There may be (and probably are) some polls being conducted that really are based on good statistical principles and have the goal of truly understanding voter preference.
But there absolutely are other polls that are “pay for play” operations that poll until they get the outcome their customer is paying for and wants to hear. Or they rig the samples and weightings to the same end.
And I am certain we are presented with “poll data” that is totally made-up, fabricated, phony-facted BS, no different than the old “hockey stick” climate data was.
So Caddell is right. The problem is, “we” don’t know which is which, and how the game is rigged. For example, a pollster could easily rig the outcome to favor Trump, only to start tilting it the other way over time as a means of discouraging and demoralizing Trump supporters. When we have no way to tell which of the above 3 categories a poll fits into, it’s impossible to trust ones that favour our side and not the others. The only thing to do is to distrust them all.
And here’s another clue: look at the outfits that put their name on these polls. How many of them are favorable to conservatives?
Well, four years ago, most all of the polls, except for 2 or 3 polls favored Romney by a point or two going into election day. Romney lost by 4 points, so the polls were skewed 5 or 6 points in the Republicans favor. They were wrong. We certainly can’t claim they were biased against Republicans. Maybe it is just too difficult now to poll accurately.
A lot of it is voodoo.
1. How do you poll people who only have cell phones?
2. How do you sample so that it represents people who will vote.
3. How do you account for voter enthusiasm?
The pollster’s preconceptions will determine a lot of that.