Polls, Polls, Polls.
Looking beyond FiveThirtyEight
As was correctly noted in comments on my post yesterday (and later also noted on Facebook), FiveThirtyEight is not the only game in town, so what are the other poll aggregators/analysts are saying on this weekend before the election?
The team at HuffPost Pollster:
Our model of the polls suggests Clinton is very likely leading. (In >99% of simulations, Clinton led Trump.)
Their national model puts Clinton at 48.2% and Trump at 42.7%, with undecideds at 4.8% in a head-to-head contest. The multi-way model puts Clinton at 45.7%, Trump at 40.2%, Johnson at 5.1%, Undecideds at 5.4%, and Other at 3.6%.
They put the odds of a Clinton win at 98.1% and project 302 Clinton electoral votes.
Moving elsewhere, Sam Wang at the Princeton Election Consortium likewise has the odds of a Clinton win at 98% and predicts an electoral vote result of 312 v. 226 for Clinton.
The NYT‘s Upshot is a little less bullish on a Clinton win, putting her odds at 85% to Trump’s 15%.
Meanwhile, the FiveThirtyEight map keeps getting pink, if not red, with Clinton having now a 64.5% chance in that model (with Nevada now in the slightly Trump category).
So, there is consensus that a Clinton win is more likely than one for Trump, but Nate Silver’s model is a lot less bullish than either Pollster or Wang.
I’ll stick with Silver, who still projects a Hillary win. I just read something from Wang about how the polls really didn’t get Brexit wrong. They did. The polls may not have been as pro-Remain as the British pundits were, but they still got it wrong and they got it wrong in the exact way poll averaging is supposed to get it right.
But we’ll all know for sure in a few days.
Mike
Nate Silver had a good post a week or two ago about why his calculation of each candidate’s odds is more conservative than others. A lot of it had to do with what curve you use (not everything has to be a normal bell curve) and which one is most appropriate for small sampling sizes relative to the total population. As a commenter on another post said, this is somewhat of an art.
Nate’s choices end up giving better odds to seemingly unlikely outcomes in either direction (like a massive Hillary landslide or a Trump victory). As Brexit showed, and as the Michigan primary showed, unlikely outcomes can and do happen. Anyway, a better explanation was given by Silver, so you should go there if you want true understanding.
And I think it was Dr.Taylor himself who said in another thread: 538 has been the gold standard for three straight elections. Until it’s dethroned, you’ve got to at least give it some respect.
I think most the difference boils to 538 accounting for the possibility that the polls are way off (as they were in 80′, although they still had Reagan winning). For me, the danger here is 2000. Polls going in showed Bush with a 3-point lead and RCP was projecting a Bush landslide. That didn’t happen. Nader polled way less than expected and there was a late shift (which Ted Frank blames on the “October surprise” of Bush’s DUI being revealed; I’m less convinced). As Silver pointed out yesterday, the *average* election has been about 2 points off the polls. That means a narrow Trump victory is possible and a Clinton landslide is possible. But anyone who put the odds of a Clinton win at 98% is being overly optimistic. We actually have *fewer* live interview polls than we did in the last cycle and that off by 3 points.
The only poll that counts will be the one taken on Tuesday.
Let me put it this way. If I were a gambling man and a gambling house gave me 50-to-1 odds on Donald Trump, I’d take that bet in a heartbeat.
@MBunge:
http://election.princeton.edu/2016/06/23/brexit-home-stretch-polling/
Wang had it 70/30 remain, calling it microscopically in favor of Bremain.
These poll aggregators (538, Real Clear Politics, Sam Wang, Huffington Post) reach varying conclusions for the individual key swing states (FL,IA,NH,NV,OH). But they all agree that Trump can’t get all of them, which is what delivers 270 to Clinton.
I was expecting a narrow popular vote win accompanied by an electoral vote landslide, but the latter may not happen. A win is a win, but the lack of a mandate would embolden Trump (and probably help him to find investors for Trump TV or whatever it is called.)
@Pch101:
I’d like to see Trump crushed and Trumpism with him, but I’ll take him just being able to have a TV network, which will hopefully fail as all his businesses have.
