Improving the Accuracy of Identifying Atheists in Survey Research

survey research

Austin Cline (About.com Agnosticism/Atheism) recently compared Gallup survey data over several decades to show that increasing numbers of Americans are walking away from religion (update: link no longer active). Good news indeed, but that isn't what I want to address here. In his post, Austin brought up the reluctance of many Americans to identify themselves as atheists, even though some undoubtedly are atheists. This is a problem plaguing this sort of research, one I've recently been considering.

From the perspective of someone writing a survey, religious identification is usually quite simple to measure. One simply writes a question that includes many different religions as response options and asks respondents to select the one that best describes their views. But the apparent simplicity evaporates as soon as one comes to atheism. A religious identification question with a "none" option is going to get far more responses for that option than would one with an "atheist" option. The stigma around atheism is so great that many American atheists simply do not perceive themselves as being atheists.

Austin asks, "…how might research firms obtain better, more accurate information about atheists?" The most important part of my answer is that researchers need to determine precisely what they are interested in measuring. How many people identify themselves as atheists is a vastly different question from how many people are atheists. Are we interested in one, the other, or both?

Assuming we want both, we need more than one question. To determine how many atheists we have, I'd do it by adding a question like this:

Do you believe in any sort of god or gods?

The response options to this question would be dichotomous (i.e., yes or no) because an "I don't know" or "I'm not sure" in this context would not make any sense. The question is about what one believes. Those who answer "no" are atheists, whether they identify themselves that way or not. This would let us know how many atheists there are.

Next, I'd include "atheist" and "none" as response options to our religious identification question. Why? Because this would allow us to determine how many atheists identify themselves as atheists. And that would be nice information to have.