You’ve certainly heard of the ARIS 2008 survey from Trinity College. One of the more interesting aspects of the survey is the demonstration that there is a sex difference in patterns of religions identification. Below I give some links where this has been discussed, but I want to note that in many discussions one of the first things people say … quite reasonably … is that the differences seem small and potentially well within the normal sampling error of a survey.
The reason people think that is because they are accustom to survey data in relation to political polling where sample error is usualy 5% or 3% because of standard methodologies and sample sizes. The ARIS survey has a much lower error rate. The lower error rate does not mean the the conclusion is stronger than it might otherwise be. It means that it is less likely spurious than it might otherwise be. But, in this case, this certainly means that the differences between groups (mainly male vs. female) can not be written off as statistical artifacts. (Though they could easily be of little consequence. Or not.)
In order to help move this discussion beyond sample size questions, I am hereby reproducing salient portions of the Methodological Note for the recently widely discussed and blogged about survey “American Nones: The Profile of the No Religion Population.”
- The American Religious Identification Survey … is a random digit dialed (RDD) survey of a nationally [US] representative sample of 54,461 adults. Of those, 7,407 are Nones, or individuals who responded to the question: What is your religion, if any? with “none,” “atheist,” “agnostic,” “secular,” or “humanist.”
- For these 7,407 people, basic socio-demographic information was collected (e.g., age, sex, etc.).
- A randomly selected, nationally representative subset of those 7,407 cases, 1,106 people, was asked an additional series of questions on behaviors and opinions that provide further insights into the profile of Nones. The subset is a nationally representative “silo” of Nones.
- Additionally, a random subset of the overall 54,461 participants totaling 1,015 individuals were asked some of the same questions as the None silo. This “national silo” is a random subsample and is representative of the general U.S. adult population.
- The 1990 data in this report are from the National Survey of Religious Identification; a nationally representative sample of 113,713 adults among whom 9,899 self -identified with one of the above None categories.
- The sampling error for the full ARIS 2008 is +/- 0.31%. For the No Religion sub-sample, the sampling error is +/- 2.38%.
The following blog posts have discussed this survey, and include interesting comments.
- Are Women Less Skeptical?
- Americans Who Don’t Identify with a Religion No Longer a Fringe Group
- American “Nones”, sex differences
- Fear the atheist
- ARIS’s ‘American Nones’ and Politics
The actual survey is available as a PDF file. Just click here and it’s yours!