Protected: Two ways to estimate primary outcomes without polls (transcript)
This content is password protected. To view it please enter your password below: Password:...
What To Look For in the Iowa Caucuses
(updated since January 29th) After the mindnumbing levels of coverage over the last year…the first actual voting of the primary season finally starts tonight, with the Iowa caucuses. To answer the simple horserace question, Donald Trump seems positioned to come out on top on the Republican side, as is Hillary Clinton on the Democratic side. In both cases, polls narrowed over the weekend, adding a dose of uncertainty.
For the long term, a better question to ask is: can we get clues about who will eventually get the Republican and Democratic nominations? (For a hint, mouse over the image.) To get a sense of where things might be headed, here’s what I will be thinking about tonight.
On both sides, the race appears to have narrowed over the weekend. Iowa is a must-win state for Bernie Sanders. Neither Trump or Clinton needs it. And what other GOP candidates need, Iowa probably won’t deliver.
First, the Republican side. In the 4 surveys that started since last Saturday January 23rd, were conducted January 26-29, Trump‘s median support is 31.0 ± 0.4 % 26.5 ± 2.4 % (median ± estimated SEM). Trailing him are Ted Cruz at 24.0 ± 1.1 % 23.5 ± 1.5 %, Marco Rubio at 15.0 ± 1.3 % 18.0 ± 1.7 %, and Ben Carson at 8.5 ± 0.7 1.7%. The weekend’s results give the impression of net movement from Trump toward Rubio (and of course a more complex reshuffling among candidates could also have accounted for such a shift). Given the Iowa GOP rules, which award delegates strictly proportionally to vote share, these four candidates should get multiple delegates. Nobody else would get more than one delegate out of the 27 that will be awarded on Monday. However, I can think of two difficulties in predicting this outcome. First, the polls might be off on average. Pollsters have to guess about who will vote (for instance, by estimating turnout and identifying newly registered voters). Those guesses can be wrong. A veteran Iowa pollster, Ann Selzer of the Des Moines Register, has twice gotten results that show Trump with about 5 percentage points less support than other surveys taken at the same time. On Saturday, Selzer came in for Trump at 28%, not far from the other surveys. So, at this point, multiple estimates have now converged.
More important than a Trump win (which seems likely) is his vote-share percentage. As I’ve written before, I have done a detailed simulation of the delegate-allocation process. I estimate that to get a majority of delegates by Super Tuesday, Trump needs to get at least 30% of the popular vote in the early states (Iowa, New Hampshire, South Carolina). He’s at 33% in New Hampshire and 38% in South Carolina. Trump seems likely to come close to 30% in Iowa as well. According to my calculations of the delegate process through Super Tuesday, any finish in the 20’s is OK for him, though a second-place finish behind Cruz (or Rubio) is bad publicity.
Working in Trump’s favor, primary voters tend to commit fairly late in the process, and before the weekend, they were moving in his direction:
It was probably a good move on Trump’s part to skip last night’s debate. That move helped lock in opinion – and also led him to own yet another round of news coverage. That guy knows how to play the media. Also, as Josh Marshall has written, the appearance of domination matters.
Regarding other candidates, Ted Cruz could extract a win, but Iowa was supposed to be a strong state for him anyway. Marco Rubio could finish strong, but his bigger problem is the deeply divided field of candidates, which Iowa will not resolve.
As pointed out in The Atlantic yesterday and by me in The American Prospect a few weeks ago, GOP delegate selection rules work against establishment candidates – and favor Trump. My delegate-selection calculation assumes that the Republican field will still have at least four candidates after South Carolina. Right now that assumption looks OK. First, we have Trump, Cruz, and Rubio. Then Kasich may finish in second place in New Hampshire, which means he would likely stay in. If Jeb Bush stays in too, that makes five. In this way, the closeness and division of the “establishment tier” (Kasich, Rubio, Bush, Christie) works in Trump’s favor.
