538 now has FL and NC tipped pink, with strong Trump momentum. Assume he wins there.
They also have him a combed-forward hair ahead in NV. But we’re told that Clinton has a considerable early voting advantage. Assume Clinton wins there.
What about NH? The only quality post-Comey poll – MassINC – has Trump +1. Again, strong Trump momentum. Assume Trump wins there.
The EV count stands Clinton 266, Trump 263.
Then it all comes down to CO.
The only post-Comey poll by even a B grade pollster is Emerson College, 10/28-10/31, 750 LV.
Clinton +3
Margin of error: 3.5
EDIT:
Another way to look at it, assuming that Trump wins FL and NC… [Since I originally posted this, NC has tilted back to Clinton at 50.1%]
Of NH, CO, and NV, Clinton must win 2. Even if she wins CO, without the other two, the College is tied at 269 and (we may safely assume) Trump wins.
As I expected way back when, Ayotte has moved into a consistent lead in the NH Senate race. Not that Hassan isn’t well-liked enough or that Ayotte leads in favorability ratings. Hassan would have been a shoo-in for a third two-year term as governor, but switching to a federal office after less than eight years in a statewide office is a high hurdle for a politician unless that federal office is an open seat or the incumbent in the federal office has demonstrated that she/he is a nincumpoop. (ref the 1958 CA gubernatorial and senate races. The popular Gov Knight and Sen Knowland tried to switch offices and both went down to defeat.)
In DC, Ayotte has demonstrated that she’s out of step with the majority in NH and therefore, was deemed vulnerable and Hassan was popular enough that objectively she could be expected to defeat Ayotte. Hassan made a huge error in this election cycle. With the NH polls from 9/15 on showing a strong win for Sanders, she pulled out all the stops for HRC. Very dumb move. Then there’s the open gubernatorial race. Looks as if the Dem nominee is going down to one of Sununu’s kids, but it appears to be less of a preference for the Republican and more a comparison between the two candidates.
That current political landscape would bode well for a decent Republican nominee against HRC. But Trump doesn’t fit that bill. Expect it to swing back in favor of HRC by next Tuesday, but while still possible, probably not enough to pull Hassan with her.
I’m sticking with HRC in CO and NH. Unknown is what is going on the NM, but as of 10/2 Johnson was fading; so, HRC should carry it. If HRC carries NV or VT, it’s a wrap.
I posted some time back that I hoped it would NOT come down to Colorado. I know that state somewhat…
It won’t. It’s practically unheard of that a candidate that has consistently led in a state for months and months in polling loses it at 11:55 pm. But she’s not helping herself by lashing out at those she wants others to see as her enemies. Too defensive and strident which flows right into the public perception of her weaknesses.
Did someone give Kaine some politically savvy advice – Politico 11/3/16 – Kaine: Comey is not trying to influence the election
Dial it way back to get less of it as leading media stories.
OTOH – the optics of Huma headlining a glitzy fundraiser in DC are not helpful to the Clinton campaign. More reinforcement of the campaign negatives. Tin ears.
Don’t forget that ME splits its electoral vote. If Trump manages to pick up NH, it would not be a huge surprise that he also gets ME CD2, so that it wouldn’t even have to go to the House. The few polls for CD2 have shown it to be competitive.
Yeah, it still comes down to CO if Trump manages to pick up FL/OH/NC etc.
In my conservative EC projection, I have ME-CD2 in Trump’s column. With NH and CO in HRC’s column and NV in Trump’s, that puts HRC at 272. Switch CO and NV (although I still think the opposite CO-HRC and NV-DJT or both in HRC’s column is more likely) then it would be 269 to 269 unless DJT doesn’t carry ME-CD2. Hard to believe that HRC’s team hasn’t seen and therefore, hasn’t been working ME-CD2 for some time.
First, stop looking at 538. I’ll explain in a bit, but Silver’s methodology has always been screwy and this year he’s being gamed.
Second, Colorado is NOT going Trump. It won’t be close. 4 years ago the polls said 1-2% for Obama – it went 4.7%. There is a Latino firewall here and if you work in the field you see the passion. The passion wasn’t there in the midterms, which is how fake Moderate Gardner won the senate against lackluster Udall, but it’s back again now in bigger numbers and with more emotion.
Latino votes have been underestimated in forecasts every year since 2008 and they are doing it again. Only the high quality pollsters know how to poll the Latino community and to estimate accurately their turnout.
