The U.S. is “retreating from religion,” according to a fascinating piece on Scientific American. Based on one expert’s analysis, it’s possible that by 2030, a third of Americans might have no religious affiliation. By 2036, the amount of people who report having no religion could be the largest “religious” group in the U.S.

The Scientific American piece was written by Allen Downey, a professor of computer science at Olin College in Needham, Massachusetts. Downey built models using data from the General Social Survey, a yearly survey that the University of Chicago has conducted since 1972. The survey asks around 1,000 to 2,000 American adults a series of questions, including ones about their religious beliefs and attitudes.

Since 1990, the number of Americans who report having no religious affiliation has nearly tripled, from 8 percent to 22 percent in 2016, Downey writes. Downey refers to these people as “Nones,” meaning that they marked no religion in response to the question “What is your religious preference?” Other options for that question include Protestant, Catholic, Jewish or other.

“Over the next 20 years, this trend will accelerate,” he writes. Based on his estimates, there will be more Nones than Catholics by 2020 and by 2035, Nones will outnumber Protestants in the U.S.

While Downey doesn’t get into why these changes are happening, he does mention some interesting trends that could help paint a more complete picture. For instance, in parallel with the rise of the Nones is a declining respect for religious authority, decreased participation in organized religion and loss of religious belief, according to the survey’s ongoing findings, Downey writes in the longer version of the article that appeared on his personal website.

Plus, there’s a definite age aspect: “Among people born since 1980 there are more Nones than Catholics, and among the youngest adults, there may already be more Nones than Protestants,” he writes. (Downey also wrote about this in another article for Scientific American focusing on how college freshmen are getting less religious.)

There are more atheists and agnostics today, Downey writes, and people’s views on God are changing. One of the questions in the survey asks “Which statement comes closest to expressing what you believe about God?” with five corresponding answers, ranging from “I don’t believe in God” to “I don’t believe in a personal God but I do believe in a Higher Power of some kind” to “I know God really exists and I have no doubts about it.”

The amount of people who chose that last statement, unequivocally believing in God, has declined: in the 1990s, 64 percent of people answered that way, while just 58 percent answer that way in 2016. Among people born prior to 1940, more than 70 percent “profess strong belief in God,” Downey writes in the blog post. “But this confidence is in decline: among young adults fewer than 40 percent are so certain, and nearly 20 percent are either atheist or agnostic.”

But it’s interesting to note that this doesn’t necessarily mean that people don’t believe. Downey points to an established pattern called “believing without belonging” that describes how religious beliefs haven’t changed that much over time, but fewer people are part of organized religion today. That could also be changing: Downey writes in his blog post that “it is clear that strong belief in God is declining and being replaced by weaker forms of belief and non-belief.”

The actual number of Nones could be even higher than the survey shows. That’s partly because of something called social desirability bias, which can affect the outcome of surveys, Downey writes. People are likely to give responses that present them in a socially acceptable light. As atheism and agnosticism become more common, more people will likely feel comfortable answering that they’re part of the “None” group.

While these are just estimates, and the implications of these findings aren’t clear, it’s an interesting insight into how our belief systems are changing, especially among younger generations.

Read the Scientific American article here and the blog post here