Speaking of the benefits of division…in New Hampshire, GOP delegates are awarded proportionally, but only to candidates who get above 10% of the vote. After that, unassigned delegates go to the top finisher. Currently, only Trump, Kasich, and maybe Cruz are above that threshold. That means that with 30% of the vote, Trump is likely to get 15 (and as many as 17) of the 20 delegates awarded. Doesn’t that blow your mind?
* * *
Now, the Democratic side. In three surveys that included January 29-31, and started on or after January 25th, Hillary Clinton is at 48.0 ± 1.5 % 46.0 ± 2.1 % and Bernie Sanders is at 42.0 ± 1.0 % 43.0 ± 2.5 %. So Hillary appears to be ahead slightly. However, Iowa Democratic party rules say that if a candidate gets below 15% in a precinct, his/her voters must be reassigned. That means most of Martin O’Malley’s support (5.5 ± 0.9 % 3.0 ± 0.4 %) has to go somewhere, and they prefer Sanders by about a 2-to-1 margin. Over the weekend, O’Malley’s support appears to have declined and the Clinton-Sanders margin narrowed, suggesting that O’Malley supporters will vote strategically. And then there are the undecided voters, who are 8% of voters. So although Clinton is still favored, there is quite a bit of uncertainty tonight.
However, I estimate that Sanders doesn’t just have to win Iowa. He has to win by a fairly large margin. Why is that? Clinton runs stronger in states with fewer whites than Iowa, which is almost everywhere else. So Iowa is a high-water mark for Sanders, at least for now. My guess is that Sanders has to win by 5 to 10 percentage points in Iowa to be competitive for the nomination. Exactly what he needs is hard to estimate, since a win will also get him media attention that could boost him further. In any event, a convincing Sanders win is the one genuinely newsworthy event on the Democratic side.
Topics:
“Iowa Caucus – What To Look For”
Final Ann Selzer poll.
I’m looking at the final Selzer poll from 2012. She did great except for Santorum. Could we be paying too little attention to Carson?
@bks: But she did capture the late movement toward Santorum (in the day-to-day breakdown). No similar effect for Carson this time around.
“I’m looking at the final Selzer poll from 2012. She did great except for Santorum. Could we be paying too little attention to Carson?”
Her one and only historical failing is a tendency to not get evangelical surges quite right. But if she’s underestimating someone in that fashion, it’d be Cruz, not Carson, IMHO.
If Trump wins Iowa (even by a little bit) that validates the primary polling to a large extent. One criticism of Trump’s poll numbers has been the uncertainty in measuring commitment of his followers. If that gets settled then the leads he has in other states look real.
What GOP result does Hillary want? Unlike the median pundit, I give Trump the best chance of beating Hillary. (And I have a rationale for that opinion, too long for the margins of this comment.) However despite my brain saying I should root for Cruz to weaken the GOP, my heart wants Cruz crushed like a bug. I’ll go with a prediction of the Selzer poll (I was told another one is coming out on Sunday, is that right?) Trump 29, Cruz 27, Rubio 13
I’ve been reading that people “in the know” (for whatever that’s worth) think Trump’s numbers are several points too high. His lead is still double-digits in Iowa so that may not matter, but the point is that if someone (likely Rubio?) sneaks up to 15-16% over the weekend, it’s not impossible that he could end up finishing close to, or even ahead, of Cruz and/or Trump. If that were the case you’d have to assume Trump’s standing in NH would automatically be imperiled, both because it would validate that his numbers are inherently too high, and also because Rubio and Kasich would both seem to have a path to winning NH.
I think Trump would need to win Iowa by at least 5-10 points to feel really comfortable about NH.
I am really curious about what quantitative argument there is that surveys are underestimating Trump’s support.
Sam: my guess is that whatever information exists to this extent is likely to be proprietary…which means it could of course be complete bunkum. What I do keep hearing is that a disproportionate amount of Trump’s support is from people who are less likely to vote than the average Republican voter.
