Part of this is a credibility signal. In 2012? there was a mid tier polling firm that was just making it up. They would take Gallop's latest numbers and make it 1.5% more GOP. They didn't publish the cross tabs, which was the first clue that something was off.

Also having done polling for a newspaper...the breathless stories about subgroups fill column inches and make the math work on paying for the poll in the first place.