NPR just put up an article on this, with a graph - much higher than previous years during the primary season, but at the moment looks not too different than 2008.
No idea and not sure how to judge it even.
FWIW here’s where Wang was in September 2012 (near end of):
And popular vote margin of about 4. The final was 332 EVs and a margin of 3. Pretty good for a September call.
Trying to find 538’s same point in September call I can find this archived page. Which seems to have him calling for an over 5 popular vote margin with an over 80% probability of Obama winning. I can’t find his EV prediction for then but I can find that his odds of an Obama win dropped dramatically for a bit in the beginning of October. Wang was not so volatile.
I see no evidence that Silver has a better record in September than does Wang.
Moreover Wang’s model is simpler (and transparent). When there is a choice between two models and neither one demonstrates clear superiority to the other the simpler model is generally to be preferred.
Being less scientifical about it, Silver today summarizes his sense as being Clinton +4 but
Whereas my read Wang is not far off from that Clinton +4 but that his version would be find both a narrow Trump win and a 10-point Clinton win to be very surprising.
But in Sep. 2014 Wang’s system failed miserably. Nate gave the GOP a more than even chance to take the Senate despite being behind in the individual Senate polls. Wang just read the polls and assumed they wouldn’t move except randomly. Instead, they all moved in the same direction, and it’s not as if it wasn’t predicted, and not just by Silver. Incumbents polling below 45% in September are usually in big trouble come November. Leads up until October are often nothing more than lack of name recognition for the opponent. Wang was treating 40-35 leads in September as unlikely to change.
That’s what makes Silver’s model a cut above the rest, he uses historical trends to predict poll movement and the behavior of undecideds. Things like the economy and Presidential approval ratings can move undecideds towards one or the other candidate.
Silver did better in the Senate call in 2014 and in 2012 Wang did better in Senate races.
No, the current divergence is not that Silver is using historical trends and Wang does not. Wang is using historical trends, explicitly and very transparently. Silver is doing that less so, and is promoting fundamentals more and stating that there is less from history to call upon with confidence, therefore relatively huge margins around his call.
Okay. He cited 18-20% undecided in his recent chat with other 538 writers as a reason for such uncertainty. In a race with even an 8 point margin for Clinton that’s no sure thing. Thus my eyebrow raising at Wang being so certain. I mean, his odds basically say, “Clinton wins. Full stop”
And in early September '08 polling had 14% undecided in some polls. Sure not all.
Actually “that’s either undecided or voting for one of the (largely anonymous) third-party candidates.”
So what will they do?
Well right now we know that if they had to decide between the two that in most polls they’d either give the same margin or a bit more to Clinton.
Maybe they’ll, as current numbers suggest, split slightly more to Clinton or evenly. Or stay home. There is certainly no historic basis for believing that undecided and third party preferring voters polled will split overwhelmingly to vote one way or the other, and no current data that suggests they will.
Probably it depends on who the undecided and third party voters are. If they do decide to vote for one who would it end up being?
Right now it seems that they aremostly those who would otherwise be predisposed to vote for Clinton.
Is there any rational reason to believe that this group has any realistic possibility of ending up coming out to vote splitting for Trump overwhelmingly?
If they come out to vote then they will either vote other than Trump or Clinton as a protest, especially if they think that the race is not close in their state (your stated position for example, addie), or if they “come home” it is per their demographic coming home to vote Democratic and against Trump. But more likely having something else to do that day.
There’s an interesting post on 538 today that touches on this. Nate Silver took a look at recent polls in his “states to watch” and ranked them. Michigan was the tipping point, if Trump could win GA, IA, AZ, OH, NC, and FL. He’s trailing by 1-2 points in NC and FL in Silver’s “simple average” of recent polls. That illustrates one plausible path to victory for Trump, but he’s got some catchup to do in NC, FL, and MI for that to happen.
2008 was a year of undecideds before the race was decided by a massive financial crisis with the incumbent party being blamed and the incumbent party’s candidate frankly not appearing up to the task (Palin’s selection probably cemented McCain’s collapse though he was in trouble anyway).
Any crisis that occurs now would likely be blamed on Obama and work against Hillary. Without a crisis, this race will be more about the candidates themselves. Both are unpopular, which is why Trump will probably keep this close until the very end.
It’s … interesting … that Silver writes that post without bothering to mention that his own polls-only places MI as a more improbable win for Trump than PA is (23% chance in PA and 20.5 in MI).
Bottom-line of that post though remains: Trump winning GA, IA, and AZ in a close election is no heavy lift. Winning both OH and FL at the same time also, a bit tougher. But assume he does. He still needs to pick up another. There are few paths for him and many for Clinton.
It is not impossible.
ISTM the path to victory for Trump runs directly through getting the enough of the undecideds to dislike Hillary enough to choose the “third party or stay home” option. Call it voter suppression by persuasion.
