I was describing a 15 point flip, but that's because I was conflating the unfavorability polls with the GE polls (which I didn't fix after realizing this with the edit). You clearly failed to spot the simple discrepancy when you pointed to the +7pt margin (I presume for the most recent CNN poll).
Read more closely. I wasn't pointing to any particular poll. I was making a general point. If, hypothetically, Clinton's actual margin is +7%, it wouldn't be unusual (or indicative of a polling conspiracy) to see anything from Clinton +15% to Trump +1%. That was in response to you jumping to that conclusion on the basis of a minor difference in two polls.
I don't require an explanation of standard deviation.
Your post indicates otherwise.
What would be more interesting, if you really wanted to dispel such a notion mathematically, decisively, would be to track the history of Rasmussen vs. CNN polling (either for general elections or favorability ratings) and establish whether-- accounting for varying margins of error-- if there is no statistical trend one way or the other that deviates beyond the expected threshold.
Again (sigh), you misunderstood my point. The fact that polling group A and polling group B have the differences in results that you outlined doesn't say anything at all about the existence of a polling conspiracy. The fact that you think it does is another example of the kind of flawed thinking that is too common here. The kind of thing that in another era would lead to you to think that a dead rooster on your lawn is a sign of a coming invasion.
Also, even a mere few weeks often separate such significant incidents as yet another mass murder by terrorists abroad. My instinct would be to presume this might be more likely to describe a reversal than polling body differences.
Well, good luck with that. My instinct is to presume that an individual layperson's analysis of an extremely difficult and complex subject like how world events will cause some people to change their minds about who they'll vote for in November is completely worthless (and I wouldn't put much stock in expert analysis either). Betting odds and polling analysts are the best guides.
http://www.oddsshark.com/entertainment/us-presidential-odds-2016-futures
http://fivethirtyeight.com/features/donald-trump-has-a-20-percent-chance-of-becoming-president/
Since both candidates are known to have a chance (a +260 underdog winning wouldn't be an Earth-shattering event) and humans generally have a great ability to interpret events as confirming their pre-existing biases, I don't expect anyone will learn anything here.
An interesting realization I had when discussing the state of the race with Judo was that if someone genuinely believes that polling is nearly irrelevant--particularly at this point in the race, but it applies broadly--
no particular outcome could ever be seen as evidence that they're wrong! That works both for "polling skeptics" and "polling CTers" (the difference is just--"polls don't tell you anything" vs. "polls are not reported accurately"). If you look at results as a whole, that's a different story (you'll be forced to the conclusion that polling works), but people don't do that.
Interesting that Nate Silver is giving Trump basically no chance.
Between his call and the polls all summer, we'll get a pretty clear picture of how things will pan out.
Silver is giving Trump a 20% chance. That's way better than "basically no chance."