2020 USA Presidential Election Opinion Polls

Collins must have 'closed the gap' rather dramatically late on; alternative these polls were utter muck.

Myself and some folks on here backed Collins at 4.5 a long time ago.

Nov 8, 2020 -- 9:18PM, frog1000 wrote:

Exactly. Our view is the polls have underestimated the Republican performance for 2 elections in a row in the battleground states.A goal is to work out why.

It's because polls tend to tighten in battleground states the closer it is to the election as the candidate/party that is behind chucks the kitchen sink in.

frog - are you employed in the polling industry?

We're all agreed that polls in the main have over-estimated Dems in key states right?

I know more about Texas than other states, but a more independent minded state like Maine can't be too surprising that it went more split ticket can it?

Do you think the flood of campaign money muddies the water? Collins's opponent had huge sums, but in the end Maine stuck with what they know and apparently like.

Quinnipiac had Gideon winning Maine by 12 and Biden winning Wisconsin by 17

maine senate republican 4.5 betway
Aug 1, 07:44PM
Betway have the republican in the Maine Senate at 4.5. The democrat has a slight edge but this is close to a toss up.

Here is the PM I sent out, dated August 1st.

politicspunter 08 Nov 20 22:19 
...
Myself and some folks on here backed Collins at 4.5 a long time ago.

Congratulations. However, I would suggest that you didn't elect to do so on the basis of the polls above.

maine senate republican 4.5 betway
Aug 1, 07:44PM
Betway have the republican in the Maine Senate at 4.5. The democrat has a slight edge but this is close to a toss up.

Here is the PM I sent out, dated August 1st.

Nov 8, 2020 -- 9:22PM, Hank Hill wrote:

frog - are you employed in the polling industry? We're all agreed that polls in the main have over-estimated Dems in key states right? I know more about Texas than other states, but a more independent minded state like Maine can't be too surprising that it went more split ticket can it? Do you think the flood of campaign money muddies the water? Collins's opponent had huge sums, but in the end Maine stuck with what they know and apparently like.

No I work in the Gambling industry.

We have done a fair bet of modelling in this area and keen to do more for future elections.

I think it could be good PR given the noise involved and the unobjective analysis that is often done in this field. The media seem fairly biased as do punters who support one candidate or the other.

Looking for analysis that can strip the emotion out of it.

Removing the ego from the situation is important I think.

Yes, I read your post of 22.25. I'm certainly not doubting that you placed the bet. I'm suggesting that 4.5 would not have represented value on the basis of the above polls.

Just to say that in that Maine senate race 538 had Gideon 59/41 favourite in their final model.

It's probably even harder to remove bias than in sports.

Collins should have been 2.44 in 538s final model.

In global sports markets not much bias exists. Sure England are often over bet by UK punters but that is offset. In this race the vast majority of bets were placed on Trump to win even though the polls suggested otherwise.

Nov 8, 2020 -- 9:34PM, frog1000 wrote:

In global sports markets not much bias exists. Sure England are often over bet by UK punters but that is offset. In this race the vast majority of bets were placed on Trump to win even though the polls suggested otherwise.

Schhh, for gods sake, are you trying to ruin me?

Nov 8, 2020 -- 9:38PM, politicspunter wrote:

Nov  8, 2020 -- 10:34PM, frog1000 wrote:In global sports markets not much bias exists. Sure England are often over bet by UK punters but that is offset. In this race the vast majority of bets were placed on Trump to win even though the polls suggested otherwise.Schhh, for gods sake, are you trying to ruin me?

Apologies.

I thought it was common knowledge over 70% of backer money went on Trump with bookies.

It is, but don't broadcast that it was nuts

Hills have Trump at 9.0 to win in 2024

Which Trump?

2024 Sky / Hills

Kamala Harris
       

4.5       

4.5       
                                                                   
Joe Biden
       

6       

5.5       

                                                                   
Donald Trump
       

11       

9   

                                                                       
Mike Pence
       

12       

11   

                                                                       
Nikki Haley
       

15       

15   

                                                                       
Alexandria Ocasio-Cortez
       

17       

15   

                                                                       
Ivanka Trump
       

26       

26

https://www.oddschecker.com/politics/us-politics/us-presidential-election-20...

AOC 17.0 !

Nikki Haley interesting.

As is Harris.

Biden and Donald a lay at those prices for sure.

DJT into 16.5/17 to win this one.
Biden out to 1.06.

This is what I do. I look at the candidate market roughly two years before the event and scatter some minimal stakes across a hoard of folks who are at least 300.0 and mostly nearer 1000.0. Sometimes you get lucky and most times you don't. However, even one or two "runners" means you can trade/arb for a profit.

That list looks very thin on the ground.

Surely both parties have some younger more enthusiastic people that could have a go who are not celebs?

Tucker Carlson and Dwayne Johnson both at 25/1!

politicspunter, I find it very surprising that 538 made Collins a 2.44 on the basis of the polls detailed below (in summary to avoid clogging up the thread) but I'm happy to take your word for it.

