Releasing "Weird" Results is the Sign of an Honest Pollster

Mish

An ABC poll has Biden up by 17 points in Wisconsin. That's not a typo.

How Polling Works

This is how polling works. You conduct the poll, make sure your methods are sound, and the result is the result.

Releasing "weird" results is the sign of an honest pollster.

But what about consistently weird like Rasmussen and Trafalgar? 

What to Make of Wisconsin?

Please consider What To Make Of That New Wisconsin Poll That Has Biden Way Ahead

Once in a blue moon, you see a poll that makes you blink twice to make sure you’re not seeing things. This morning’s ABC News/The Washington Post survey of Wisconsin was just such a poll. It showed Joe Biden 17 points (not a typo) ahead of President Trump, 57 percent to 40 percent, among likely voters. To put it mildly, this is a stunning margin in what is supposed to be one of the most competitive swing states in the country — a place that Trump carried by less than 1 percentage point in 2016.

And this is not an easy poll to disregard. ABC News/Washington Post adheres to what we consider the gold-standard methodology (meaning they use live phone interviewers, call cell phones as well as landlines and participate in the American Association for Public Opinion Research’s Transparency Initiative or the Roper Center for Public Opinion Research archive) and earns an A+ grade in FiveThirtyEight’s pollster ratings database.1 The poll single-handedly increased Biden’s lead in our polling average of Wisconsin from 7.1 points to 9.0 points, and it is currently the most influential poll in our forecast of Wisconsin, where Biden’s chances of winning the state have reached a new high as a result — 93 in 100.

Not Isolated Actually

Nate Silver goes on to explain recent trends. This move in Wisconsin is not isolated.

  • RABA Research was out with a poll yesterday showing Biden at 50 percent and Trump at 46 percent in Iowa. Last week, the A+ pollsters Monmouth University and Siena College/The New York Times Upshot found similar margins. Iowa, remember, is a fairly red state that Trump carried by 9 points in 2016. And for the first time, our forecast now gives Biden a better chance than Trump of winning the state, although it’s still basically a coin flip (Biden’s odds are 51 in 100).
  • Yesterday, Gravis Marketing released a survey of Minnesota in which Biden led Trump by 14 points. Biden’s chances of winning Minnesota have now reached an all-time high of 94 in 100.
  • On Sunday, we also got a Gravis poll of Michigan that gave Biden a 13-point lead. That’s on top of last week’s Fox News poll of Michigan showing Biden 12 points ahead. And this morning, ABC News/Washington Post also released a Michigan poll giving Biden a smaller 7-point lead. Overall, our forecast gives Biden a 94 in 100 chance of winning the Wolverine State.

Late Momentum

This surge in momentum for Biden was expected in this corner. 

We see it in Georgia as well where A+ rated pollster Monmouth University has Biden +3. 

My Fearless Election Forecast and Range of Outcomes

My Fearless Election Forecast October 13

On October 13, I posted My Fearless Election Forecast and Range of Outcomes

I moved Iowa, Ohio, and Georgia into Biden's camp expecting momentum to go in Biden's favor.

I expect all three to break for Biden unless he makes a major gaffe in the next debate, catches Covid, or some other unexpected event happens. ... Even Texas is in play for Biden.  

Spotlight Iowa

On October 21 I commented Spotlight Iowa: Monmouth University Has Biden 4 Points Ahead.

I expected the polls to change for Biden, and Iowa just did.

Silver waits for the polls, I will take a reasonable shot when I see one. I expected the Monmouth poll and there it is. 

Confidence Level 

Q: How confident am I of Iowa, Ohio, and Georgia? 

A: Not very and neither is Silver.  

Silver has Iowa 54-46 Trump, Georgia as 52-48 Trump, and Ohio as 50-50. 

I have all three as 55-45 Biden

Despite the Monmouth poll, my odds did not change because I already expected such a result by a high quality pollster.

If we see further momentum in Biden's favor, expect to see Nate Silver at something like 55-45 Biden for Iowa, Georgia, and Ohio as well.