@MBunge: I’ve heard two different takes on “polls were wrong on Brexit”–the polls may, indeed have been wrong or selective reporting on them made them seem as though they were wrong. Much the same way as you are more supportive of Silver than the four or five other aggregations that are more optimistic.
Understand, I’m not supporting a Hillary blowout or a total repudiation of the GOP because of Trump–personally, I think that has hit on who the GOP (including the “establishment”) are pretty well. I don’t have any take to speak of anymore. I expect that, starting in January, we will be witnessing Act III of the clown show that passes for governing now, but I’m just gobsmacked by the reality that 40% of my fellow citizens actually believe all of the nonsense of Hillary as Marxist destroyer of economies/spawn of Satan and Donald Trump as a credible candidate for President. It’s just breathtaking.
@Just ‘nutha ig’rant cracker:
Having served 20 years in the military, I have a lot of Republican friends, and if my social media feeds are any indication, most of them believe a lot of things that are just plainly false. They buy Trump’s statements without a second thought even though 90% of what he says ranges from questionable to outright bullshit.
And a lot of these people are far from stupid. I’ve known them for many years, worked with them on missions and projects that took a significant level of intelligence and independent thought. That’s the part that gets me–people who I know are smart, but for reasons I’ve yet to figure out, they still drink the Kool-Aid.
@Mikey:
As is the case with many of the buildings that carry the Trump brand, I would expect that it would be largely a matter of licensing his name in exchange for money and some gratuitous participation (a seat on the board, some TV appearances.) The operations would be handled by the right-wing media professionals.
There is actually now early voting results to go along with polls. They show a huge Latinx wave in AZ, TX,NV, NC, and FL. They also show a surge in black early voting, after disappointing early results. All this looks good for Clinton.
I am convinced that Clinton will outperform her polls , which generally undercount minority voters in Presidential years. Between that, and better Clinton GOTV, I’m betting Clinton will do 3-4 points better than her polls.
@Pch101:
Bannon and Ailes, presumably.
@Mikey:
It’s the power of belief. Facts don’t matter much when the political system is turned into a battleground for a culture war.
In a contest between facts and gut feelings, gut feelings almost always win.
A lot of intelligent people have said that Christ will return…before the Cubs won the World Series!
So where is he?
@stonetools:
You’ll be amused to learn that Breitbart, that paragon of responsible journalism, is breathlessly “reporting” that Trump has twice the support from African-Americans as did Romney.
I imagine this is pronounced “La-tinks”. LOL
If we know anything about cognitive biases, we know it changes how you even see the incoming data. It doesn’t much matter how smart you are if you’re working from bad information.
And there’s just a lot of straight-up propaganda and lies and misinformation. I know a retired electrical engineer from Motorola who was genuinely puzzled when I told him that you don’t get welfare simply for being black. No joke.
“If I watched Fox News, I wouldn’t vote for me either.” -Barack Obama
And there’s just tribalism. Like how every month when Doug writes up the Obama jobs numbers post, you wait with amused anticipation to see what the obligatory negative comment will be by the second or third sentence.
@dxq:
No, it’s actually pronounced La-TEEN-ex, and is preferred by some Hispanics over “Latina” and “Latino.”
@Turgid Jacobian:
And Leave wound up winning by a full 3.8 points, with the poll average being off by over four points. That is exactly the kind of result that people doing what Wang is doing are supposed to avoid, particularly when they start assigning winning percentages to the whole thing. It’s one thing to predict someone winning by 50.5 and they lose by 49.5. It’s another to say that there’s a 70% chance someone will win, then he loses and does so by a significant margin.
And Wang gave Remain a better chance of winning than Silver is giving Hillary.
The other thing that strikes me is that Wang and the other Hillary 90%ers are basically forecasting the same electoral vote result in 2016 as in 2012, after missing that mark in 2012. I know we’re supposed to trust the stat heads for everything but my common sense tells me that Clinton/Trump should not produce the same result as Obama/Romney.
Mike
@dxq:
These people actively seek out bad information. The confirmation bias is self-inflicted.
There is a right-wing media industry because there are people who want to believe what it has to say. Facts won’t sway them because it was never about the facts.
