Submitted by Salil Mehta via Statistical Ideas blog,
Sea of faulty polls
In this article we cover the theoretical bases for two interconnected ideas that we've discussed recently:
(a) that the empirical polling results are not as dire as current landslide mainstream media projections make it out to be, and
(b) many polls are oscillating about impossibly low probabilities right now for Donald Trump.
This year is genuinely unique in merging several fundamental aspects, with a largely disenfranchised voting base across the country (i.e., record undecideds), and pollsters unable or unwilling to properly assess the true probability for Mr. Trump (and their incoherent polls evidence this). This is not a matter of apologizing for the ground-level odds currently shown by mainstream media, or that the average Hillary Clinton lead is merely unsustainably high. This loses the forest through the trees, as we theoretically prove here.
Start by studying a sample of the general election polls below, taken in just the past couple days.
Do you see anything wrong there? If you don't, then you have no business being around polling data.
http://www.zerohedge.com/news/2016-10-22/statistician-warns-americans-ignore-capricious-polls?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+zerohedge%2Ffeed+%28zero+hedge+-+on+a+long+enough+timeline%2C+the+survival+rate+for+everyone+drops+to+zero%29
| 
