January 16, 2016
Sunday night on NBC (9pm Eastern), Democrats meet for their fourth and most important debate. Round One was Waiting For Joe Biden. Round Two was on a Saturday night in the wake of the Paris attacks. I can’t even remember Round Three, also on a Saturday night when I should have found alternative entertainment.
Watching that one was an act of civic obligation and bad judgment. I’m legitimately curious about this one. Plus it’s on after football on Sunday night, so no real activity conflict. This one will have a bigger audience. The race is getting closer and a lot of people will be at home.
In honor of actually looking forward to the event, we’re doing multiple prep pieces. Tomorrow morning, I’ll cover what questions might come up, what advice I’d give the candidates, etc. Today we look at where they currently stand.
Who is actually leading in Iowa? Are Bernie or Hillary positioned better? Is she trying to protect a slim lead, or is it even worse? This matters for at least a couple reasons. It’s another excuse to look at polls and data, and any of the strategic decisions for the debate are based on present position and short-term outlook.
Bernie MUST win Iowa, so Hillary realizes the value of cutting this off now. It’s worth digging herself another hole for the general election. Meanwhile, he realizes this is for real. No sense leaving any cards unplayed.
FiveThirtyEight has an excellent forecast model. It’s a modified version of what Nate Silver developed for the 2012 primaries and it attempts to balance the various polls by weighing them instead of just averaging as the Real Clear Politics or Huffington Post averages do.
For your prognosticating enjoyment, there are two settings, one that only takes polling into account, and another which adds in things like endorsements, which have some predictive value.
As of the moment this is being written (the link above will take you to whatever the latest update is), Hillary has 82% odds of winning Iowa with everything being taken into account, and 66% just looking at the polls.
I disagree. You may wonder why I built up the forecast, only to argue with it, but it’s my nature. Plus, it’s the best one out there and they’ve done way more math than I have. But they needed to construct something that will work over multiple contests in multiple places over multiple months.
In trying to achieve that balance, there’s a chance they wind up missing a specific variable that doesn’t apply everywhere. That’s what I’m for, to catch the goofy outlier. In this case, it’s undecided voters.
Polls often do a poor job of showing them. They usually attempt to get voters to pick someone, asking who the respondent would vote for if the election was today, rather than trying to get them to speculate on what they might do later.
A separate question will ask if there is a chance they’ll change their mind. Looking at extra data will tell you x percent of voters might shift. You can find out that one candidate has a higher percentage of voters that are locked it. But the gradations are missed.
Let’s say the following classifications exist:
Completely locked down for a candidate
Strongly favors a candidate
Leaning towards a candidate
Slightly favors a candidate
Most polls will wind up with all but the completely undecided in one column or another. Looking at the recent Iowa polls, anywhere between 86% and 100% of voters picked a candidate to support for now.
Those are entirely different things. There is no way 100% of poll respondents are even halfway sure who they’ll vote for. What happened is the pollster made them choose someone. Completely undecided voters are included.
Hold that thought for a moment.
One of the best things about FiveThirtyEight’s model is the adjustment for house effects. If a pollster has consistently shown Bernie doing better than the median result, and he gets a good result in their new survey, it’s less meaningful than if it was from one that leaned towards Hillary. Makes sense, right?
This is extra important in Iowa this year, as there are two clearly separate tracks. Some pollsters have regularly shown Hillary with large leads, while others have indicated there’s a real race. With much of Bernie’s shot based on turning out less frequent voters and Independents, you can argue for either model.
If you think he’s going to rally the troops, believe Quinnipiac, ARG, and CBS/YouGov. If you think only loyal Democrats who usually vote can be counted on when it’s dark and cold outside, agree with Gravis Marketing and Loras College.
The Iowa Poll, done for Bloomberg and the Des Moines Register, under the guidance of Ann Selzer, is the one most observers pay the most attention to. Her track record is excellent, and FiveThirtyEight gives them an A+ in their pollster ratings.
Until the just released poll, they’ve found themselves on Team Bernie, showing a competitive caucus for months (albeit with Hillary in the lead), when others were forecasting a wipeout. Because of that record, the FiveThirtyEight weighting adjustment counts a 42/40 Clinton edge as bad news.
While that’s still close, based on previous numbers, you would have expected them to come in with an edge for Sanders if he’s in fact still surging. The easiest reading of this is his momentum in Iowa has stopped or slowed.
His forward progress likely has. Trump isn’t talking much about Bill Clinton this week, busier with Ted Cruz being Made in Canada, and making sure American sailors aren’t humiliated by Iranians in the future.
Hillary has pushed aggressively on gun control, trying to get to Bernie’s left. In general she’s trying to tie herself to the still generally popular with Democrats president, leaving Bernie on the outside looking in.
I would believe she’s halted the slide. But there’s good news for Bernie in the poll results too. Selzer is much more conservative about getting voters to pick a candidate. Her poll is the one where only 86% have chosen. Of those, 4% are Martin O’Malley fans, but as I’ve mentioned before, in Iowa, they’ll need to pick someone else at the caucus, unless their particular location has at least 15% support for him.
So really, 18% are up for grabs. More voters than that could shift, but these are definitely available. Selzer’s poll was never part of the Bernie-leaning group. That’s what it looked like, but what she was doing was counting undecideds and slightly leaning Hillary voters as undecideds, when others had them with Hillary.
Now, those voters have temporarily abandoned her, but she never really had them in the first place. Bernie-favoring polls are listing some of those voters for him now, but they’re still undecided or just slightly leaning his way.
Selzer took a middle-of-the road approach on turnout of irregular or first-time caucus participants. Her results just looked like those who expected large turnout because she wasn’t counting mostly undecideds as Hillary picks.
If you don’t try to force voters to pick someone unless they really favor someone, and you take a mid-range view of turnout, you get an effective tie. Hillary and Bernie are tied in Iowa right now. That’s my story and I’m sticking to it.
Here’s the catch:
Bernie has a choice to make. He can win if he has great turnout of people who definitely favor him, but may not vote, while grabbing about a third of the undecideds.
He can also win if turnout is better than usual, but nowhere near what Barack Obama managed in 2008, as long as he captures more than half of the undecideds.
To push turnout, he needs to do the Full Bern. To grab more than half the undecideds he needs to seem like a really plausible nominee, someone who could go up against Trump, Cruz, Rubio, whomever and have a definite shot of winning.
In the previous three debates, he’s reached 70-80% effectiveness on firing up the troops, and about 30-40% effectiveness on convincing wavering voters. Any post-debate polling movement has favored Hillary.
On Thursday, Donald Trump managed to hit both marks, both saying things to get Trumpists to caucus and appearing more palatable to waverers. If Bernie can pull of the same balance, he could actually have his turnout and undecideds too.
If Bernie hits neither, Hillary has an important edge coming out of the final debate before the caucus. Which way does he go, and what are his chances of executing? That’s our topic for tomorrow.