Speaking of dysfunctional tribes, there needs to be MAJOR housecleaning at the FBI. Those deranged republican idiots are literally using Breitbart as source for investigations now.
@stonetools:
One of my correspondents on Twitter made this point. It’s kind of silly to believe that millions of people’s votes wax and wane that fast. Probably what the polls are measuring is enthusiasm (e.g., when Clinton is doing poorly, her supporters are less likely to answer polls). But in the end, everything will come down to turnout. Trump doesn’t have a ground game, but he’s running a very different campaign. Black turnout is down. But if the reported wave of Latino turnout is really that big, it could make Clinton outperform her polls and doom Trump in Florida.
Huh. It’s almost like it would have benefited the GOP to NOT nominate a foaming-at-the-mouth anti-immigrant demagogue.
@CSK: Latinx rhymes with Kleenex? Wow, that’s terrible. I get it from the tendency to accent the penultimate syllable in Spanish, but it’s really odd in English.
I assumed it was pronounced “Latin X”, like “Malcom X”, which is also odd for something spelled as a single word, but the single word is spelled oddly itself, and clearly isn’t pronounced the way it is spelled.
English has enough screwy words — can’t we make up new ones that aren’t screwy?
@MBunge:
Your common sense? More likely that’s the voices in your head telling you things.
@Gustopher:
The word was created in response to the fact that Spanish is a gendered language, and that Latinos as a group are denoted by the male plural, so it’s presumed (by proponents of the term “Latinx) to be innately sexist. And yes, it is La-TEEN-ex. I had to look up the pronunciation.
I should add that I’m not crazy about having to write sentences such as “a student should go to their class,” but it’s required in some instances now to demonstrate inclusiveness to those who feel they have more than one entity dwelling within them.
@MBunge:
Wang’s prediction for the electoral vote:
Barack Obama 303 EV, Mitt Romney 235 EV….has a 22% chance of being exactly correct. The next-most-likely outcome is Obama 332, Romney 206 EV
The actual result: Obama 332, Romney 206 EV. Note the comment immediately above. (The difference between Wang’s most expected and second most expected outcome was Florida, which Obama won with a spread of less than 1%.)
Wang’s prediction for the popular vote: Obama 51.1%, Romney 48.9%.
The actual result: Obama 51.1%, Romney 47.2%.
http://election.princeton.edu/2012/11/06/presidential-prediction-2012-final/
Aside from demonstrating your inability to accurately read and process information, what exactly have you proven here?
Snapshot on the ground reporting from NH: I was in one of the population centers in the southern part of the state for most of this morning. The Trump bus was apparently driving around in circles, and the coffee shop I was in was packed with GOTV ground troops sporting Clinton gear, with clipboards, walk lists, etc. Both campaigns look very active. We’ll see.
@Jen:
As your neighbor just to the south, I can testify that no politicking, none whatsoever, is going on here. And really, why should it? Clinton is going to win here.
Huffington Post’s final electoral vote prediction for 2012:
Obama: 281
Romney: 191
Tossup: 66
Every state that Huffington Post called for a particular candidate was accurate.
It also favored candidates for each of its four tossup states:
CO – Obama
FL – Obama
NC – Romney
VA – Obama
Huffington Post got all of those right. So in essence, Huffington Post predicted the electoral vote with 100% accuracy.
http://elections.huffingtonpost.com/2012/romney-vs-obama-electoral-map
I guess that the sandstorms on Planet Bunge must distort what gets through on the interwebs.
Just two more things.
1. When it comes to pollsters, always go with the guy who has skin in the game.
2. The Clinton campaign has spent hundreds of thousands, maybe millions of dollars on polling that is supposed to be better and more precise than what is available to the general public. Are they talking and acting like people who’ve been told they have a 90%+ chance of winning?
@MBunge:
Like pros who care about the downballot, especially the Senate E.g., the attention paid to AZ.
@MBunge:
All pollsters and poll aggregators have skin in the game.
They want people to go and vote. If it’s close, more people will vote. It’s also more likely that voters who don’t want to vote for Trump but would prefer not to vote for Clinton, will vote for Clinton if the election looks like it’s going to be close.
