Going in the right direction – but I wonder if some of that might be fallout from the Khan affair, wherein Trump hits a new low of such despicable callous boorishness that it’s surprising even for him.
It might well be. Combine that with a convention bounce - not yet determined - for the Hill-meister and I’ll take it.
Silver was on On Point a couple days ago, he said that at least historically, polls tend to “firm up” about three weeks after the conventions. He said 2012 was kind of an atypical year, in which polling largely firmed up earlier and never changed much the entire election. Romney got a bump after the first debate but it was more of a “media bump”, while he did get a bump in a lot of polls the 538 model never showed him winning even after the debate.
But he did offer the caveat–he’s just going off past elections, where a couple weeks after conventions polling tends to stop moving very much, and this election isn’t guaranteed to be anything like the prior ones.
Which is funny, because everyone always assumes any unpredictability favors Trump. Which I can understand, but if we’re REALLY throwing EVERYTHING we know of political convention out the window, surely Clinton must be the recipient of SOME of those breaks.
Looking better, but still essentially a coin flip. Those numbers still scare me. Remember, that’s chance of winning, not share of electorate. Hillary at 49% vs 51% makes no practical difference. Trump winning 4 out of 10 does not comfort me, either, but we’re still far enough away from the election. Let’s just hope this is the beginning of the trend line bouncing back to where it was in June.
The Tump vs. Khan thing began, what, yesterday? Id be surprised if an analysis of polls just released this morning reflects the impact already.
The Khan speech was Thursday, the last day of the convention. I’m not sure of the timing of the subsequent events and some of the reporting is confusing, but it seems that by Saturday night Trump was already trying to backpedal so his offensive comments would have been earlier, and might well have influenced the latest poll.
Regardless, this really is a despicable new low for the guy.
ABC released the Khan statements Sat morning as a tease to their Sunday show, which is when the real controversy started. So while any impact Khan’s speech may have had may be reflected (which I doubt would actually happen), I don’t think the controversy has had time to make an impact.
This is a simply false statement. He specifically has done that analysis using two different time frames. That is specifically what makes up the difference between the random drift ans his Bayesian odds, it is specifically how often a polling lead this strong has held up in the past. Silver misrepresents Wang’s model.
Silver’s system is simply statistically noisier. It responds aggressively to short term changes in polls which serves Silver well as it will thus change in response to blips much more strongly, is more volatile, and make the data more exciting, giving readers more of a reason to give more clicks, which Silver needs and Wang does not.
Trump got a bump that was average to low average for convention bumps … to report that such increased the odds of his winning highly significantly is simply silliness done to increase reader interest. Clinton’s bump the other way is looking to be a bit bigger and certainly not smaller. Polls doing pretty much exactly as you’d expect should not result in the odds of the race changing by much.
Neither system is perfect and both perform well. Silver’s will pick up on a real change earlier and report insignificant changes as potentially meaningful more often. Wang’s will be slower to respond to real changes and less likely to overreact to noise.
You’re misrepresenting 538. The model (which is different from the nowcast, which they explain is just for shits and grins), does I ignore the convention bounces.
I am sorry but you are mistaken. It does not ignore convention bounces. Perhaps you are thinking about how the “polls plus” approach attempts to adjust for them (not ignore them), along with adding all of the non-numbers based secret sauce “fundamentals” punditry?
Hence “polls plus” runs 61% Clinton, not too much changed from where it was before the conventions (July 14 was 63%), while “polls only” has changed more dramatically (from 66% July 14th to 51% now).
I don’t know why the Khan thing is supposed to change anyone’s mind when encouraging his supporters to beat the shit out of people they disagree with didn’t matter and saying that Muslims need to be registered in a database didn’t matter.
Maybe more people are paying attention now?
The people who don’t vote in primaries are paying attention now.
And a nameless faceless “other” is easier to demonize than one with a face and a voice and who shares and articulates the values that we only hope most other Americans are patriotic enough to share.
As of 9:00 CST 538’s nowcast has it thusly:
Clinton - 63.6%
Trump - 36.4%
Polls only is at 53.3/46.7 and polls plus is at 61.8/38.1.
Looks like Hillary is getting a pretty good post-convention bounce.
I suspect Trump is also experiencing a rather hefty post-Khan drop.
And Trump’s statements about the Khans probably haven’t even hit the polls yet. He only started talking about them (and therefore prolonging the story) on Saturday.
Khaaaaaaaan!
Sorry.
It had to be said. 
This. The winner of this particular election will be the candidate who is able to convince undecided voters that they are the less insufferable of the two, and things probably just became a lot easier for Hillary.
FWIW Wang estimates Clinton’s bounce to be about 7 points, which “by current standards … is large.” Net from conventions is into positive territory for Clinton. And yes that is before Trump’s horrific attacks on Mr. and Mrs. Khan have had meaningful impact. (Still concerned about how well Mr. Khan was vetted, Derek?)
His model has been pretty stable reading it as 60 to 65% under random drift and 80 to 85% assuming regression to the mean of long term polling and historic standard deviations from that (Bayesian).
Even by RCP rolling average technique the race is minimally back to where it was before the conventions and near to before the DNC email leak.
538’s “polls plus” (which includes attempts to adjust for convention bounces, and some “fundamentals”) has stayed within a few points of Clinton with a 62% chance since July 14, which is also where it was in early June.
The odds of the race results as told by number crunching have been pretty dang stable: no slam dunk but Clinton a solid favorite. Team Clinton needs to execute and not commit any major errors; needs to exploit the errors that Trump makes (often despite his team I think). But actual large swings in the race are greatly exaggerated.
I have a fair degree of confidence that the race will settle back into polling Clinton +5 to +7 range with rare excursions of aggregated results outside that range from here. As it does that much closer to the race the odds of its changing much goes farther and farther down.