FWIW, apparently Ann Selzer (whose track record is, for a pollster, exemplary) pegs Trump at roughly 5-6 points below what the polls currently say…
bks, I think Hillary’s team would still be rooting for Trump even if they shared your view that Trump is the worst overall candidate for her to face in the general. Cruz and Rubio both have (some) institutional support, traditional infrastructure, and campaign cash to stay in for the long haul, so neither one is going to be knocked out after Iowa.
Trump, on the other hand, has been running largely on the fact that he’s been winning. A loss could pierce that balloon, tank his soft support in later states, and drive him rather quickly from the race.
So if you like a long Republican primary — and I take it as a given that Hillary does — then you ought to be rooting for Trump on Monday.
Bush just passed Cruz at Predictwise (7:30 pm pacific)!?
http://predictwise.com/politics/2016-president-republican-nomination
I wrote to David Rothschild and he says it’s due to Cruz tumbling (vs. Trump) in the prediction markets:
https://twitter.com/DavMicRot/status/693239480673771521/photo/1
I think Sam’s last sentence should have been developed further, and I disagree that Iowa is must win. Now, if Bernie doesn’t win NH, then we’re in a different kettle of fish.
http://socraticgadfly.blogspot.com/2016/01/a-word-or-two-with-you-about-democratic.html
Actually, I reached my conclusion before even taking into consideration the superdelegates.
The basic issue is that Iowa Democrats aren’t like Democrats nationally – they’re whiter and more liberal. If Sanders can’t whale on Clinton here (“whale”=10 percentage point win), then he is probably doomed.
Not buying it, especially since Sanders is showing a surge in African American polling numbers, at least according to Fox.
It’s all about expectations, and framing, which is a subset of spinning.
Clinton had a lead of more than 100 superdelegates over Obama going into the 2008 Iowa caucus, and that was with her being less of a quasi-incumbent than now, and with Obama in far stronger standing coming into Iowa than Sanders was this year.
Let’s tackle that more.
Bernie Sanders does not come from a state neighboring Iowa, unlike Barack Obama.
Bernie Sanders was not “anointed” and given a prime speaking spot at the 2012 Democratic National Convention, unlike Barack Obama in 2004.
Bernie Sanders was not a presumed presidential candidate at the start of 2015, unlike Barack Obama at the start of 2007.
And yes, Sanders did what he did in Iowa last night.
Don’t be so sure Trump is an almost certain loser in the general election. One major terrorist attack could change that.
Or a new economic crash and recession.
There seem to be a lot of white working-class voters in places like NH that are actually for Sanders or Trump. It’s very different from the Obama coalition.
Sam, any chance that you could take the horribly depressing 2014 info down from the top of the page? It’s giving me indigestion.
I agree.
With all of this talk of momentum and Trump’s possibly marching to the nomination, I would like to add that the Republican Party will likely do all that it can to stop him and that will result in a significant campaign funded by the elites and moderates who have the most to lose. They are well-funded and they have time after NH to coalesce around whichever candidate–Rubio, Kasich, maybe Bush–they believe will stop Trump. The party will also pressure the bottom feeders to drop out quickly for the good of the cause.
There are many more anti-Trump voters out there but they are being steamrolled by Trump’s press and the numbers that Sam mentions (the 30% threshold). The 70% of GOP voters who don’t support Trump will come out and vote against him if they are given a credible alternative. This points to a long process for the GOP and one that I believe will result in neither Trump nor Cruz being the nominee.
Rubio is not a moderate. He looks moderate next to Ted Cruz, but his voting record is to the right of every Republican nominee ever.
That sounds like a brief summary of punditry on Trump a few months back (I could particularly single out a few pieces at 538, with almost identical content). By now, all those who raised this argument revised their assumptions (in Nate Silver’s case, he made his switch following the observation that the Republican establishment does not in fact seem to wield whatever power it has against Trump, or at least has avoided doing so up to this point).
I think that the Republican Party elites are starting to reconcile themselves to Trump, and convincing themselves he’s not so bad.
