How Many Atheists Are There — and How Do You Count Them

Based on a video lecture from 2020.

Atheism is a fascinating phenomenon from a sociological standpoint. It is the rejection of religion — and while in some countries this carries no particular stigma, in many others it remains a loaded word. Few people will voluntarily call themselves atheists, even if that is effectively what they are.

Here I want to talk about the difficulties of counting atheists, primarily in the United States, and to work through one scientific paper with an interesting methodological finding.

Why Atheists Are Hard to Count

Problem one: unbelief can still carry reputational and social costs. When Dawkins publicly identified as an atheist and criticized religion without restraint, he was attacked from all sides — by atheists who considered him too radical, and by believers who could barely believe what kind of person he was.

This is connected to the fact that religion is still strongly associated with morality. For many people, a non-religious person is an amoral person, someone without ethical grounding. There is actually a body of research documenting this association in public consciousness.

Problem two — and this one is less obvious: even in anonymous surveys, atheists tend to describe themselves as believers. People present themselves as society expects them to be. This is a well-documented phenomenon in sociology.

The result is very different figures depending on method:

• According to Pew Research Center (2015), 3% of Americans identify as atheist in a direct survey. By 2023, this figure had risen to 4%.
• When offered a binary question — “Do you believe in God, yes or no?” — 11% answer no (Gallup).

The range from 3 to 11% looks significant, but these are still very small numbers. Which raises the question: what would happen if we could methodologically eliminate the social pressure factor?

That is precisely what the authors of the paper in question attempted.

Hypothesis and Method: The Unmatched Count Technique

The paper’s hypothesis is simple: if atheism is measured using methods that exclude social pressure, the figure will be substantially higher.

To test this, the researchers drew two samples totalling 4,000 respondents (data provided by YouGov, a reputable panel) and applied an indirect method called the Unmatched Count Technique (UCT).

The logic is as follows. Participants are split into two groups:

Group one receives a set of innocuous statements — “I ride a motorcycle,” “I exercise regularly” — and marks how many apply to them.

Group two receives the same statements plus one socially sensitive one — for instance, “I use substances.” They also mark the number that apply.

The difference in average responses between the two groups indirectly reveals what percentage of people use substances — without ever asking them directly.

In this study, the sensitive statement was: “I do not believe in God.”

The Experiment

Sample 1: participants were randomly divided into three groups.

• Group one chose from 9 innocuous statements those that do not apply to them.
• Group two chose from 10 statements (the same 9 plus “I believe in God”).
• Group three gave a direct binary answer.

Sample 2 was designed for cross-validation. Three groups: baseline (6 innocuous statements), critical condition (6 + “I do not believe in God”), and a reliability check group (6 + a self-evidently absurd statement, something like “I do not believe that 2+2 < 13” — to test whether the method itself was functioning).

The authors applied Bayes’ theorem for probabilistic statistical estimation.

Results — and a Puzzling Discrepancy

The authors’ overall estimate: 26% “indirect” atheists — against a maximum of 11% from Gallup. A factor of 2.4 to 8.7 times higher. Striking.

The explanation for the gap is likely the stigmatization of atheism: people do not call themselves atheists even anonymously, when the question is asked directly.

There is also one outright puzzling finding in the data: in the reliability check group of Sample 2 — the one containing the absurd mathematical statement — participants selected more applicable statements than in the other two groups. The authors have no explanation; they note it requires further investigation. Neither do I — if you have a theory, I would genuinely like to hear it.

Conclusions

The UCT approach clearly needs further refinement. But the underlying logic is an important step. The point of science is not to build a pretty model — it is to get closer to the truth: how many people actually do not believe in God but do not say so out loud?

While direct surveys give us 3–4%, indirect methods suggest the real figure may be several times higher.

Reference: Gervais W.M., Najle M.B. (2018). How many atheists are there? Social Psychological and Personality Science, 9(1), 3–10.