Highly unlikely. The Real Clear Politics chart shows that the polls favoring Clinton covered only though Sunday (and in one case Saturday), hardly enough time for Trump’s statement (announced on Saturday but formally aired on a Sunday morning talk show) to get effective blowback. These polls are almost entirely convention bounce. It’s the next set of polls that will include the reaction to Khan.
You mean where the red line looks like a sperm cell that lost the race? Facing the utter futility of brief existence?
tim314,
What I am in reference to is what underlies the Bayesian model. Details here, here, and here. He has calculated what the standard of error of current polling at various points has been relative to predicting actual results.
The snapshot is used as the starting point but otherwise is not too much different in spirit than 538’s nowcast. His probabilities of what happens in November explicitly take into account how much chance there is, based on long historical view, for how much change, and explicitly includes an understanding that states may move together and are not independent events.
When Silver states that Wang does not take into account how accurate polling averages have been in the past he misrepresents Wang’s methods.
As to the concept that polls in general may be skewed somehow and that Wang does not adequately allow for that systematic error … I fail to see any empirical evidence that polling in aggregate has ever been significantly skewed in U.S. Presidential general election. Okay, maybe in 1980. I would agree with the point that it could happen, that pollsters in aggregate could be using an incorrect likely voter poll, for example … as occurred in the Democratic MI primary … which 538 did not allow for either.
Not sure though how 538 can quantify that uncertainty. Actually pretty sure they cannot. It is just an opaque adding in some level that he decides feels right, or more precisely, makes up. Wang, for better or worse, goes with the approach that there is no evidence that it does exist to a significant level and that the only completely transparent manner to handle it is to assume that pollsters will do as good of a job as they have in the past. Whether or not this election has the electorate behaving differently than they have in past elections, screwing up the likely voter screens, is not something that can be empirically numbers based and is more punditry. Does not mean that it cannot happen but it does not belong in the systems that Wang and Silver put forth. (Unless Silver was explicitly transparent about how much co-variation he was assuming was reasonable to expect and provided the data to back that up - if he has done that I have missed it.)
More simply, when Silver claims that Wang assumes the error in polls is independent from state to state he misrepresents. Wang’s accurate take is that the snapshot EV result is the same whether co-variation is zero or is 2% one way or the other and that the much larger source of variation from current to Election Day that will vary state results is the national vote … adding in co-variation as a factor adds much complexity for relatively little added probability value.
FWIW historically Silver has less confident error bars than he should. I distinctly remember in 2008 that he was right in more states than he should have been given the probabilities he gave for each state. Yeah, everyone was impressed he was right so often, but if his model was calibrated correctly he should have been wrong a few more times.
As to dinging Wang on being a scientist rather than a “statistician” … you do realize that Silver does not have a degree in statistics either don’t you? He has a BA in Economics. His great experience in statistics was honed not as an academic but as a baseball fan. They both still know their shit.
He geeked me. I’m totally covered in green math nerd slime.
The McCain incident should already have done that. Methinks this is a convention bounce.
I agree. The damage to Trump caused by his comments haven’t had time to be polled yet… those might come by Friday, definitely next week.
No.
There won’t be any. He already questioned a white guy’s service and got away with it.
What was striking to me is the difference in favorable/unfavorable ratings. Clinton’s unfav was higher than fav, but marginally, whereas Trump’s unfav was not only higher than his fav, but far more disproportionately so.
Re: fall off from the Khan thing
Yes, but that was long ago, political-time-wise. Last year some time. Trump’s numbers went way up in June/July so those are new supporters, many who may have forgotten the McCain thing. He could lose many of those new supporters.
