Fatal Flaw....Romney winning women by 3% and losing men by 7%? That is where you see a sampling error. And a fairly substantial one I would say.
Polls aren’t necessarily accurate when divided up into smaller demographic groups. That doesn’t mean the overall total isn’t accurate. Naturally, the sample of men or women in the poll is only half the total sample. And if the total sample was just the minimum quantity needed to be accurate, half of the total will be only half what is needed to be accurate about each gender’s preferences. Rasmussen would need to double his original sample size in order to get accurate numbers for each gender. Otherwise the original poll can only be accurate for the group it defines, i.e. voters, not for smaller sub-groups within it.