Poll Added Date/Pollster/Rating/Gideon-Collins/Margin
01 Nov Change Research C- 54-46 Gideon +8
01 Nov Change Research C- 46-42 Gideon +4
31 Oct Emerson College A- 48-46 Gideon +2
28 Oct SurveyUSA       A  51-49 Gideon +2
28 Oct SurveyUSA       A  46-45 Gideon +1
27 Oct Colby College   U  47-43 Gideon +3
[14 Oct Gideon +7; 05 Oct Gideon +1/+1; 30 Sep Gideon +8/+5; 24 Sep Gideon +4; 22 Sep Gideon +0...]

Regardless, it doesn't seem to alter the fact that even the best of these polls missed the outcome by a considerable distance.

Biden may be forced to offer to stand again to defeat threat of trump

Quite a bit of bluff, perhaps.


Does trump win rep nomination,
Or do they look to go more mid stream

Dave, modellers give extra weight to how highly they rate the pollster and how near the poll was carried out to election day. At a glance there there are some A rated pollsters giving Gideon a lead of 1 or 2 in the last week of campaigning. That's well into toss up country.

Trump should be ok to stand as Florida felons now are able to vote post conviction.

politicspunter, Whilst I was happy to take your word for it, I have now confirmed that, exactly as you stated, 538 did indeed make Gideon 'only' a 59/41 favourite over Collins (1.69/2.44). However, it doesn't alter the fact that those polls were dreadful. The polls were only marginally better in North Carolina.

In fact, 538's projections for the Senate, presumably as a result of the faulty polls on which they were based, appear to have been some way off the mark with the 2 Georgia run-offs and Alaska TBD.


Republican Seats

In fact, 538's projections for the Senate, presumably as a result of the faulty polls on which they were based, appear to have been some way off the mark with the 2 Georgia run-offs and Alaska TBD.

Georgia seat- 538 final model...

Chance of winning

Perdue 57
Ossoff 43

Popular vote estimate

Perdue 49.3%
Ossoff 49%

Current Count in seat..

Perdue 49.7%
Ossoff 47.9%

Dave, Now I would say that's fairly close to the 538 model final estimate?

politicspunter • November 8, 2020 10:22 PM GMT
Trump should be ok to stand as Florida felons now are able to vote post conviction. 

Don't they need to pay their fines though?

I wondered who would be the first to notice that

Kanye will stand his fine.

I know a decent amount about politics, but defer to some of you guys like PP and frog in polling. Is it not very difficult in a red state like Georgia that hasn't voted Dem for Pres since 92 to gauge how the other races will go in the state? I mean traditionally Americans did a lot of split tickets anyway, but with a wildcard like Trump did that make a split ticket vote harder to judge in the polls?

I know a decent amount about politics, but defer to some of you guys like PP and frog in polling. Is it not very difficult in a red state like Georgia that hasn't voted Dem for Pres since 92 to gauge how the other races will go in the state? I mean traditionally Americans did a lot of split tickets anyway, but with a wildcard like Trump did that make a split ticket vote harder to judge in the polls?

Well, red state or not, here is the 538 final President model for that state...

Chance of winning

Biden 58
Biden 42

Popular vote

Biden 50.1%
Trump 49.2%

Current count

Biden 49.48%
Trump 49.27%

Yes polticspunter, I agree that 538 seem to have done ok in regards to the above Georgia race. What I meant, by the above, was that the two Georgia seats are going to a run-off and therefore they and the Alaska race are yet to to be determined (TBD). All of the other Senate races appear to have been called on RCP (although some of these have not yet been called by all of the Cable/TV Networks).

There are, however, multiple 538 Senate race forecasts that were way off the mark. North Carolina being one of the most glaring. That is not surprising given that their projections are largely based on the guff the pollsters [sometimes] feed them. Their overall Senate control forecast appears some distance off the mark too - even in the unlikely event that the two Georgia run-offs go to the Democrats (seemingly Gross has not yet given up on Alaska but a Trump 2020-2024 2nd term would seem almost as likely as the Democrats winning that from their current position).

* Yes 50/50 would give them control, on the assumption Trump is unable to prove that the Democrats cheated in the Presidential race. However, they are unlikely to to obtain 50+ seats and, even if they did, 538 would still have been off having given the Democrats a 75% to 78% probability (Deluxe model/Lite model) of obtaining control. They gave the Republicans a 21-24% (Lite/Deluxe) probability of retaining control of the Senate.

North Carolina, the Republican Thom Tillis looks to have done very well there but he is not a huge amount ahead of the 538 forecast (50.5% democrat/47.3% republican) current count is Republican 48.7%, Democrat 47.0%. Besides, there are still about 3% to count which, as they are mail-in, could be favourable to the democrat. Alaska is impossible to know as they wait till the 10th before counting mail-ins I believe but it looks tough for the democrat to win although to be fair he is the underdog.

Control of the Senate still looks tough to call.

Just to say that in North Carolina senate race, the Libertarian candidate has done well, picking up 3.1% of the vote so far which may have hurt the democrat.

In fact, 538 had other candidates picking up 2.2% in North Carolina Senate but they currently have 4.3%, roughly double the estimate.