I noticed 538 has both NV and FL tinted pink. Early voting strongly suggests Clinton has both of these pretty well locked down.
http://www.ktnv.com/news/ralston/the-nevada-early-voting-blog
@Turgid Jacobian: “they got it wrong in the exact way poll averaging is supposed to get it right.”
what does that mean?
@PJ:
Not Sam Wang at Priceton, one reason he is the best.
The need for clickbait is one reason 538 has become so unreliable. (I pretty much disregard 538.)
@charon:
Bragging rights 🙂
@CSK: I understand the rationale for it, and I wholeheartedly approve of the idea — both in Spanish and English. I simply find the spelling and pronunciation to be ridiculous.
English needs a pronoun for “that person over there, who probably has a gender, but I don’t really know or care what that gender is” (singular “their” is probably winning there), and Spanish has much greater problems with everything being gendered.
I’ll eventually get used to it, just as I have gotten used to singular their, I just wish it followed normal spelling rules, latinex or something. Secret vowel sounds in a string of consonants is one of the few problems neither language has.
Somewhat OT (although it does concern the election), but this was too hilarious not to share.
Enjoy 🙂
@CSK: There are reports of literal busloads coming in to help knock on doors from MA–no politicking there because all of them are here!
@dxq:
“Drink!” (It’s why I don’t read those posts at work.)
I think this article is overly harsh to Silver, but it makes the point I was trying to make yesterday. By adjusting every single poll based on ‘trend’ it exaggerates trends and introduces uncertainty and volatility, rather than reducing it to more realistic levels.
As far as Brexit, I think there’s one thing that explains the polling situation that I never see mentioned: Jo Cox. She was leader of the “Remain” campaign. Leave was narrowly leading until she was assassinated. Her death shocked the UK and dominated the news in the days leading up to the vote. I suspect that her assassination had the effect of discouraging “leave” voters from responding to polls. After her death, polls swung back to “remain” but people hadn’t really changed their minds.
I view Brexit as a caution against response bias.
@MBunge:
Because they’re not taking it for granted and giving up? GOTV is going to be key. HRC and her people – who aren’t stupid – are well aware of that. I’m tickled blue – dark blue – that they’re not wasting these last few days assuming the win is theirs.
@Gustopher:
There was a movement 6-7 years ago to replace the standard English personal pronouns with “ze,” “zir,” and “zes,” but it seems not to have gotten far.
@HarvardLaw92: I hadn’t seen that yet. Thank you – it was hilarious! I especially liked the line about four years of the apprentice. Well played, songwriter(s).
@Jen:
Do you have any sense of whether the southern invaders are Trumpkins or Clinton supporters?
Something I found truly shocking was the margin by which Trump won the Republican primary in Massachusetts: 49.3%, which according to several sources, was the highest in the country. Kasich was a very poor second at 18%, and Jeb Bush garnered an abysmal 1%. Something like 22,000+ Mass. Democrats changed their registrations to vote in the Republican primary (whether they did it to jack up Trump’s numbers, because they knew he’d be a loser in the general, and then went back to being Democrats, I can’t say.)
Clinton, on the other hand, had a very tough primary race against Sanders in the Mass. primary.
CNN reporter Eric Bradner tweeted this early this morning:
The idiot Trumpkins in the follow-up tweet stream are comedy gold.
https://mobile.twitter.com/ericbradner/status/794863521272098816
@Kari Q:
Turnout for Brexit was 72%, versus 58% for the 2010 and 2015 general elections. So below-average turnout was not an issue.
But seniors aged 65+ had a turnout of 90%. Between the xenophobia and scaremongering about the NHS, they strongly supported Leave.
https://www.theguardian.com/politics/2016/jul/09/young-people-referendum-turnout-brexit-twice-as-high
What surprised some pollsters is that a large proportion of the considerable undecided vote chose Leave instead of going with the status quo. If there is a parallel to be found in the US polling, then it could be with surprises from our undecideds and third party voters (although I personally doubt that they will be so skewed that they hand a win to Trump.)
