Protected: Two ways to estimate primary outcomes without polls (transcript)
This content is password protected. To view it please enter your password below: Password:...
Indiana may not matter any more
Media types want you to get your knickers in a twist about Indiana. However, the data suggests that it doesn’t matter any more. Rationally speaking, it is probably time to stop writing so much about the Republican race for delegates. Also, may we have a moratorium on “brokered-convention” articles please?
Today I write about the PEC delegate snapshot. It is based on data posted here. All polls are current, including Trump +6% in Indiana (n=3 polls). Based on Tuesday’s voting, in which Cruz underperformed polls by a median of 4 percentage points, I will no longer assign a Cruz bonus. Note that Trump overperformed polls by a median of 8 percentage points.
As of today, for recently-unpolled states (NE,WV,OR,WA,MT,NM,SD) I will start using Google Correlate-based estimates. Of those states, Trump is favored in West Virginia (34 delegates) and is near-tied in Oregon and Washington (proportional representation). The rest are Cruz states.
Put through the PEC delegate simulator, the median delegate count is 1333 (interquartile range 1304-1339). The probability of getting to 1237 delegates is 98%:
What if we assume that Trump will lose Indiana? In that case the median drops to 1284 delegates (interquartile range 1278-1287). The probability of getting to 1237 is now 97%:
The 1% change in probability is inconsequential. The main effect of forcing a Cruz win in Indiana is to reduce uncertainty in the delegate count, which you can see in the narrowing of the historgram.
Close states (Oregon, Washington, and New Mexico) happen to use proportional rules, so they contribute very little uncertainty. Winner-take-all or nearly-winner-take-all (i.e. district-level rule) states are either strong Cruz (Nebraska, Montana, and South Dakota) or strong Trump (West Virginia, California, and New Jersey).
Most of the remaining uncertainty comes from district-level races in California. With California polls showing Trump +18% (Google Correlate says Trump +31%), it will take a highly coordinated effort by Cruz and Kasich to pick up many of its 53 districts. They would use geographic information like this Sextant Strategies survey to guide their efforts. At the moment, the likeliest outcome is for Trump to get at least 160 out of 172 delegates in the Golden State.
Topics:
How do the very strange and complex delegate selections rules in West Virginia affect things? The rules make it possible that ballots are disqualified, etc. I’m not sure why that would favor a given candidate, but without that sort of chaos Trump clearly wins in WV. With it, it seems like odd things could happen.
I just code it the conventional way as indicated in the rules: 25 delegates statewide winner-take-all, plus 3 delegates in each of 3 districts, winner-take-all. I think further attention is overkill. At some point one must rise above the dreary detail and make a decision about what is important.
Trump now has a team in place to minimize the damage as much as Cruz does so I’d expect a wash on this front.
ITYM … effect of forcing a Cruz win in Indiana is to increase uncertainty slightly…
No, it’s correct. Extended a bit now.
My mistake!
http://physics.nist.gov/cuu/Uncertainty/glossary.html
Thanks Sam!
Really appreciate today’s post. It’s nice when you tackle the media narratives head on.
While I’m from Virginia, I grew up in Indiana. Instinct tells me that Trump will win. Your analysis only reinforces that.
This is going to be an interesting election. Sam, what will you do with all this unpredictability that will shortly manifest itself? For instance, many Republican voters may stay at home. In this uncharted territory, how can you quantify what that number will be? Crossover voters? People who’ve never voted? Pissed off women?
All may shortly become chaos, from a statistical perspective. We shall see.
Very interesting from a political science perspective.
Keep up the excellent work.
The presidential elections will be well polled. LV (Likely Voter) is a category that pollsters shift too late in the election cycle. That should pick up any Republicans that decide to stay home and any of the other categories you mention. Polls only should serve quiet well.
Thanks Sam,
It is a great feeling getting access to unvarnished facts and logical predictions. The media and pundits love a horse race.
It would seem that the Cruz and Kasich teams must have some inkling of the writing on the all. The announcement of a VP selection by Cruz seems bizarre, and why would Carly attach herself to a boat anchor? Does Kasich really want to be the 2016 GOP VP nominee?
