news articles say it says.
I think the internet is being blamed for natural causes. I’m not sure what all the natural causes are, but as an LDS who is in the 7th decade of life I can tell you that the “church” 50-60 years ago was a very ‘safe’ place for most people. There was much less social and economic stratification. I can’t speak for other religions, though I know a few evangelical Christians who are unhappy with the aligning of their churches with corporate interests and neo-conservative politics. People who question and seek truth (even before the internet) have perhaps never been very safe in a homogenous religious setting, but it’s even more important now to navigate cautiously if a person wants to remain part of a faith community. The internet merely provides support for those who wonder if there is a place for them at ‘church’.
Comment #1 is reminding me that I downloaded that book. Now I need to read it.
Aaron, I’ve been grappling with this question too. One thought I have had, not to go all BOM here, is that it also correlates with prosperity. At the risk of sounding like a crotchety old man, kids these days often suffer from entitlement and not having to work for things. It’s easy to be secular and liberal under those conditions. I don’t really feel that strongly about that, but I have wondered how much that correlates with the secularization of society when it is the wealthy nations (and the wealthy among mixed economic nations) that are more likely to become atheist. Religion is still plenty popular among the upwardly striving poor. Given that observation, though, atheism looks mighty good by comparison.
Angela, there is some good empirical evidence that this is the case, i.e., as GDP rises within a country then religious affiliation declines. However, one problem with the argument is that for it to fit the data there should be some cut off point at which GDP per capita begins to create the security necessary to not affiliate with religion. This is difficult to prove.
So it’s not the devils doing?
Another theory is that it has been so long since the world was in a major state of war ( think WWI and WWII). People born in the 1970s were the first generation for decades not raised by parents who lived through a war. Of course, I’m not including the Korean War or Vietnam. World war has a way of humbling people in a way that leads them to a closer relationship to God. As we get further and further away from that kind of influence we are probably likely to see an ever growing group of agnostics.
Brian T – the old “no atheists in foxholes” argument. Another good one.
Knowledge kills religion and the internet provides knowledge our religious masters have tried to keep from the population.
Interested in the claim about being a correlated Mormon may have an effect on whether someone stays with the faith of her childhood…of course my own story, but I was a mid 70s baby, and I was pretty aware of the “difficult” stuff by about 19 years old… I read a lot, got lots of data, and made value judgments based on what I was exposed to… Mostly interesting to read everyone’ s perspectives.
I know plenty of non-Mormon youth who say that they are ‘spiritual but not religious’. They seem to believe in something higher, but don’t like being forced to go to a building every week to prove their spirituality, they don’t like the politicized nature of religion, they don’t like the ways that religion tries to tell them what to eat/wear (or what they shouldn’t eat or wear). They also grew up in a modern developed society with many choices. To them, church isn’t the only place to potentially go to if you want to meet people or have a strong social network.