While simultaneously rousing his band of rabble that usually don’t vote at all to bother showing up this time while filling the social media space with noise in the months prior. I’m *not *suggesting that all Trump supporters are rabble. Merely that in contrast to, say Romney or McCain, in addition to conventional R voters there’s this new demographic that’s being energized and enfranchised for the first time. Call it the opposite of voter suppression (voter amplification?) by persuasion.
My concern for both Silver and Wang is that while “rouse our team as much as we can without excessively rousing the opposition’s team” is a tried and true political tactic with lots of relevant history, this time is truly different. Quantity has a quality all its own and we are in uncharted waters (at least over the last century-ish) as to the quantity of disaffected / swing / newbie voters. We’re certainly in uncharted waters as to the power of social media and biased news source echo chambers to amplify a signal.
IMO Wang is being silly to project such a small dispersion = tall skinny bell curve on his range of outcomes. Silver’s much flatter, wider curve with larger tails is a more accurate description of reality itself. Albeit not of the polling data viewed in isolation.
Well from a practical matter it makes no difference whether the undecideds stay home (which I think most will do) or if they all come out but break evenly does it?
Silver justifies his uncertainty to a large degree on the possibility that they will break Trumpward. Given that he is even more unpopular with this group than is Clinton and that these are disproportionately younger voters, if they do vote for one of the two of them the odds are it won’t be for Trump.
So let’s assume, for the discussion, that they don’t break on way or the other - they stay home, or vote third party.
Will Trump inspire enough increased turnout and margin of White non-college-educated voters (a group that historically has poor turnout) to make up potential increased turnout and margins in some minority groups, the loss of GOP margin in the White college-educated demographic, and the impact of superior Democratic GOTV infrastructure on the ground?
That is the question and is the limitation of polling right now. How likely is each voter and how much do you factor in GOTV? Polling may be off some from election day. Which way and by how much?
Based on where Trump is right now, it wouldn’t be enough for him to convince the undecided to stay home. That might help prevent things from getting worse for him, but he’s still behind in decided voters, too. He needs to convince currently-decided Clinton voters to stay home, or to convince some of the current undecideds to break for him, or he needs to convince a lot more of his stay-at-home supporters to come vote than the models expect.
I’m going to contradict myself some … polling can still help out some.
Most recent numbers I can find are from this Bloomberg poll 2 days ago.
So let’s go to that 538 app and let’s make some very generous to Trump assumptions:
-
He actually does significantly better with non-college educated Whites than that, heck, instead of 4 points under Romney’s move him to 4 points over Romney’s.
-
Those voters turnout much better than they have before: instead of the 57% turnout they had for Romney 67% of 'em come out to vote! He gets them to the polls like they’ve never gone before. That’s a lot more of his stay at home supporters coming out. The power of social media to amplify the signal.
-
And no other demographic changes at all. The current flip in White college-educated from the 12 point advantage Romney had to even or sometimes even slight advantage Clinton does not occur, no increased margin or turnout among any minority demographics. Trump does as well as Romney did among them all. Improbable but play along.
Trump still loses. A close one to be sure, a near tie in the popular vote (0.1% advantage Clinton) and 279 to 259 in the EV. But even under those absurdly generous assumptions he loses.
Here’s another article, a bit older, from Brookings.
The big shift after the conventions … mainly increased Clinton share of college educated women.
The fun comes with the simulations. They simulated a 2016 election using 2012 voter turnout rates and pre and post convention preferences by demographic group. Pre-conventions Clinton would have been winning by 4.8 million votes; after-conventions by 10.2 million.
But they said, what if White non-college educated males come out much much higher than ever before? How about as high as White college educated men - going from 2012’s 55% all the way to 79%?
Clinton post-conventions wins by 5.3 million.
Even if you moved White college-educated women all the way back down to the pre-convention level, and kept that high hypothetical turnout of White noncollege men … Trump would, in those simulations, lose.
Nothing written in stone but his task includes more than keeping undecideds from breaking to Clinton and getting his vote out. He has to also either win over the lion’s share of undecideds and/or flip some sizable number of White college-educated voters back to Romney levels and beyond.
I wonder how surprising it is for you to write something like that. 6 months ago, or 3 months ago, did you think it’d be this close in September? I think most people did not.
I cannot speak for anyone else but I have been very consistent with my assessment. He has IMHO had a longshot chance all along and still has that same longshot chance.
You can get a sense of what people were thinking 4 months ago here.
In May roughly half felt the final would be more than Clinton +6 to roughly half less than Clinton +6. I took the under position and I do not believe that the race has fundamentally changed since then. To me the reasonable over/under line remains Clinton +5 and the chance of a Trump win remains a bit less than Silver’s degree of uncertainty would have it be.
I am also expecting that election day will have the polls having underestimated Clinton’s win by a couple of points.
You’re probably right since Stein is taking 3% right now and will probably be closer to 1%. So you can probably give Clinton a couple of points from her on election day, assuming a high percentage of those aren’t “never Clinton” voters.