And, to add to that, last night a huge number of Latino voters showed up in Nevada on the last day of early voting. 82.5% of them were flagged as low propensity voters, thus were screened out by the likely voter screens. We’re going to see the same but not to as extreme a degree in NM, CO, FL, TX, AZ, CA, etc.
Back to 538. Most polling aggregators try to keep things simple. Complex models tend to introduce problems and inaccuracies. Nate Silver is the exception. His background is baseball stats where some of that complexity could be justified. Baseball is a sport where a poor at bat can result in a big hit, such as a blooper or lucky bounce, while a great at bat can result in an out. This randomness factor caused Silver to look for every data point he could to improve his predictions. And in 2008 he hit the jackpot when he stood out among baseball prognosticators and accurately picked the Tampa Bay Devil Rays, who’d never had a winning season and were buried in the same division as Boston and the Yankees, for first place. Suddenly he got a lot of new attention.
At the same time he was playing around with his new 538 web site so the attention spilled over. He’s nothing if not thorough and his web site was more detailed than any competing web site plus he offered a daily commentary that was very readable. While being openly pro-Obama, he also gave an objective analysis and a lot of GOPers flocked to his site as well. In the end he made a name for himself. The NYT picked up up, and later he bolted for ESPN and even bigger bucks.
Of course, 2008 was an easy election to pick, so inaccuracy wasn’t an issue. In 2012 it was still an easy election to pick, and he was fronting for the NYT and appearing on all kinds of TV shows. Meanwhile, more accurate prognosticators were getting little attention.
But even then the problems were evident. In 2008 he had Obama at only 75% way too late in the election cycle. Eventually he had him at 95% but even that was wrong. With a 7 point national polling lead McCain’s chances were a fraction of a percent. Sam Wang and others pointed this out. Silver responded by saying 95% is 1-in-20 and if you look back at the past 20 presidential election there was indeed one election that had that big of a surprise. I forget whether he was talking about 1936 or 1948, but for either the comparison was absurd – polling was so minimal back.
Sam Wang, who has a website that could have been deployed in 1996, and is anything but user friendly, has a simple state-only poll statistical model since national polls introduce a lot noise. He includes each poll only once and doesn’t weight for recentcy. He throws out crap polls. And given probabilities of swings in voting preferences and behaviors he is able to much more accurately predict results. His model does critically rely on LOTS of state polls. His model is not reliable in mid-terms – and this has been shown – due to far fewer polls.
Nate Silver models uncertainty by including every factor he can, including polling behaviors from 50 years ago. In response to criticisms of him including things like economic reports in his model he now also offers a polls-only mode. But even so there are fatal errors in that model.
For one, every poll is weighted for recentcy. Thus, a crap GOP push poll in a state that just came out can dramatically swing that state in his model. Silver tries to account for that by weighting the value of the polls with grades (like “C+”) but that’s pretty weak. And in 2016 a flood of crap low-sample size, pro-GOP polls have been released overwhelming his model.
Another problem is bleeding. Yesterday a single crap poll for Utah came out bumping up Trump’s lead over McMullin from 5 to 7 points. Nationally that bumped his probability of winning in Silver’s model by 1.2%. That’s nuts. But his model assumes that a poll of one state has implications for the surrounding states, so adjusts the chances for the candidate in those states as well. Thus, the Utah poll – in Silvers’ model – improved Trump’s chances in Colorado, AZ, NM, and NV and other states. Now, there is a situation where this kind of method can make sense, and that is early when there are few state polls. But at this point Utah is 100% GOP and 90% Trump – no new poll in Utah should shift the overall likelihood for Trump, and in most models it doesn’t. A week ago we saw something similar happen when a poll showing Trump up 14% in Oklahoma gave him a 2% point improvement in likelihood nationally. Oklahoma. Polls from states like Oklahoma, Wyoming, and Rhode Island should have zero impact on the national model since those states are clearly decided.
Follow mystery pollsters’ twitter feed. Check out Sam Wang or the huffington pollster. But don’t sweat over 538.
Well, except nationally most (all?) states move in the same direction from the prior election. If the national trend shifts red, red margins increase in red states and blue margins decrease in blue states and vice-versa. Tend to doubt that neighbor states bleed more than the national trend, and who is to say which state is the bleeder and which is receiver?