The thing is, whether Tommycat had a source or not is irrelevant. The data are not available for us to look at. Indeed, I'm highly skeptical of the claim that atheists drop out of school at a higher rate than the superstitious. At the very least, correlation doesn't imply causation and other considerations might be controlled for if the data were available, like were the "atheists" from traditionally religious or irreligious homes; economic status; ethnicity; gender; etc.
The one source that Tommycat provided was a tertiary source that used the General Social Survey to show a trend of conservative/liberal men/women. The trends shown claimed to be IQ but the methodology is questionable since it relied upon several vocabulary questions rather than an actual test instrument designed to gauge a g factor. The variables used were WORDSUM and POLTVIEW if anyone bothered to track down the survey and do it themselves.
But, and more importantly, the survey was not relevant to the current topic nor the subtopic in the other thread which correlated cognitive function to conservativism since the tables at the tertiary source evaluate conservative males vs. conservative females and liberal males vs. liberal females. There was no comparison between conservative to liberal that I saw (though I would need to go back and look again). Even if there were, the GSS survey questions are not a valid or accepted instrument to gauge g factor. The CAT and the ASVAB are.