Ohio is not going as planned but Georgia and Iowa are.

Silver is now 53-47 on Georgia and 50-50 on Iowa up from 54-46 Trump. 

Silver has Ohio 60-40 Trump. I will wait for more polls and reassess Ohio.

Modeling

"Given Biden's strong polling in Iowa lately I suspect that our model is lowballing him in NE-2, in fact. Getting another nonpartisan poll there before the election would be one of the more helpful things right now."

Reasonable Shots

Some people wait for polls others are willing to go out on a limb and take a stab at anticipating them.

Let me reiterate my October 21 comment: "Silver waits for the polls, I will take a reasonable shot when I see one."

A couple of friends questioned my statement.

But if all I do is wait for and parrot Silver, I may as well be Silver. 

I strongly disagreed with Silver in 2016 regarding Trump's odds of winning. 

Flashback November 2, 2016

Mish in 2016: Assessing Trump’s Chances With Six Days Remaining

Without a doubt, the decision by FBI director James Comey to announce the discovery of new emails related to the Hillary Clinton has seriously damaged Hillary’s election chances.

This may hand Trump the victory.

I disagree with Silver on Nevada and North Carolina. Earlier today Silver flipped on Florida, and before his flip, I disagreed with Silver on Florida.

Mentally place those states in Trump’s column, and we are pretty much where we were heading into the first debate: with Trump needing one more state.

Momentum

In the end, I expected Trump to fall one state short. But in real time I accurately assessed momentum swung to Trump and that Comey was a huge factor.

 When momentum breaks hard, it is a mistake to average polls especially when there is a dearth of them. Stale polls play too big a role.

In 2016, there were not enough quality state level polls and they undersampled low-education white voters as well.

Then because of a dearth of polls at the state level, the Comey effect did not show in the states that mattered.

Eight Reasons This is Not 2016

Believe what you want but there are "At Least" Eight Reasons This is Not 2016.

  1. There are far fewer undecided voters this year. For discussion please see Who are the Undecided Voters and How Many Remain?
  2. There is no last minute Comey Bombshell (sorry Trump fans, the Hunter Biden Story is Irrelevant)
  3. Hillary was 2-3% ahead nationally in 2016 but Biden is 8-12% ahead.
  4. There are more state polls this time.
  5. In 2016, the state pollsters underweighted uneducated white voters.
  6. It's likely that pollsters overcompensated for the education gap this time thereby underestimating Biden's lead.
  7. Hillary was more despised than Trump. Now, Trump is far more despised than Biden.
  8. The massive amount of early voting is Democratic. These are not "likely voters". They are 100% guaranteed voters. The pollsters are underweighting "guaranteed" voters as "likely" voters. In 2016 a big chunk of these voters sat out the election.

Bonus Reasons

Those who expect a "Bonus Reason" I just happen to have one: Covid Records Shattered In The US and Europe

Mish

Comments (63)
No. 1-21
ToInfinityandBeyond
ToInfinityandBeyond

Maybe the Republicans should have spent more time working on a better Covid response instead of rushing through their Supreme Court nomination.

PecuniaNonOlet
PecuniaNonOlet

I am not sure why Trumpers are shocked by these polls and call them "fake news" given that even red states such as georgia, texas and arizona are now in toss up or leaning Biden. Turnout has been huge and that doesn't bode well for voter suppressor Republicans. People aren't standing in lines for hours to vote for more Trump.

Even Trump warned about this a few months ago.

“The things they had in there were crazy. They had things, levels of voting that if you’d ever agreed to it, you’d never have a Republican elected in this country again,” Trump said during an appearance on Fox & Friends.

randocalrissian
randocalrissian

Nice work on campaign analysis this cycle, Mish. Impressed. If the polling is as accurate as it historically has been, excluding 2016, then you're going to come out of this smelling like roses.

My new worst case for Biden has Trump winning PA, GA, FL, and NC, and still losing 275-262 with Biden carrying MN/WI/MI, IA, NV, AZ. But if Trump also flips NV and one NE single electoral vote, it could be 269-269.

