“”no self-respecting pollster” weights their polls to make party ID even. “
That, of course, is pure b.s.
I happen to know quite a bit about statistical surveying (although, I’ve never done an election poll). There are many potential sources of error. Most media outlets at least know about “sampling error” — or, they at least know enough to report the “margin of error”. Unfortunately, sampling error isn’t the only, or even the most serious threat to the validity of a survey (poll).
“Non-response bias”, and it’s opposite: “voluntary response bias” are common, potentially far more serious than sampling error, and much more difficult to detect or fix. Based on several posts on FR, it appears quite likely that these biases are operating in many polls in this election.
For instance, conservatives are probably more likely to hang up on a pollster — in the belief that the poll won’t be fair. They thus become “non-respondents”, and introduce a “non-response bias” against conservative voting results.
Meanwhile, “progressives” acting in the belief that pollsters are fellow travelers, are more likely than average to respond. They want their views to count; and thus introduce a bias in favour of the liberal candidates. This “voluntary response bias” is almost certainly in play during exit polls. The conservative voter believes his voting record should be kept private; meanwhile, the liberal will saunter past the New Black Panther Party intimidators, and button-hole the hippy with the clip-board. Result? Worthless exit polls.
Contrary to what Wallace says, a “self-respecting” pollster will try to anticipate and resolve these bias problems. In the case of election polling, the pollster should compare known percentages of registered Democrats, Republicans, and independents with the percentages of each in his sample. If these percentages don’t match closely, it’s likely that some sort of response bias is operating. The pollster should try to fix that. One fix (not always the best) is weighting the samples.
Relying upon voice phones as the now somewhat antiquated technology to conduct polls themselves is flawed IMHO, especially with modern alternative communication technologies like texting, emails, or forums/blogs that many younger folks & some older have begun to favor at the expense of voice communications. Armed with now almost universal caller ID, myself & many others that I know rarely answer a phone anymore for a voice call. Only if the caller ID matches an actual contact or is a person I recognize - else leave a damn voice message if it's important. And, those I do know are aware that I am much more likely to respond in a timely manner if it's a text or email, so I don't get many voice phone calls to begin with. I suspect I'm not alone.
Moreover, with caller ID I think the chance of someone even inclined to answer a phone call at all from someone they either don't know, or that is obviously a telemarketer/money-begger/pollster, is a strong function of personality. I like to be left alone in general, as I suspect many non-collectivists do. IMHO, phone polls are going to be dominated by touchy-feely-talky types that are inclined to answer an out-dated anonymous voice call in the first place. We're not in the 20 century anymore - technology has changed, and poll technology somehow needs to change with it.