If you’ve paid any attention to surveys that ask what religion people are, you know the responses depend a lot on how the questions are worded. Consistency matters. Not being prodded to say “Christian” by default matters. Who’s doing the asking matters.

So when you’re trying to find out what percentage of Americans are non-religious, there are a lot of factors you need to consider.

No one has dug as deeply into this as the ever-prolific Ryan P. Burge of Eastern Illinois University, who just posted about two major data sets he’s been analyzing for a while: the General Social Survey (GSS), which has gone out at least every two years since 1972, and the Cooperative Congressional Election Study (CCES), which has been around since 2006.