BUT - 304 right now feels like the floor for Biden.

Eddie_T
Eddie_T

Yeah...I attribute the Wisconsin shift to COVID...seems like a no-brainer on that one.

It seems more and more like PA is going to be very critical...it does not make me happy that they might be a couple days behind at least...in announcing the winner. That sounds like the best we can hope for.

9 million registered voters in PA.

Slightly over 3 million mail-in ballots requested total.

Nearly 2 million returned so far, 1.35 million were from registered Democrats.... 2X the number of Republicans and "other", to date.

It would take a real landslide to make the results knowable on election night. I guess it's possible, looking at the high numbers of democrats who voted early.....maybe?

But that is also possibly a red herring.

Rocky Raccoon
Rocky Raccoon

It just seems to me there are many strong efforts to make sure Trump is defeated next week. I don't think I have ever seen traditional Democrats and Republicans working together on this level in the past for a common goal all the while Trump has done so little to expand is WWE base.

Roger_Ramjet
Roger_Ramjet

Yep, momentum seems to be shifting to Biden. In fact, Trump's recent rally in Omaha left supporters cold, literally.

numike
numike

White House admits report that listed 'ending' COVID pandemic as Trump accomplishment was 'poorly worded'

Sechel
Sechel

It's the Pandemic, stupid

njbr
njbr

How much can you deny what is right in front of people who have to make a decision?

Turn the corner 4 times, and you're back in the same spot you were in the spring.

Talk about your imaginary 2 Nobel prizes and people begin to wonder...."They didn’t cover the two Nobel Prizes. I got two in one week, did you ever hear that? For two totally unrelated things. And my only complaint is I should have gotten seven or eight, because, if you knew, some of the other things I’ve done are much better. But the fake news didn’t even put it on. And somebody had a show where they said the amount of time devoted to Donald Trump’s Nobel Peace Prize—two of ‘em—zero, on the networks. Zero. "...

He may be a good show on the road, but just like the Rolling Stones concert, it ain't your real life and it doesn't pay the bills or make you well.

Like it's said..."That boy aint' right in the head."

njbr
njbr

We're about to find out what an authoritarian narcissist will do when he receives a vigorous rejection via a popularity poll...

TimeToTest
TimeToTest

Just saw where Alex Jones was on the Joe Rohan Experience. #CancelSpotify was trending afterwards.

I am sure people are very forthcoming about their political views to pollsters at this point.

Jackula
Jackula

I agree with Mish on the Covid point, Trump's inept handling of the pandemic with Covid cases beginning to explode is gonna turn this election into a rout of the Republicans.

LawrenceBird
LawrenceBird

416-122

CzarChasm Reigns
CzarChasm Reigns

Timing is everything...

(CNN) On Monday, the Supreme Court issued a ruling prohibiting Wisconsin from counting mail-in ballots that arrive after Election Day. In a concurring opinion, Justice Brett Kavanaugh raised troubling concerns about whether he'll be independent of the man who named him to the court.
Kavanaugh wrote, "If the apparent winner the morning after the election ends up losing due to late-arriving ballots, charges of a rigged election could explode."
(Kavanaugh quotes from a law review article by Professor Richard Pildes, a current CNN Contributor, whose words were actually in support of extending mail-in ballot deadlines; it is one thing for a professor to theorize in a scholarly piece, and quite another for a justice to take that language out of context and to use it in a decision of the Supreme Court).

LawrenceBird
LawrenceBird

Pollsters have to earn a living between election cycles. Being horribly wrong is not a way to get clients.

Carl_R
Carl_R

Another shocking poll out today showed Nebraska at +5 Trump. Normally a Republican would win by 20 points or so. In the Nebraska Third District, Trump is probably +60%. That means that Biden will win the Second District easily, and that he will probably carry the first district as well. That would make the Nebraska split 3 for Trump and 2 for Biden, moving another point to Biden. It's still only 1 point, but it is shocking.