Mostly, I thought this as positioning for which Politician inherits the remains of the GOP establishment AFTER the elections.
This has not been a normal Presidential election year… not many sane players
Forcing a Cruz win in Indiana while holding everything else fixed seems non-physical. If Trump loses Indiana it may well be a national slip — Indiana drop might be correlated with other states. So what happens with a X% shift nationwide towards Cruz? (X= 6, 8, 10…).
I ran a google correlate experiment using % of delegates won by state for 43 states (removing Guam, PR, Northern Marianas as google correlate does not consider them states).
The following are the 10 best keywords for trump.
Interesting that Degrassi shows up again. Its a canadian tv show on which Drake used to star when he was younger.
0.7521 the anthem
0.7387 45 years old
0.7346 degrassi season 13
0.7332 new years eve 2013
0.730 165th birthday
0.7269 dj drops
0.7260 degrassi season 11
0.7252 the hit
0.7241 alien surf girls
0.7238 chaise lounges
Fresh polling out of Oregon at Huffpollster has Trump/Cruz/Kasich at 43/26/17 with 13% undecided.
Proportional, so will really not matter at all.
Needed a break from the current primary race (Indiana) and needed to fill my mind with something new. Have you used Google correlate to run numbers on a possible Trump vs Clinton race and if so, will you share?⚖
I have not had any good thoughts on how to do that. In the primaries, Google Correlate worked well because the primaries are a drawn-out process. Therefore we can get information about the web search/candidate relationship from some states, and then impute them to other states.
I actually think Trump v. Clinton can be addressed by normal polls, even this far in advance. Working on that.
There are some states where there has been little or no polling, and I suppose Google Correlate could be an interesting thing to try to apply to estimate the situation there. Most of them are places not considered swing states, but this cycle might open some of them up.
Exceptionally good to hear that you are working on Trump v Clinton. The signal to noise ratio everywhere else is horrible.
My prediction is that the polling will change during the campaign. I do not believe the opening polls will be the same as the closing polls. All indications are this will be a Presidential Campaign for the ages.
That said, I await your analysis eagerly.
Like Matt suggests, I assume you could use this method to “fill gaps” in state polling, perhaps with more reliable results than previous attempt to accomplish that.
But of course by the time the election is near, it is unlikely there will be material gaps in states relevant to the presidential outcome.
However, I do wonder if there is a way of applying these techniques to Senate (and other state) races, and House races.
I don’t do any of the sophisticated modeling that Sam Wang does here, but I’ve been tracking all the state level general election polls and doing simple averages in each state to try to get a view of how Clinton vs Trump (and other combinations) look and just as importantly how that changes over time, since we do indeed still have 193 days until the election. In any case, for anybody interested, it is at http://electiongraphs.com/2016ec/ .
You can always choose none of the above. A Poll..a poll with interesting questions.
http://www.rasmussenreports.com/public_content/politics/elections/election_2016/24_opt_out_of_a_clinton_trump_race
It would be interesting to see how this tracks .. It does not “feel” like a low turnout election..
Just wondering if the people here agree with the probability I am assigning of each of the following outcomes. Just off the cuff.
1. Clinton landslide – 70%
2. Narrow Clinton win- 19%
3. Narrow Trump win -10%
4. Trump landslide – 1%
Trump has been universally attacked in the media… and in the political realm. Yet, he is going to win the nomination. Only one conclusion can be drawn: he is a brilliant campaigner. Of course, because it is forbidden to call him a brilliant anything, this will be quickly denied. But if he isn’t a brilliant campaigner then he has a very, very special relationship with the tooth fairy, because there just aren’t any other explanations.
Hence, the Presidential *Campaign* will be very interesting and I don’t thin it can be handicapped right now.
What did you give as the chance that Trump would be the GOP nominee in September 2015?
To the “nobody knows anything” sort of argument I’d respond: Trump’s dominance in Republican primary polls was established early on and he never really gave it up (the closest anyone came to taking it was a brief Ben Carson boom, and there was only ever one outlier poll that had Carson as the “frontrunner”).
