Well, that's only a piece of things. When you are trying to create a vision of consensus your statistics are no better than the hype created at the horse races to coerce bets. If polls are that limited they will only be effectively accurate through scientific means of repetition. One poll is worthless alone and with the role of bias, it's hard to say if the averages across the board are even right. If you discard the results that are way off, maybe.
I'm curious, do they poll people at random based on their registration or did people sign up? Polling descriptions are often vague. I think the latter would be more successful.
so again, to reiterate
1) you only need a sample size of around 300 people to be accurate for a population of millions within a relatively small margin of error. 1000 basically is a 99% confidence interval
2) any poll that has any merit, even polls rated with like an F by nate silver, will use random sampling. By definition any scientific poll uses a random sample. If they end up over sampling a particular subset of the population, they then weight the results. I say this all the time on JPP and no one ever listens.