Sorry, you’re right - though source aside, I found it a good summary of links to Nate’s various statements of caution on his own work.
I predicted that they would be virtually tied with some credible polls showing Trump in the lead. I also said (not sure if it was in the same post) that the maximum impact of that story - provided there was no additional follow-up - would be felt by week’s end. I was off but not by much. An earlier ABC/Post tracking poll indicated a brief Trump lead (though within the MOE). I admit to calling it wrong on the national RCP average, but not by much.
On the flip side, Nate Silver’s number crunching seems to indicate that the race is trending toward a virtual tie in terms of the electoral college. Like you, though, I’m struggling to figure that one out myself. I agree that the race is tightening but I think Clinton but I think the she is beginning raise the nose of the plane. I’m not just looking at polls either – I’m also considering things like education levels, diversity, and past election results. If someone told me I had to make my picks now I’d actually probably give Ohio to Trump, but Florida would probably go to Clinton in my view. I’d probably also give her NC at this point though my gut is telling me ‘no’ based on the state’s voting history. Of course that could change. Final picks Monday night.
I admit to being off, but I wasn’t off by much at all, though. In fact a “virtual” tie is essentially anything within what could be expected to be within polling error range. There’s actually a pretty good article about that on 538’s site today.
RCP’s averages put Clinton ahead by +2.4 (today). Polling was off by 2.7 in 2012 as a point of reference. Two current polls show a tie. A ABC/Post tracking poll actually did show Trump with a lead earlier in the week.
I’m wondering if this is why Silver is increasingly bullish on Trump in North Carolina because the polls seem to be showing some good news for Clinton the last few days.
It’s kinda weak tea, imho. The whole two statements from Nate they bring up are about possible polling issues not about his own methods. Lag time for major events and likely voter screens. The rest of the article is just listing things they think Silver is not accounting for that all just happen to be saying Trump’s chances are overestimated.
Now don’t get me wrong, I think Nate is overestimating Trump’s chances too and a couple of those article’s points may be valid. But that is a lying click bait title and it’s just like that dude’s opinion, man.
But all those points really tied the article together.
Eta: OK, three statements. They note Silver admitted there hasn’t been much movement in voter preferences. Still nothing about voicing cautions about his model.
Who me? When I made this topic it was in July when the results were showing him up. That was then.
Personally, for both candidates, I wish them well. I really don’t care, I’m just posting the data up so I can see what people think of the results.
Ha ha, no not you. I was talking about the Raw Story article that squidfood linked.
Phew, you made this law-abiding anarchist worried for a moment.
Someone mentioned about the Wang model, anyone have a link for the model? I’m not googling it myself (you can understand why).
Trump could win, but looking at poll data since 2004, it would be an extraordinary upset if it actually happened. Assuming historical trends remain true this year, the question in my mind becomes more about how convincing her victory is and whether the election results embolden republican obstructionism at the federal level or whether they perhaps back off the threats and continue to be functionally dysfunctional as they’ve been with Obama.
The problem is that his model is the only one doing it, while every other model has much higher numbers.
You can see a comparison at the New York Times prediction page. The average is 87%, or 89.5% if you take out the outliers on both sides.
Silver’s is not the only data-based prediction model out there. And he’s been awfully “pundity” this election, given his articles.
I still can’t get past him acting like New Hampshire had flipped when his models said nothing of the sort. He knows better than to bring up individual polls.
There was a short snark war earlier today on Twitter between Nate Silver and Sam Wang, something like this:
Nate: “Princeton Baseball Consortium probably had the Indians 99% favorites to win.”
Sam: “Well, you’re the expert on sports statistics.”
It would not be an extraordinary upset due to the high number of undecideds and third party voters. 35% chance of a Trump win sounds about right. Clinton is polling only 45% in the 4-way, 46.4% in the 2-way. Undecideds and Johnson voters don’t have to break THAT hard for Clinton to get pulled under water. Plus as 538 has pointed out, the map is actually problematic for Clinton, not Trump. She could very well win the popular vote and lose the election.
The biggest threat to Clinton that I’ve seen the last few weeks is Johnson voters, – more likely to be fiscally conservative but socially open-minded – decide to become #neverHillary voters. Indeed it looks like some of Johnson’s voters have come home for the hold-the-nose-and-support-my-ideology vote. But not all of them are going to abandon Johnson. I suspect most of the movement has occurred. The same forces that created Trump have also created stronger protest candidates this year.
It’s certainly within the realm of possibility that Trump could win – I won’t completely write him off like Sam Wang. But when you look past the polling decline for Clinton and her current numbers, a couple of things stand out: Trump still hasn’t taken a lead, and by election day anywhere from 35-45 percent of the electorate will have voted. What’s more, it seems like the democrats have gotten off to a good start in the early results. That means Trump probably needs to have a decisive turnout on election day. Now in some states like PA, NH, and MI, this in and of itself isn’t a problem for Trump as there’s no early voting. However, the polls don’t provide any real evidence that he’s winning. He’s been consistently losing in the polls in each of these states - yes NH has a poll or two showing some strength for Trump but the general data and history still lean toward a Clinton win.
But here’s another important factor, and it might be the most critical one: Clinton’s ground game and organization. Clinton has a sophisticated information gathering apparatus, which might not matter so much if she were really losing the race. But all signs are that she’s protecting a lead - a slim lead and a gap that has been tightened, but a lead nonetheless. Obviously that won’t win her anything in deeply red states, and it probably won’t matter in a state like Ohio, where Trump has really tapped into the white working class without much push back from traditionally strong democratic constituencies (On a sidebar note this year might be remembered as the year that Ohio officially became a true ‘red’ state permanently). But in other states, like Florida, Nevada, Pennsylvania, North Carolina, and New Hampshire, Clinton’s machine can be the difference maker. I suspect it probably will be.
Though I concede this has been a wacky election and anything’s possible.
Some of that should already be captured in polling though. I can’t think of an election where the polls were off because one candidate’s ground game was better than the other’s. Of course, a good ground game can be decisive in a close race, but we should already be seeing that in the polling and we probably already are seeing it in the polling, thus her clinging to a lead.
Undecideds this point Obama v Romney: 4.5
Undecideds today: 5.5
It’s more but not dramatically. The available evidence is not consistent with any realistic chance they suddenly break much more one way or the other, if they come out and vote at all. Of note in 2012 8% were “late deciders” deciding in the last week. Yes, hope sprung eternal then too for Team Romney that they’d break for him.
Thing is that in general the undecideds are predictable. Half stay home pretty reliably, for example.
Yes there are also a total 6.8% selecting a different choice.
What do we know about how those selecting a different choice would choose if they had only the two to select? Well some polls offer the question both ways and the results have generally been either the same difference between Trump and Clinton or within 1 sometimes one way and sometimes the other. Those who claim they prefer Stein overwhelmingly would go Clinton and the less tiny numbers who say Johnson are more evenly split. If they had to choose between Trump or Clinton.
And of course many will actually make the choice they currently state and not break at all.
Obama’s ground game was better the Romney’s, and his data collection and analysis were definitely better – like way. Trump is casting his line out into the water somewhat randomly – definitely better with Conway on the team but still somewhat winging it and catching up. Clinton is going after voters with surgical precision.
Good post.
Also, I think it was Sam Wang who said that the late deciders who actually do vote peel off evenly for both candidates for the most part.