"Each state poll was conducted by phoning 625 likely voters in the final week of September"
What are the chances that those were "registered" voters?
Exactly...and 625, or even 1000, as a sample size is oh so representative /sarcasm.
Polls are only good for strippers.
What's the difference between "likely" and "registered" voters? Is it that "likely" may not be "registered," and, if they're not "registered," why bother to ask them? They can't vote unless they're registered. what am I missing here?