@Pch101:
My comment wasn’t about turnout, though. It was about not responding to pollsters asking how people plan to vote. The point was, even though “leave” voters stopped talking to pollsters in the wake of Cox’s assassination, they hadn’t changed their minds and they were still going to vote. They just didn’t want to talk about it.
Silver is saying all the polls are off…I just doubt that. Some of them I am sure. But all of them? No. But then again, we don’t know what will happen until it happens.
@MBunge: On Brexit, the polls were heading in the right direction. If the vote had taken place a few days later and another poll had been done it might well have shown a win.
@MBunge: @MBunge: On Brexit, the polls were heading in the right direction. If the vote had taken place a few days later and another poll had been done it might well have shown a win.
@MBunge: Keep in mind, that a lot of Republicans are not voting for Trump. I have voted a straight Republican ticket for years, but no more. I voted today and I did not vote for Trump. And I am not alone. But Democrats are turning out for Hillary. That gives her an advantage.
@CSK: Clinton won more votes in the primary season than any other candidate, Republican or Democrat.
@Kari Q:
The numbers would suggest that your scenario did not take place.
For example, if you take these polls, eliminate those that did not measure undecided voters, and then perform a running calculation of their medians and averages (I measured batches of eight), then you’ll find that support for Remain was running in the mid-40s while the undecideds were in low double-digits.
https://ig.ft.com/sites/brexit-polling/
Remain led more polls than Leave, but few polls showed majority support for Remain in a contest that consisted of two only choices. The undecideds made the difference.
These polls – too many, and too frequently. It used to be just the Gallup Poll and maybe one more. And that was every few weeks. That was enough. Today there are so many polls and it seems they are conducted every other day, if that is possible. The concern here is why – why so many and so
often ? What is the reason ? Think about being bombarded with these numbers and “trends” every day for months on end. You see what could be going on here ? A type of mind numbing control, or influencing by psychological means. Think about a time when you saw a few people looking up in the sky at something. Soon, the whole crowd is looking up.
Or you are on an elevator. It stops, the door opens, someone gets off, and everyone else starts to get off even though it is not their floor. The crowds that wait in line over night waiting on the store doors to swing open on a Black Friday sale; to purchase merchandise that is not very important to them or not needed. People start to hear that so and so has this locked up over and over. After months of this stuff they start thinking maybe they had better just go ahead and vote for that person, everyone else is and they are going to win anyway, who wants to vote for a loser ? So you see what starts to happen here.(* see below). And those electoral boards that are on large touch screens. The moderators just obsess over the numbers, the scenarios, the “route to 270”, thousands of different numbers combinations, the effects of rain or even sunspots. You wonder what those people will have to play with after the election. Maybe they should go to Las Vegas and get a job or teach 6th. grade geography.
A good idea would be to go back to one or two polls, conducted no more than twice a month. This should be taken up by the elections board or whoever runs the elections.
* “Monsters Are Due On Maple Street”: a story written by Rod Serling and presented on his “Twilight Zone” tv program. This episode has been studied and analyzed by sociologists and behavioral psychologists. This deals with crowd behavior and mentality; and how it can be controlled, manipulated, and coached to an incredible degree. There are even some good teaching lesson plans available on line.
@Tyrell: More information is better. A such, more polls is a good thing, although the quality of some of the polls are questionable.
We were definitely worse off with just one major poll to focus on (although even back in the day there was more than just Gallup).
@Jen:
Once again, the invasion of the m#&^holes. Build that wall 🙂
@Steven L. Taylor: Thanks for the comment. It seems we are in information overload. A lot of it is good, but some is not and the challenge is sorting all through it.
@CSK: The people I saw today from Mass. were Clinton supporters who are volunteers and part of a (very, very) organized GOTV effort.
@Hal_10000:
Yes–and that’s one of the reasons why I think Sam Wang is way too overconfident with his 98% estimate. That’s why his model made such a fool of itself in the 2010 Senate race between Angle and Reid. First of all, polling upsets happen on occasion, and they happen often enough that it’s foolish to assign 98% to any race except maybe some deeply gerrymandered House districts. And given that chart at FiveThirtyEight about the level of error in past presidential elections, coupled with the unusually high level of undecided voters in this race, that would seem to make a very good case for caution. This all seems very obvious, and the only reason I can imagine we’re hearing all this FiveThirtyEight-bashing and praise for Sam Wang here is wishful thinking on the part of Democrats–which strikes me as a deeply unhealthy attitude. I still think Clinton is going to win and that a Trump win will be a disaster, but believe me, we have to prepare ourselves for anything.