The guy they all really hate is Ted Cruz, and I think it’s because they regard Cruz as a true believer who they can’t control. Whereas Trump has no real ideology other than a belief in the awesomeness of Trump, and they imagine that when he’s President he’ll be eager to take their advice. They’re willing to run with Trump as a Cruz-killer.
I think they’re mistaken about this, personally. But a lot of them do personally know him and may like him, and they know and hate Cruz.
Feels GOOD!!! Just getting back to actual voting again rather than just pundits. Sam, I like your call on the effect of Trump skipping the debate. Listening to CNN and FOX analysts and the voters selected for interviews, it was going to be an out&out disaster, worst mistake he ever made. How about the Frank Luntz focus group? Good to come here for some sanity.
I am really interested in how well Trump does because I really do not know if he can deliver the GOP voters who did not vote in 2012 or Blue collar Independents who can choose to vote in the primary. Iowa and NH will set the stage for the next wave of discussion. Similar with Bernie. There is lots of enthusiasm — can it deliver a big win? If we could predict the increase in number of voters over previous years, we could easily track the winners.
I’m so glad you mentioned the Frank Luntz Focus Group. Is that meant to be taken seriously?
I’m not from the US, and came across this group on Fox a while back. I wasn’t sure if they were simply injecting some sort of comic relief, or if it was meant to be legitimate. I am now pretty convinced, having watched several more Focus Group sessions, that their intention is for it to be taken seriously.
I mean really, what the heck is that all about????? It’s like a weekly support meeting for bad actors.
Eva,
I really liked the concept. Whenever you have a political debate or a primary election, select a representative sample of voters who are most likely to vote or support the debaters, and ask them how they view their candidates performance, This is very entertaining and much better than the talk of pundits or those who are clearly partisan to a candidate or party.
BUT Frank has the same problems as all the rest of us — what exactly is a representative sample?
In the Iowa GOP Debate, how do you insure you have the correct number of committed and uncommitted voters for each candidate. Frank never states that he has matched a specific poll or poll of polls — just a group of Republican and Republican leaning Independents from Iowa who are interested in voting. So in a group of 35 Iowa potential GOP voters, Frank asks how many think that Trump made a terrible mistake by skipping the debate, all but 3-4. What about Trumps 30% in the Polls — Does this mean taht 60-70 % of Trump voters feel he made a serious mistake skipping the debate OR did Frank just not select a representative sample?
So, it is great entertainment, but I treat any insights as questionable.
The best part is when analyzing what answers in the Debates get the most/least approval from Frank’s audience. This is very credible.
@JayBoy2k: Actually, even if the sample for the panel was representative, the question itself could easily have introduced bias. It’s perfectly possible that a majority of Trump supporters, if asked point blank (and publicly too) whether he made a mistake by skipping the debate (especially if they take the person asking to be assuming a positive response is the “correct” one) would admit it was a mistake (at least if asked so *before* the post-debate analyses in the media, that by now mostly switched to the “Trump won the debate and Fox News lost” narrative).
But giving such an answer tells us nothing about how Trump’s move affected their level of support for him. They may admit it was a tactical mistake, or they may tactically admit it was a mistake, but keep supporting him nevertheless.
“I’m so glad you mentioned the Frank Luntz Focus Group. Is that meant to be taken seriously?”
Actually, yes! Not as a measure of anything in the real world. It’s pure propaganda, not even barely disguised.
But it’s to be always to taken very seriously as the definitive guide to Roger Ailes’ agenda.
and they (O’Malley voters) prefer Sanders by about a 2-to-1 margin.
The DMR poll yesterday said that O’Malley voter split evenly for Sanders and Clinton.
“When O’Malley’s voters are asked their second-choice option, they split between Clinton and Sanders, Iowa Poll shows”
https://twitter.com/JenniferJJacobs/status/693581326188871680
No, that’s wrong. That claim is very poorly sourced (read the replies).
