More discussion of how wording affects polling results. Unfortunately, as Liberman has pointed out, none of this addresses the fundamental question of why. Why do the words "homosexual" and "welfare" cause more negative polling results than "Gay Men & Lesbians" and "aid to the poor"? My own weak attempt at a first pass answer (in the LL comments) is that "in both cases cited ("Homosexuals" vs. "Gay Men & Lesbians" & "welfare" vs. "caring for the poor"), the first, seemingly more controversial term is a single word and the second is a phrase. It may be the case that we silly humans find it easier to attach strong emotional semantics to a single lexical item. One could imagine a study that looked at the role syntactic heaviness plays in survey response."
It may also be the case that longer phrases are harder to categorize.