Protected: Two ways to estimate primary outcomes without polls (transcript)
This content is password protected. To view it please enter your password below: Password:...
Iowa discussion thread
Follow the returns at HuffPost, the New York Times, and Red Racing Horses.
“Will Marco Rubio come in a strong third, or a weak third?” That was an actual pundit question. Data-ish punditry welcome below.
>>>
12:09am, Tim in CA: “Iowa and New Hampshire may represent a high water mark for Sanders. And in the case of Iowa, that high water mark is a tie. That result doesn’t look good enough for him in the long run. New Hampshire is even more favorable to him, so he will probably win it. But then comes SC and NV, and after that the Super Tuesday states. The tide will start turning to Hillary. That is the big unreported story, right?”
12:07am: The Duke asks, “Will we get a ‘I was overconfident in the polls and my model was wrong’ post from you?” My short answer: no way!
Firstly, let me get out of the way that this misusage of “model” really grates on me. When I report simple poll medians, there’s no model to speak of. This colloquial terminology has always bothered me.
But to address what the Duke meant to ask: my estimate of probabilities was simple, and seems OK. I suggested that the Iowa data indicated a 70% probability of a Trump win, 60% for Clinton. These probabilities were based on the idea that poll medians could be off by an average of 5 percentage points. For now it appears that the polling error was about 2 percentage points for Trump, the same for Clinton. These are actually pretty small errors. The big story tonight was a net move from Trump to Rubio, as I pointed out. Also, think about it…those probabilities would only be both right 70%*60%=42% of the time…in other words, a 58% probability that either Trump or Clinton would lose. That’s the way it goes when outcomes are uncertain.
I will stick with one statement: Sanders needed a clear win tonight, and he didn’t get it. For the moment, I see tough sledding for him in the weeks ahead.
Here is one error I may have made: I didn’t think Iowa would narrow the field. But Rubio’s strong performance might just do that. We will see.
11:25pm: NYT now has delegate projections. On GOP side (97% reporting): 8 Cruz, 7 Trump, 6 Rubio, 2 Carson, 1 Paul. Democrats (93% reporting): 21 Clinton, 21 Sanders.
11:17pm: I agree with Pechmerle – the “strong third” thing ended up not being that silly. If it narrows the GOP race to three candidates soon (Trump, Cruz, Rubio), that’s the most likely route to a not-Trump outcome.
10:45pm: The Des Moines Register site was terrible.
10:25pm: Cruz, then Trump, then Rubio. With a chance that Trump will fall to third. That is about a 5% deviation from polls. Wow.
10:20pm, Froggy: “A reminder that Ann Selzer had it Trump 28, Cruz 23, Rubio 15, Carson 10. Oops!” but “Not a total loss for her – Carson does have 9.3%.” ow.
Topics:
How have Iowa entrance polls been done in the past?
MSNBC reporting 43% Repubs 1st time voters. Beats the 38% of 2012 which was all time high of Republican Iowa voters
Entrance poll so should be taken with many grains of salt: Dems under 30 caucusing are 91% for Sanders, 8% for Clinton, 1% O’Malley.
Are Iowa precincts approx. the same population size?
Even if they were (at least regionally they could be), precincts are the same for both parties. Therefore the number of Democrats (or Republicans) per precinct will vary. For example, college-town precincts will have more Democrats and fewer Republicans.
I was just wondering if 20% of precincts reporting implied about 20% of the population. I understand the clumpiness issue.
Apparently the population size is fairly variable. See item 15.
Another bad night for the pollsters? Larger turnout but Cruz is outperforming. Wasn’t large turnout supposed to be good for Trump?
I think a large turnout may have been necessary (but not sufficient) for Trump. That is, if much of his support came from first-time caucusgoers, then lack of high turnout would be a bad sign, but the converse is not necessarily true. Rubio seems to be outperforming as well. Maybe the “desperate” “establishment” conjured the high turnout?
Generally I am surprised by the results. I do wonder whether there’s something of a reverse shy Trump effect going on. Even with the secret ballot, the caucus is a pretty social environment — I’m sure people are busy asking each other who they’re voting for, and I assume people are reluctant to say that they’re voting for Trump.