I agree with Eddie - Covid is the most likely explanation for the shift in the upper Midwest to Biden.

AshH
AshH

I see the betting markets (PredictIt) have the odds of a Biden win at about 60 to 65%.

BUT, I've never seen so many prominent members of a party support the opposing candidate (Lincoln Project, and traditionally conservative publications) like today. I've never seen the country in worse shape, from struggling with a pandemic, to the economy being in terrible shape for a large segment of the country (except the stock market has become decoupled from the economy), to trade wars with everyone, to racial injustice & protests. It feels like this election should be a blowout, if it weren't for Trump's cult of personality and the FNC's unwaivering loyalty.

T. Cruit
T. Cruit

I just can't get past the enthusiasm gap between Trump and Biden. I watched a Biden Rally and it felt like attending a funeral of someone nobody cares about. Watching a Trump rally is uplifting and makes me think better days lie ahead for the US. When I seek out the videos that show the crowd; the only logical conclusion is that pollsters are lying. Yesterday I watched a Trump parade one Rodeo Dr in California and thousands participated. Also it was white, black, Latino, Asian, Jewish, and others.

jibob
jibob

Man you guys are dumb. My "fearless prediction" is that trump wins. Polls are garbage. You even think that Biden's lead is being underestimated ?!!! Oh, Ok, you're smart. LOL

Hayseed
Hayseed

I'm sure Minneapolis, St. Paul, and Duluth, will vote mostly for Biden. But I'm not sure about the suburbs. I have driven around the northern and western burbs a lot this Fall, and the signs are ten to one Trump. I doubt many of them are for Trump because of how great he is. It's more of an F you to Washington, the Media, and Celebrities. Why couldn't the Dems allow someone better onto the ticket?

Has anyone released the scripts the pollsters are using? What I read, and what I see don't match.

Carl_R
Carl_R

This election will be an interesting test for pollsters. Polling used to be accurate, but anymore, it is much, much more difficult for a variety of reasons. First, people often don't answer calls from unknown numbers, which is a systematic bias. Next, with the polarization, an increasing number of people are unwilling to give their genuine opinions. Finally, the turnout varies widely from year to year to year. There are also polls taken over the internet, and they have different problems, with self-selection bias, and the like.

Pollsters try to adjust for these things, but how effective their adjustments are is unknown. In 2016, the polls were fairly close, but a shift of only 2% was enough to change the results in a few key states, which in turn changed the winner. There is every reason to believe that the difficulties in obtaining an accurate poll have become worse, not better, and therefor the polls may be off by even more this cycle. But, in which direction?

There are reasons to believe that the polls have a bias in favor of Biden, and that Trump will outperform his polls. The primary one is that Trump supporters are less likely to state their opinion, or to put up yard signs for fear of retribution. I don't happen to be a Trump supporter, but if I was, I certainly would not want it known as I might lose customers over it. As a secondary reason why Trump might outperform his polls is that his supporters tend to turn out, and they are committed to do so this time, too.

There are also reasons to believe that the polls have a bias in favor of Trump, and that Biden will outperform his polls. The key one is that, just as Trump is able to motivate his followers to turn out, he is an incredibly polarizing person, and is equally effective in motivating people who dislike him to turn out to vote. Trump won in 2016 partly because, while his voters turned out, Hillary's voters had a lower turnout. I don't believe that will be repeated. Based on the incredible number of early votes already cast, it would appear that this year will have a record turnout.

One final thing that can skew the real results from polls is a late shift. In 2016, the Comey announcement probably caused a shift of a couple points at the last minute. Thus, perhaps the polls were actually correct, but it was just the late shift that changed the outcome?

My guess? That's a tough one. My guess is that the factors able will be offsetting. On the one hand, there will be a hidden wave of Trump supporters who have camouflaged their intentions, so the Trump vote will be higher than expected. On the other hand, there will be a much higher than expected turnout on the Democrat side, too. I predict that the two effects will net out, and the results will end up close to the actual polls.


Global Politics

FEATURED
COMMUNITY