i know that people will point to cell phones as a problem with polling, and a lot of the concerns they point to were at one point valid, but online "polling" is simply not a way forward. the best polling firms have actually found ways to poll cell phones and don't really see it as a serious barrier anymore.
the problem with the online polling is that there's no concept of randomness. the math underlying polling needs a random sample to make any sense. you can basically take anything with a "credibility interval" and toss it in the trash - the correct credibility interval for an online poll is not credible. ever. if it's close, it's by chance.
the rcp average is consequently just a polluted mess that should be avoided. i don't know if they've addressed this or written about it. but, they shouldn't be aggregating credible polling with this online propaganda-generation bullshit.
what that means is that the number of polls you should take seriously is a lot smaller than the media is going to throw at you. when the media throws a poll at you, immediately check to see if it's telephone polling or an "online panel" and respond accordingly.
the reason they report "credibility intervals" is that they can't measure error because the sample isn't random. it's just a pr tactic to make "online polling" look more scientific than it is.
it is not the same thing as a margin of error. don't be confused by this.