Considering that O’Malley is at 3% and the survey claims 602 respondents, any claim from the Iowa Poll would be based on about 18 respondents. That would lead to a fairly uncertain result. Based on binomial sampling statistics, if the split was 50-50 between Sanders and Clinton (9 O’Malley respondents for each of those two), that could easily be 65-35 (one-sigma confidence interval, i.e. 68% of the time, the actual split will be between 65-35 and 35-65).
We do know that the PPP sample split toward Sanders, as per my citation.
Carson seems dead. Rand Paul may outpoll him. We shall see.
Sam, two points:
1) You don’t say anything to the contrary, but FWIW, the IA Dem race is not proportional in ways far beyond just re-allocation.
Even without the O’Malley factor, the Dem vote share could be 52-48 in favor of of one candidate, the reported results could vary immensely from that that figure.
First, the vote share is not even counted, let alone reported. Only the delegate share is counted and reported. Second, votes are weighted very differently in different precincts. Third, delegate allocation is complex within each precinct, and can vary widely from vote share within that precinct.
The Bleeding Heartland blog has an excellent rundown of all this, should you care, which I’m not sure you should. But one important upshot is that if the Sanders vote is concentrated in liberal bastions like college towns, it will badly hurt his reported results. In ’04 Dean almost definitely had a higher vote share than the reported results gave him.
2) While I’ve been strongly predicting Trump was the favorite since early August, I still think an 80% probability is too high. Not radically high, but still a bit.
The only way Trump loses, other than just losing his core vote of course, is if the race gets down to 2 candidates pretty early in the front-loaded process. And we’re starting to see late efforts to make that happen for Rubio.
The donor class finally seems to be coalescing around Rubio. Fox, (whose importance in GOP dynamics can’t be underestimated), is finally going all in on Rubio. Assuming Trump wins IA, the media ‘two stories out of Iowa’ rule seem to lean toward a ‘Trump Wins’ and a ‘Rubio Outperforms’ story, which Fox will help push, which could help boost Rubio to a clear 2nd in NH.
For the Fox point, in the last debate for example, you could see that in two ways. The Luntz “focus group” was essentially an advertorial for Rubio. And contrary to every bit of CW I’ve seen, Trump didn’t ‘skip’ the debate, Ailes forced him out with that weird baiting statement that pushed Trump into a corner. There’s more from Fox beyond the debate.
I don’t think the ‘winnow to Trump/Rubio early’ plan is likely to work for a bunch of reasons, including that Cruz is not a team player. But it’s looking more possible than it used to. It still relies on a number of bank shots, like kill Cruz now which will build up Trump, so we can then kill Trump. And it’s happening later than it should have. So while I do think Trump is a heavy favorite, there are still things that make him weaker than a more consensus candidate like Bush ’00, who really was an 80% favorite going into IA.
1) Yes, that’s a gerrymandering-like effect where Sanders supporters are packed.
Probabilities…hmmm, maybe. I guess I think 65% for Trump, 80% for Clinton. A big uncertainty in my mind is the actual vote on Iowa, with ensuing press reaction. If all goes as expected, these numbers would change afterward.
The hardest thing to gauge right now is what happens after Super Tuesday, when the field will narrow, but where will the candidates be? Will one have an actually-insurmountable or psychologically-insurmountable lead? Will the increasing validity (and frequency) of Clinton-v-Trump polling data affect the minds of voters? Come to think of it, maybe I should stick with just the Super Tuesday prediction, which has variables I can see. That’s an easier calculation.
Hillary will win the delegate count tonight even if Sanders has more voters because of distribution . She has the better organization this time and paid attention to Obama 08 and Paul 12 . Precinct leaders have been instructed to keep O’Malley viable (15 per cent) to keep Sanders from evenly splitting the delegates just like Obama did to her with Edwards .
Gallup has an interesting graph today. Sanders has overtaken Clinton in Net Favorable rating among democrats.
http://www.gallup.com/opinion/polling-matters/188957/voting-begins-sanders-popular-clinton-dems.aspx?utm_source=alert&utm_medium=email&utm_content=heading&utm_campaign=syndication