[justlurking]

Another point - when they take a poll, they use the percentage of people who self identify to a specific party. If they randomly call 800 voters, if 8% self identify as more democrats then that is the percent that is used.

Yes, and that's why the party identification is important. If the ID breakdown is significantly different than recent comparable elections, you have to at least consider the effect on the outcome of the poll.

("Push polls" become obvious when you consider this metric. Some pollers conduct these "polls" to reinforce a pre-determined outcome, and make their candidate look good.)

And, you have to consider whether this is a registered or likely voter sample. At this point, registered vs. likely is probably meaningless, because I don't know how one can reliably measure that 5-6 months early. But, it can be a useful measurement of voter enthusiasm.

But, since a random phone poll only uses people that accept the call, it tends to over-sample people that are at home and answer the phone. This almost always oversamples Democrats.

[Dave W]

Agreed. However, I believe all polls use cell phone numbers now. They were forced to since so many have cut their landlines. If I remember correctly all of these polls are RV and for the reason you stated, most polls don't switch to LV until Labor Day. I do believe there are bad polls out there, but the big names are reputable not that they don't try to twist the meaning of a poll.

I think one of the big stories 50 years or more from now is just how corrupt the media is. I work in the media, but I am ashamed of how biased the media is in general.

Of course, the media has always been biased - yellow journalism in the 1800s, etc. Nothing new but it is so pervasive and blatant.