Poll Diving – part 7
Disclosure. A few weeks ago, an election scenario intrigued me and then I noticed that I wasn't alone in this consideration. And if it came to pass would there be any way for the left to punch above its weight. However, I first wanted to assess the possibility of such an outcome. My first pass on all the key states suggested that it was possible, but within a few days the polls suggested implausible and a few days later not worth considering. Thus, my first call on this presidential election is that we won't face an electoral college tie. That's good because given the current generalized stress level in this country, it would have been really ugly situation.
As of today, 30 October 2020, there remains indecision among enough voters in many states to accurately project where they'll end up. The election pros seem to be hedging on more and more states. Perhaps they're overcompensating for their 2016 flubs. OTOH, it never hurts to ask if there's something brewing that can't yet be seen in the numbers and if there is what is the most probable direction and is it regional or national.
The softening for Trump and shift to Biden – and it appears to me to have been in that order – has been slow. More drip, drip, drip than whoosh. It only gathered some steam in MI, PA, and WI after Biden broke through a barrier. However, the gap between Clinton and Trump was the least in those states and therefore, fewer drips were required.
The pattern isn't different in VA that Clinton easily carried by 5.3% points. Then it was Clinton 49.7 and Trump 44.4% with 4.5% third party. Now it's Biden 51-53% and Trump 39-41% with 3% third party. Will VA 'bleed' or 'leak' into NC? So far Trump's best recent poll number is off 2% points from 2016 and Biden's worst poll number is up 2% from Clinton's final.
There is more disparity in the FL polls than NC. However, Trump's range is worse and Biden's is better. Only one percentage point separated Trump and Clinton, but it was a solid point that could be read before the election. Still, Biden carrying FL seemed unlikely a few weeks ago. But it (whatever it is) is also happening in GA that Trump carried by four points in 2016. Trump's poll numbers in GA are worse than in FL, but Biden's aren't quite as good. Georgia flipping would be as big a surprise as MI in 2016. And contrary to my earlier assessment, Ossoff could win the senate seat.
Trump's wins in IA, OH, and TX were more solid, nine, eight and nine points respectively. So, the drips haven't added up to as much as in other states. Trump lead in most of the OH and TX polls even it it's not by that much. IA is closer to being tied. Factors that have shifted the polls in WI aren't absent from IA and could break the tie in the next few days.
AZ and NC, both favored Trump by four percentage points, look much alike in polling numbers. Both also have a sitting GOP Senator that is being effectively challenged. Otherwise these two states are very different. At the risk of getting this backwards or completely wrong, I'm still seeing a Biden win in AZ and NC remains too-close-for me to call.
In conclusion, on my cheat sheet, I only have ME CD2 and three states in play: IA, GA, and NC. That would put the EV at 318 Biden, 182 Trump, and 38 undecided.
Comments
Iowa and Covid
Iowa has a large older population and is experiencing a high number of deaths and infections. I believe this will put Iowa in Biden's column. Also, Iowa voted for Barack Obama twice. Polls are really close in Iowa. I never try to predict Southern states for Dems, although these two are less "Solid South".
I can't claim to have a good read
on Iowa. Can see that it's #4 on the cases/M list but its deaths/M and CFR are far down on the list. Reports from the ground in WI are the covid is a major factor there, but in 2016 Trump's margin was tiny compared to his margin in IA. So, in WI based on current average poll numbers he's down by 3.3% and Biden is up by 3.8% (compared to Clinton) and above 50% which gives Biden a 6.4% lead. In IA, Trump is down by 5% and Biden is up by 5% but both seem stuck at 46% or 47%. The WI polls are more numerous and current than those in IA. So, maybe everyone is waiting for DMR to weigh in.
Des Moines Register released
its final poll. They apparently found all the Trump voters that other pollsters missed. All 1.5% points of them. What DMR couldn't find that other pollster did find were 6% points for Biden. So, DMR has it as Trump 48% and Biden 41%. IOW, Biden is doing 0.7% worse than Clinton and 2020 Trump is doing 3.2% worse than 2016 Trump.
This is very odd -- even in AL where Trump is down by 3.1% points and Biden is up by 4.6% points. Not that either matter since Trump's winning margin was so large in 2016. But the pattern of Trump down and Biden up is consistent in red and blue states.
FFS
The DMR??? The same propaganda mill that withheld their final primary poll which NO DOUBT showed Bernie way ahead.
Not even interested in anything they have to say----now or forever.
NYCVG
The same.
DMR has long enjoyed a good rep in IA. My observation is that their polls skew conservative and they aren't above fudging a bit with too high a number of undecideds.
That said, they may not be off by as much with this latest poll as a first glance would suggest. Will attempt to explain in a new essay.
Michael Moore warns, "Don't believe these polls, first of all."
Your analysis looks good to me. Even Allan Lichtman agrees with your outcome.
But what if the people are lying to the pollsters like they did in 2016? Moore is convinced that is the case. Would it stay within the margin of error?
.
On the other hand, maybe Trump voters didn't lie this time. And maybe basing your campaign on the fact that you are not the other guy is all it takes.
