To address Crazy Anglican's sampling concern:
"In our sample, 23.9% of households identified as Christian (n = 280), 43% as Muslim (n = 510), 27.6% as not religious (n = 323), 2.5% as Jewish (n = 29), 1.6% as Buddhist (n = 18), 0.4% as Hindu (n = 5), 0.2% as agnostic (n = 3), and 0.5% as other (n = 6). Results from an independent samples t test, comparing altruism in children from religiously identifying (Msharing = 3.25, SD = 2.46) and non-religiously identifying (Msharing = 4.11, SD = 2.48) households indicated significantly less sharing in the former than the latter (p < 0.001). To further investigate these effects within specific religions, three large groupings were established: Christian, Muslim, and not religious; children from other religious households did not reach a large enough sample size to be included in additional analyses."
So they threw out the Jews, Muslims, Buddhists, agnostics, and "other" households. Elsewhere in the study n=323 is used for comparison of the "not religious" group to the Christian and Muslim groups, so the households in the "not religious" group used for statistical analysis are the same households that originally identified as "not religious."
To be clear, I don't take the study as dispositive and I don't take for granted that the methodology is sound, but sampling bias has nothing to do with it.
Also, re: the voter's box--IMO it has a lot more to do with trying to figure out the candidate most likely to be elected and implement the policies most in line with what you want government to be doing/not doing. I don't really care if the President is a scumbag except insofar as it affects his policymaking and diplomatic relations.