@Pch101:
Yes, but they underestimated Obama’s margin of victory in a lot of those states. Just because this didn’t change the ultimate outcome doesn’t make it unimportant.
@Kylopod:
It’s a poll aggregator.
Virtually none of the 2012 polls had Obama north of 50%.
To project Obama at something greater than 50% would have required a considerable amount of tweaking, which is something more than aggregation.
@Pch101:
But the simple fact is that their 2008 estimates (well, Pollster.com’s–it hadn’t yet been bought by HuffPost) turned out more accurate than their 2012 ones. Which is really a reflection of what’s already been pointed out: for whatever reasons, the 2008 polls were on average more accurate than the 2012 ones.
@Pch101: I had not seen that page, thank you for sharing the link. My impression was formed from looking at Pollster. However, I believe your link does not disprove my thought.
I agree that ‘remain’ initially led. Then, at the end of May to the middle of June, suddenly polling strongly favored ‘leave.’ This continued right up to Cox’s death, when polling switched back to remain. Why did they switch? I’m not well informed enough to know, but I have heard that the leave campaign began to change advertising tactics to emphasize immigration. I don’t know if that’s true, I will gladly accept correction for that.
In any case, from May 22 till June 17, ‘leave led or tied 21 of 29 polls. Clearly, ‘leave’ had the momentum at that point. Then Cox was killed. From that day till the vote one week later, leave led in only 3 of 10 polls.
Given that leave won, it seems clear that the polls taken before Cox’s death picked up a real swing in opinion. That swing toward leave was masked in the polls taken after Cox’s death. My argument seems to be supported by the results of individual polls shown at the link.
I still believe that the death of Cox caused leave supporters to be less likely to respond to polls, although they had not changed either their opinion or likelihood of voting. The appearance of a swing back to Remain was caused by nonresponse bias, just as this election appears to swing wildly from a double digit lead for Clinton (when negative stories about Trump is dominating the news) to tied (when negative stories about Clinton dominate) based on the willingness of each candidate’s supporters to respond to polls, even though no one is changing their mind. As Yougov says here, when things are going badly for their candidate, people are less likely to take polls.
https://today.yougov.com/news/2016/11/01/beware-phantom-swings-why-dramatic-swings-in-the-p/
@Kari Q:
The numbers didn’t change significantly after Jo Cox’s death, so your theory doesn’t really work.
I would suggest that the lesson of Brexit polling is to disregard the polls that force an answer answer, particularly when the subject is confusing or controversial. The surprises come from undecided voters, and it is important to gauge how many of those that there are.
Just about every poll that accounted for undecided voters showed that they were in the double digits. Combine that with high turnout, and you end up with a big wild card that needs to be understood if the polls are to be analyzed properly.
At no point was Remain averaging above 50%. That should made everyone wonder about how secure that position was.
It was presumed that the undecideds would gravitate toward the status quo, but that presumption proved to be wrong; more effort should have been made to understand them. I hope that this lesson does not apply to our own election on Tuesday.
@MBunge: Until the last two days or so, I’d have rounded 538’s forecast to 70/30 Clinton, too. So these predictions aren’t as far apart as you are saying. Probabilistic forecasts are just like that.
@Pch101:
I respectfully disagree:
https://elections.huffingtonpost.com/pollster/uk-european-union-referendum/edit#!minpct=35&maxpct=55&mindate=2016-05-15&maxdate=2016-06-23&smoothing=moderate&showpoints=yes&showsplines=yes&hiddenpollsters=&hiddensubpops=&partisanship=S,P,N&parties=D,R,I,N&selected=remain-in-the-eu,leave-the-eu,undecided&fudge=0.9
@Kari Q:
As I keep telling you, remove the polls that don’t have an undecided option.
You will notice that were virtually no polls that included an undecided option that put Remain above 50%. People should be asking why that was the case.