Cities around the country are boarding up their windows. This is one of those times I wouldn't want a Biden yard sign.
"Don't count the underdog out. Upset wins are surprising but
not impossible."
So says 538.com, where as of today they project Trump has a 10% chance of winning the election.
I think it is worth remembering that statistical projections about who is most likely to win are not the same as predictions of who will win.
Long-shots DO sometimes win. Even the best projections don't say Biden has a lock on it, just that he is favored to win according to the data currently available.
So whatever happens, they cannot actually be wrong. That's a pretty slick trick.
The last upset win was in 1948.
And that was only because Gallup didn't know how to poll voters.
Too close to call based on EV isn't an upset except in the minds of the loser who expect to coast to victory.
Vandalism ...or worse.
.
I live in the Houston area. Two things I have not seen during this election cycle:
1) A Biden bumper sticker.
2) A Biden yard sign.
Actually, while I've seen Trump bumper stickers and yard signs, I've seen very few of them, especially compared to 2016. Perhaps more will appear in the next couple of days, but I kinda doubt it.
Not only have I not seen any Biden signs/stickers, I have not seen these for any Democrat, save a handful of small yard signs for a local state representative race.
While I haven't been able to come up with a decent explanation for the sparsity of Trump signs, I have for the absence of Democratic signs. I've talked to neighbors, etc., who know my political leanings as to why they don't have signs posted - when they usually do. The answers were all the same. They are afraid of, mainly, property damage should they display a Biden sign. A few even fear personal attack.
Two of these folks (might have been three, can't remember...) expressed to me that they had suffered property damage in 2016 that they attributed to the display of signs. In these cases, it was damage to vehicles parked in public parking lots. Each had to take their vehicles to body shops to repair paint finishes. One of these had obscenities scratched into the paint, and a second had the word "Trump" scratched into the side of the truck. In both cases, the political stickers had been defaced at the same time.
With the level of animosity as high as it is, not many are wanting to take the risks that displaying a political sign entails. I don't blame them.
Of course, the unknown factor is whether there will be organized violence on or after election day. While I have seen some calls for peaceful protest should a fair election be thwarted, the threats of voter intimidation by heavily armed right-wing paramilitary groups at polling places is so well documented - with tacit encouragement by Trump himself - that it needs no link.
Support for Biden is a miliimeter thick.
Someone should have made NOT TRUMP signs and stickers because that's what is winning.
The Trumpsters have been encouraged to be violent by their two-bit carnival barker leader who still thinks he won in 2016 and is the "favorite President."
Moore is wrong. Voters didn't lie in 2016.
Moore to his credit did forewarn about MI in 2016 based on what he could see in his neighborhood. However, his view was biased and could never have been sufficient to detect such a narrow win. What was off in 2016 was 1) polling in MI and WI was sparse and not of high quality and 2) the DP has never acknowledged that Hillary (nor Bill for that matter) wasn't well regarded by the general public.
Hillary was going for a landslide win. After all outside the deep red states, nobody was going to vote for her chosen "pied piper" loser opponent. Thus, she put her time and money into FL and NC. And eastern PA where she would pick up two suburban women for every GOP voter in the west. Didn't help in PA that the party jammed a weak female candidate through the Senate primary to challenge a weak incumbent GOP Senator. She did hold a couple of rallies in OH late in the race. Another total waste of effort because as with IA it was gone before September.
Two weeks out from election day, Trump was only short one state, So, even overlooking MI and WI, polls did reveal that PA could go either way. And if it went to Trump it was a wrap.
I don't see where
in the interview that Moore specifically says Trump voters lied to pollsters
in 2016 (and I just went back and checked a second time)
How can he be wrong about something he *didn't say*??
Given media hysteria, TDS and frequent cases of outright violence directed at Trump supporters it is quite understandable that at least some percentage of people who actually plan on voting for Trump will decline to state so openly. Moore is hardly the only one saying this, the real debate is over how significant a phenomena it is.
It was I who said people lied to the pollsters.
Here's why:
There was a rise in populism that year. Outsiders were in. Sanders, Trump. People were very energized by this new sense of independence and power over the status quo.
This was also happening in Europe. Brexit was a revolution of sorts. The People were demanding to break from the EU. They could not see the consequences or they didn't care. They were taking a stand against many things. There was violence and anger over it. the Brexiters were reviled as ignorant or dumb. The vote took place before the US Presidential election. Poll after poll said that Brexit would fail, but Brexit won the day by a good margin.
The people had deceived the pollsters because they didn't want to be treated like Trumpsters. In the US, support for Trump was tearing up families, causing chaos in the workplace, breaking up friendships. It resulted in prejudice and shunning in communities and social organizations. Trump supporters were bullied and belittled. Again poll after poll had Hillary ahead. The upset over Trump winning caused even more chaos. The nation's expectations were based on the polls, and people were in shock.
After that, there were many discussions about people misleading the pollsters in both nations. Pollsters talked about ways to ensure that their subjects were more reliable. Downticket polling was much more aligned with the voting results. The lying seemed deliberate and methodical, perhaps as a reaction to the persistent bullying and name calling.
