Razib's regression analysis on religion and abortion attitudes across countries is pretty illuminating in this regard.

1) It is clear that religion correlate with opposition to abortion in the vast majority of nations.

2) But, the attitudes of religious people and non-religious people track each other so that the irreligious in nation X may oppose abortion much more than the religious in nation Y.

The hypothesis here isn't that personal characteristics determine policy preferences directly, but that they predispose a person to be more receptive to (e.g.) left vs. right arguments within their society. Context is everything, though: leaning conservative in Sweden means endorsing policies that would be radically left in the USA.