I don’t think it works like that. The new support came from fickle voters who want to participate in the political process but don’t really know much about the candidates or the issues, or they are refugees from the camps of other fallen republican opponents and even some #neverhillary voters. Sure, some will be so repulsed by the Donald that they probably simply won’t vote, or will vote for Gary Johnson. But the majority, I’d say, are not turned off to the point of voting against him. These voters are surely aware of the ways in which he has gone way past conventional boundaries of what is considered inappropriate, and in many ways that has been his appeal. A lot of people – even people who aren’t necessarily tea party types committed to seeing the extermination of liberalism – appreciate that about Trump. He’s ‘different’. He talks like a guy in a bar who’s had a beer or two and tells you candidly how he feels about whatever you want to talk about. Sure, at times he talks like he’s had one beer too many, but you forgive him the next morning, chalking it up to occasional excess. Trump may be flawed, but the minds of the voters who will determine this election, he’s authentic, which unfortunately matters a lot, even if he’s not actually authentic in any way. Donald Trump has established baselines for his behavior. We know that this is how he behaves. We don’t care.
We’re entering new and dangerous territory in our political and social thought and discourse. We’re becoming a society without any sense of self-discipline, decency, or shame. There are fewer and fewer boundaries that define what is acceptable and what is not, and that’s a dangerous, dangerous thing.
I blame wearing hats inside. And thinking that a t-shirt is appropriate to wear anywhere - playing baseball, going to work or church, a wedding - a funeral.
This is what bothers me about 538’s methods, even the less volatile polls plus. It is known that polls right now are volatile with bounces to the conventions and news cycles and that polls during the conventions and until a week or more after are less predictive than what the longer term average of the polls were going into the convention season. By two weeks after the conventions the polls typically settle into a more steady state and have much greater predictive value, but the polls in the conventions period should not be altering the odds very much. What are seeing is not a dramatic bounce but a return to baseline after convention volatility: aggregate average on their way to returning to Clinton +5 to 7.
Silver’s criticism of Wang actually more applies against himself: he is using polls during the conventions (even to some degree in the polling plus model) to make predictions with no regard to how poorly polls at that time have performed at predicting November results (significantly worse than the average of polls over a time period before the conventions).
Right now, someone in the Cialis marketing department is coming up with a new ad. It probably features Miss Universe.
The McCain incident didn’t bake into the numbers because it was early. People weren’t taking Trump too seriously, and only political nerds were really paying attention. Plus, it was easily written off by people who thought favorably of Trump as a way to make himself stand out from a large field.
Here is the thing to keep in mind - none of this is black and white (or blue and red). Most people don’t start paying much attention until now. Most people don’t care about the primaries. Most people aren’t sure where Canada is on a map :).
Gallup has been doing polling on the reaction to conventions for a long time. It’s still one poll from one organization, so it carries less weight than an average of polls.
The results are still startling. For the first time ever, people became less likely to vote for a candidate, Trump, based on the convention, and by a large margin, 51-36.
This is somewhat mitigated by looking at the history, which clearly shows the growing polarization of the voters by the single digit gains of recent years.
I find the negatives utterly astounding. I would have said beforehand that it couldn’t happen. It either means that the poll is seriously flawed or else the rollout of a Trump candidacy to the larger American public is.
Yeah I’ve got to say that despite the respect he’s due, that Now-cast of his is shameless click bait.
I find it useful to measure temperature and hype. i.e. if my Facebook feed lights up one direction or another, I can check that against the now-cast - because Facebook is of course, reflective of me.
But yeah, predicatively lousy - but its also good to show people “see, what you are feeling right now might be temporary…”
Years ago, I had a good (I thought) friend who would do or say some not-good things, to me and to others. I always rationalized these things away and remained faithful to the friendship. Then one day, in a phone conversation, she casually dropped one more betrayal on me, and from one word to the next, the friendship died. It was as if a kaleidoscope had turned and everything fell into a new pattern – an ugly, repellent one that cast a harsh light on the past. I simply could not deny the reality of who and what she was any longer.
Call it the scales falling from my eyes, call it the tipping point, call it the straw that broke the camel’s back – for sane people, there comes a time when an ugly reality can no longer be denied. I suspect the Khan attacks will prove to be this for some people; that more such episodes will turn other people off; that while Trump will never lose his hard core of support, he will lose more and more prospective voters, in an election where he’s got to expand well beyond his base to win.
I suspect, Exapno Mapcase, that those astounding poll results could be coming from that sort of experience for a lot of voters.
So, nobody’s going to explain to me how this “Nowcast” stuff works? Well, fine, then, you ever have a desperate need for a tie-dye shirt, don’t come crying to me!