In highly contested elections, will this sort of behavior need to be factored into polling results? We'll know the answer to that, shortly.
In case you haven't noticed,
I'm not picking and choosing among pollsters for my state projections. I'm using the RCP average of the latest polls and looking at trends and outliers and toss in anything else I may know about the state that could move it in either direction. A good pollster that gets a crap sample will always perform worse than a crap pollster that gets lucky with a good sample.
As for your precious Trafalgar in NV 2016, they were out in left field.
Actual:
Clinton: 47.9%
Trump: 45.5%
Johnson: 3.3%
none of the above: 2.5%
Trafalgar (final): MOE 3.02
Clinton: 45%
Trump: 50%
Johnson: 3%
Stein: 1%
Well outside its MOE on Trump and barely within it on Clinton. Emerson was the only one that got it close to right: HRC 47%, Trump 46%, Johnson 4%.
In PA Trafalgar got the order right and well within the MOE. However, as all the other pollsters had it with HRC in the lead (most within their MOE to the final results), undecideds remained far to high for me and therefore, I viewed it as too-close-to call or 'flip a coin.'
NC and NH, in both states, Trafalgar had the opposite problem of other pollsters; they couldn't find the HRC voters and generally found too many Trump voters. In NH, Emerson again got the shape right and close to the mark. But like PA, getting close but the opposite isn't a fail.
Appears to me that they have a sampling and interpretation bias. In 2016 that gave them an advantage in one too close to call state (PA) and a disadvantage in NV and NH.
Odds are better that GA flips than MI going with Trump. (and while worth watching, I don't have GA in a toss-up position.)
Predictive Technologies
As I said earlier, I favor your analysis. It's logical and you are clearly knowledgeable about the stats and the past. I don't really know the contorted rules and mutant processes of US elections, My interest is in the final outcome and in methods that can predict those outcomes accurately. So, your narrative, which looks at polls but is mindful of the political dynamics in play is particularly informative to me. I mentioned Lichtman's predictive model based on historical patterns earlier. And then there are tracking polls and polling averages whipped into a predictive smoothie. I'd call Trafalgar's system psychological or trust polling.
.
The video interview that @Blue Republic posted goes deeply into his weaknesses. But like all predictors, he just brushes it aside and presses forward.
.
Yes, I noticed that. It's a Best Practice.
Just notice today that there's a
Wikipedia entry on 2016 polls Scrolling through the 2016 FL polls there made we wonder for a moment how I became so confident in calling it for Trump about a month before the election and never reversed. Until I recalled that my earliest 2016 call was "watch the Senate races." Rubio was so far out in front that a disprortionate share of the undecideds would break in favor of Trump. (The earlier polls also had more in the 3rd party column than is normal for FL.)
I wish I had such a handy dandy guide for 2020 as it projected every state with a Senate race (none that year in MI), but I don't. So, the poll numbers are guiding me somewhat more. The pattern that has developed in red and blue states is less for Trump and more for Biden compared to 2016. There's some correlation between the poll swings from 2016 actual. The larger Trump's 2016 margin the larger the swing, but in those states Trump continues with a lead and mostly a strong lead. The three exceptions are MI, PA, VA, and WI. Clinton easily won VA and the swing is 6.4% points further from Trump. The swings in MI and WI are over 7% points and PA is a more modest 4%.
The above may be a function of age demographics. I'm going to play around a bit with this.
There is a hot new item going virus
The young, smart movers are giving away trump tee shirts with a bulls eye on the back.
Guaranteed to confuse the militia. Exponentially increases your chances of not being shot.
While voting.
"Garbage" pollster strikes back - CNN interview
Marie has dismissed Trafalgar Group's polling as "garbage" (Dive #5) and Nate Silver and some other pollsters have been about as critical.
Yet Trafalgar accurately predicted not only the correct outcome in 2016 but also nailed the exact EC outcome (although they were off on the outcomes in three states, they predicted a Trump loss in WI, and wins in NH and NV which exactly cancelled each other out).
They performed well in 2018 senate races also.
In any case, they are currently projecting Trump wins in AZ, TX, FL, GA, NC, OH, IA and MI.
They have PA, MN, NV, WI and one Maine CD rated tossups.
If they are correct about their win predictions then Biden could take all the latter and still lose in the EC.
They could be wrong (or *way* wrong) then again, if they are right (again)...
Here is their chief pollster Robert Cahaly explaining/defending their methodology and record in this CNN interview:
That was an interesting interview!
Thanks for posting it. I cannot figure out how they give the subjects that they call a sense of 'anonymity,' and he didn't explain it. But that would elicit more truthful answers by my reckoning.
What interested me most was that they were gauging the psychological stressors in the social environment and anticipating certain human reactions like lying and defensiveness. That's a great way to get a gut feeling about what's really going to happen, but they must fall back on the real responses, just like any other poll. They don't appear be modifying numbers to compensate for the mood. That means everything depends on how skilled they are at getting truthful answers from a wary public.
Like the interviewer, I too, questioned his republican bias. But he gave a good answer to that: "I only make money if I am right."