Actually the New York Times here (http://www.nytimes.com/2016/02/02/us/politics/iowa-caucuses-faq-our-man.html) says that at some Republican caucus sites people express their preferences by a “show of hands,” i .e. a public demonstration of one’s opinion. That might also hurt Trump.
Isn’t the final vote done in a booth on the GOP side?
I can’t find an authoritative source describing the Republican caucus procedure. NPR says in past years caucuses used either a show of hands or a paper ballet, depending on the size of the caucus, but now all caucuses use an “informal paper ballot” (http://www.npr.org/2016/01/30/464960979/how-do-the-iowa-caucuses-work). The NYT on the other hand still seems to think that a show of hands is used in some caucuses. This may all be much more informal and a little arbitrary at the actual precinct level.
I don’t see anything about booths — could it be that people fill out the ballot in their seats?
Perhaps I am wrong about that. There does seem to be a show-of-hands step.
“Shy Trump voter” — now that seems like a pundit-ish fictional construct attempting to fit the results. I’ve not seen a single man-in-the-street interview that suggested anything shy about Trump supporters.
Far more likely: Cruz’s ground game was, as suspected by some, better than Trump’s.
In my defense the Washington Post ran a piece earlier today on how Trump polls better in live phone polls than internet polls: https://www.washingtonpost.com/news/the-fix/wp/2016/02/01/how-the-secret-ballot-in-iowa-could-help-donald-trump/. So it’s not as though I pulled the shy Trump voter out of a hat.
That said, this is certainly some ex post facto justification — I wasn’t expecting to make this argument 12 hours ago.
A reminder that Ann Selzer had it Trump 28, Cruz 23, Rubio 15, Carson 10. Oops!
Not a total loss for her – Carson does have 9.3%.
Eh–she was off by 4% for Trump, 5% for Cruz, 8% for Rubio and 1% for Carson. The average miss for a primary poll is 8%, so this poll, while not one of Selzer’s better efforts, is certainly, at a minimum, par for the historical course, if not better.
Assuming these totals hold, what’s the delegate breakdown for Cruz/Trump/Rubio?
It’s strictly proportional.
Cruz between 56 and 111 votes from ekeing out an eighth delegate.
“Will Marco Rubio come in a strong third, or a weak third?”
I don’t see anything silly or “unquantitative” about that point. It was known to be Rubio’s goal to accomplish at least the strong third. Now that he is only about 2,500 behind Trump, vs. nearly 25,000 ahead of Carson in fourth, surely that qualifies as a strong third in anybody’s book — in a multi-race event. Rubio’s campaign — not just as spin — will take some comfort from tonight’s developments.
The betting markets see tonight’s results as a huge victory for Rubio. Over at Predictwise (http://predictwise.com/politics/2016-president-republican-nomination ), the chances of winning the GOP nomination early this evening were Trump 52%, Rubio 32%, Cruz 10%. Now it’s at Rubio 55%, Trump 29%, Cruz 9%.
Trump did not outperform. The expectation was that he could bring out voters who had not participated previously. He will now be under the gun to prove that his voters will come out for him. Everyone will depreciate his poll numbers. Cruz had this great ground game and was in one of his best states. It was Rubio who really outperformed and joined Cruz and Trump.
Sam,
Will we get a “I was overconfident in the polls and my model was wrong” post from you?
No.
The Times seems to have left out 6 delegates? The results sites were terrible. You can represent
the full results with a few hundred bytes of text, but the web designers insist on larding the pages with 10 megabytes of widgets and angry fruit salad graphics.
Overall though, don’t we need to come back to Sam’s assertion that Bernie needed to do better than this? According to Nate Silver’s calculations, Iowa ranks third out of 50 for having the highest percentage of white liberals voting in the democratic primary (only NH and VT have higher proportions). As Sam has said, Iowa and New Hampshire may represent a high water mark for Sanders. And in the case of Iowa, that high water mark is a tie. That result doesn’t look good enough for him in the long run. New Hampshire is even more favorable to him, so he will probably win it. But then comes SC and NV, and after that the Super Tuesday states. The tide will start turning to Hillary. That is the big unreported story, right?
This is almost certainly the correct takeaway from Iowa for the Dems. Which is why it was so infuriating to hear the talking heads on CNN and MSNBC spinning it into variations of “Sanders will be challenging Clinton for months and months to come.”