Trump currently does not do well in head-to-heads against Hillary Clinton, despite the fact that he’s well-known to be the most probable nominee. It’s not at all analogous to him taking the lead over, say, Jeb Bush; that happened almost immediately after he announced.
In September 2015, there were 16 possible next Presidents. In April 2016, there are basically only 2. So comparing September of last year to April of this year is a bit misleading I think.
It’s also entirely possible to handicap the general election–betting markets have been doing it for months now. Generally, Hillary seems to be trading at around 70% to win–not too far off from Mark’s thinking.
I think it has a very low chance of being a landslide in the Reagan/Nixon/Johnson ’64 sense, in which one candidate takes almost all the states. Partisan alignment and hatred of Hillary Clinton among Republicans are just too strong.
I think there’s maybe a 50-50 chance it could be a thumping as strong as Bill Clinton’s wins in 1992 and ’96, without a third-party candidate confusing the issue (the map will be a little different, though).
…The rest of the probability would be dominated by the scenario of an Obama-scale or narrower win. A 10-11% chance of Trump pulling out a win is probably about right–I think the prediction markets are too timid; if I played that game I’d buy Clinton right now, because she’s probably underpriced at 70 cents on the dollar.
I concur re: the odds of a Reagan ’84/Johnson ’64 victory. Too many states are far too conservative for that to happen. Even a GOP candidate as weak as Trump is going to get 150-200 EV just for not being the Democrat.
That said, a Hillary win on the order of Obama ’08 or Clinton ’96 would give her close to 400 EV, which would be about as big a margin of victory as a D candidate could reasonably expect to get today.
…On the other hand, with prediction markets I guess it depends on what time horizon you’re playing. There will be a point at some time in this campaign when the conventional wisdom is that Trump is doing unexpectedly well and may actually pull it off; if you’re planning on dumping your Trump shares on a greater fool then, buying him now is probably a bargain.
If you give Clinton every state where she’s led or tied in some recent general-election poll, I get 390 electoral votes. It’s conceivable she could get a few more with super-high minority turnout. But, yes, the vicinity of 400 is probably a solid ceiling.
…Much better results than ~390 electoral votes basically require flipping Texas, which I think is too much for Democrats to hope for in the 2016 cycle, though it could get close. Mobilizing Democrats in Texas might be worthwhile anyway for the sake of downballot races.
Matt. I read a while back that flipping Texas may be closer than most think. The underlying demographics are favorable. Voter registration and turnout are the real problems. The SCOTUS weighing in recently on the Texas voter ID law may end up being a big deal with Latinos registering to vote against Trump. Stay tuned for that story.
I think that’s right. There are a handful of true tossup or GOP-leaning states that Hillary would win handily in a Clinton ’96-type victory: CO, NH, IA, VA, OH, FL, NC. That gets her to about ~360 EV. But to win states like IN, MO, GA or AZ, she’d have to win the popular vote by a truly impressive margin–probably >10%. And she’d need some of those to get to 400.
In polling averages I’ve seen so far, it looks like she’s got a 8-9% lead over Trump with about 10% undecided. If those undecideds break mostly for Hillary, I think she gets to 400 EV.
Yeah TX is definitely closer to the median state than many other Southern/Plains states–IIRC Romney won it by about 15% in 2012, so it’s probably closer to 18-20% to the right of the country as a whole. Over the last two decades, it’s probably moved about 5-6% leftward.
Voter turnout and enfranchisement are certainly key issues. It’s still out of the Dem’s grasp this cycle, but if current demographic trends continue and if Latinos there become better voters (I read somewhere that only about 50% of TX Latinos vote versus 60ish% nationally), it wouldn’t surprise me if TX became purple in 3-4 cycles.
That seems improbable, given that an incumbent only got 332 EVs in 2012 and won by +4 after polling not much higher for much of the prior six months. The polling showed a smaller undecided at this point in the race (late April and early May) for both the R and D frontrunners during 2008 and 2012 though.
Trump/Clinton has bigger undecideds at this stage than either of the past two Presidential races at the same point in time.
Obama was more personally well-liked than Hillary Clinton, and in 2012 he was the incumbent, but he didn’t have Donald Trump to run against. That makes a huge difference. 2016 is going to be a contest between two candidates with historically high negatives, but one is higher than the other.
One of those states where there’s no polling yet, and we might be able to use Google Correlate to see something interesting… is Texas.
My off-the-cuff guess is that it’s still safe Republican for 2016. But the margin is shrinking. Hillary is leading in current polls in Arizona, and the probable nominee is the guy who got crushed in the Texas primary.
Then on the other side of the coin, there’s the question of how well Trump could do. It’s harder to tell that than to estimate Clinton’s ceiling, because I think Clinton is actually very close to her electoral-vote ceiling now. Whereas a Trump win would take some kind of significant material change in the political situation (I’m thinking, probably, multiple changes at once: say, a major economic crisis kicking off a new recession before November, and a major ISIS-sponsored terrorist campaign causing mass casualties in the US, and Bernie Sanders going off the rails at the Dem convention and urging his supporters to storm the convention center… some improbable cluster like that).
It would be good to have a common definition of “Landslide”. What is yours? There have been 57 Elections of which exactly 19 have been won by more than 10% in the popular vote — the last one Reagan by 18.2% in 1984.
I would say that anything less than 10% would not be a Landslide. i.e Clinton over Dole or Bush with a 3rd party candidate does not qualify.
Having said that, I come to this site, to get my own bias and lack of knowledge on importance of various factors validated by facts, polls , and Dr Wang.
Please raise this topic again after Sam’s 1st set of polling data. For the record, I think the current situations favor Hillary. I am just not sure how much.
Respectfully, I’d add Obama ’08 to the “landslide” category; he won over 360 EV, including solidly Republican places like Indiana and part of Nebraska. Obama’s winning margin in that election was about 7%.
I think the general closeness of recent presidential elections, and particularly Bush’s wins in 2000 and 2004, has caused us to define landslides down. I remember the Eighties, and particularly the dread I felt watching Reagan sweep the whole country except for DC and Minnesota on election night 1984; to me it ain’t a landslide unless nearly the whole map looks the same color.
I trust everyone has noticed that the Electoral Vote site is quoting the essence of this post, complete with Dr Wang’s graphs. Nice to see them recognizing the value (and by no means the first time they have noted this site).
What is with the great Satan?
He shows polls showing Trump ahead in Indiana and then in polls “plus” shows Cruz winning.
Trying to sell more newspapers?
Well, it is a test of the demographic approach.
The Argental Satan (love that name, Sam) has things set up so no matter what happens he wins. His baseball and sports stuff is much better.
The problem with coming here is that it’s like going to a horror movie and already knowing that Fredie Kreuger is going to get it in the end. Or reading a mystery novel and flipping to the end to see who killed whom in the kitchen with a farm implement.
Although it is real fun to watch Ted flop around like a fish out of water. He and his crew HAVE to know…
NYT article says that Cruz delegates are not as rock solid as assumed. There will be no brokered convention, but the damage to GOP unity will have a big effect on the November election. I don’t see where Trump is going to pick up enough votes in swing states to offset the number of Republicans who just won’t vote for him.
http://www.nytimes.com/2016/05/02/us/politics/ted-cruz-delegate-count.html?_r=0
Got that right.
Sanders seems to be winning the Indiana Dem primary: another modest but significant polling miss.
Quick question: How could the Indiana polls be as far off as they were for Hillary Clinton and Bernie Sanders? The average of the 6 most recent polls on Huffpollster gave HRC about an 8 point lead, and looks like she is going to lose by 6 – 7 points.
People freaking out in various ways now all over the Internet. I think we are officially entering the fog of war at this point. Data, we need data.
Did Indiana matter?
no
It seems to have popped Ted Cruz’s delusion that he could still take this. But it would have happened sooner or later.