Sorry, I'm not clear on your 50.5/47.3 forecast figures?

538 appear to have given Cunningham (Democrat) a 68% chance of winning in their final pre-election North Carolina Senate forecast.

I'd have to argue about lib hurting the Dem pp, most libs would be more right leaning than left.

Ah, I see, the forecast vote share. Nonetheless, I think it's fair to say they were out on this one as were the polls.

They may be a little out but it's not a massive miss. Besides, as we know, if the polls that are driving their model are badly wrong, so will they be. There are still a few votes to come in though.

You could argue that the Libertarian/other candidate has taken up to 2% of the democrat vote but it's tough to say for certain.

From my time when I lived in Texas Libertarians would split around 80/20 right and left - could be different in N Carolina, but unlikely.

True. Those votes might take it marginally closer to the forecast vote share but it's highly unlikely to change the ultimate outcome. With the exception of Fox calling Arizona, the Networks have been incredibly cautious in calling races in these elections, which is probably better than when the reverse was true such as Florida in 2000.

Well Cunningham the democrat is 95K behind currently. The "others" candidates have currently 235K and they were estimated to get around 120K. It's possible that it may have made a difference to the result but obviously unclear.

On October 3, the New York Times wrote that the race had fallen into "utter mayhem" within a period of a few hours after Tillis tested positive for COVID-19 and Cunningham admitted to exchanging romantic text messages with a woman who was not his wife, damaging an image that leaned heavily on his character and military service. Days later, the woman confirmed that she had a consensual physical relationship with Cunningham in 2020. The Army Reserve started an investigation into Cunningham. The husband of the woman who confirmed having an affair with Cunningham, himself an Army veteran, called on Cunningham to drop out of the Senate race. Asked repeatedly whether he had had other extramarital affairs, Cunningham declined to answer.

Tillis's diagnosis, which came after an outbreak at a White House ceremony for Supreme Court nominee Amy Coney Barrett, temporarily threw Barrett's confirmation into jeopardy, as two Republican senators had already stated their intention to vote against.

Looks like a good battle there!

Talking of the polling industry, whilst I don't always agree the Guardian's articles, this one has my full backing:

.
https://www.theguardian.com/us-news/2020/nov/04/polling-industry-us-election-2016-debacle-repeats-itself
.

Can anyone get hold of any explosives?

Perhaps the Horse Racing forum would be a better bet? The closure of Black Market Reloaded was most inconvenient.

Nate Silver
@NateSilver538
·
8m
Estimates of the polling error in swing states, making educated guesses about what the final numbers will look like in states that aren't done counting yet.

WI Trump+8
IA T+7
FL T+6
MI T+5
TX T+5
OH T+5
NH T+4
ME T+3
PA T+3
NC T+2
NV T+2
MN T+2
VA T+2
AZ T+2
NM T+1
GA T+1
CO B+1

Interesting. There seems to something of a pattern in those states! I seem to recall that Trump was also under-estimated, in some cases by a rather large margin, by the polling companies in a lot of the battleground states in 2016.

I wonder if Nate will be quite so willing to defend the polling companies this time?

Nate isn't a pollster remember.

Trump did even better than this in Republican states where, because the result was a formality, there was next to no polling.

I know he's a modeller, not a pollster, but he does rather depend on the accuracy of their data. I seem to recall that he was rather defensive of the polling companies in 2016. In fairness, he has, at least occasionally, criticised them in the past for activities such as herding.

This is no more than a guesstimate but it appears to me that folks who were undecided in the final polling week, voted for Trump.

Florida is a little bit unique because of the Cuban American vote Trump attracted.

Regarding Florida, natioal exit polls say Biden won Hispanic/Latino voters 65-32. In Florida he only won them by 52-47.

It could be a contributory factor but I'm not sure that would justify how far out the polling was in some of those states.

I suspect that it's tempting for the polling companies to attribute the differences between their final polls and the actual election result to a 'late swing' or such like. It would rarely be accurate to do so. In their report into the polling companies performance in the UK General Election 2017, if I recall correctly, the BPC suggested that a maximum of 1 point of the amount by which the pollsters had under-estimated the Conservative vote could be attributed to a late swing to the Tories.

If the opinion pollsters did exit polls
They will know where they went wrong.



I think a lot of first time voters
must have turned out for trump, and
these are usually valued down
by pollsters. (or very irregular voters)

They seem to have got Biden vote right
As the total votes is a ball park of
what they predicted.

I never saw any predictions of
Total trump votes, despite asking
folk too.

yhtl, I'll await the report into what went wrong with the polling in 2020 with interest. However, a very lengthy analysis of the 2016 failings doesn't seem to have resulted in much, if any, of an improved performance.

They reckon they switched about
2% Biden from hilary

Biden was forecast to get
around 80 million votes.

So a +2 Biden = +4 hilary poll.

Most modellers I followed on twitter had
Biden 290-300,...

Wiki say 95% vote counted

Biden    75,677,793    trump 71,264,360
Biden     50.6%    trump. 47.6%

There is roughly 1.5m votes to count in California and about a million in New York which should slightly increase the popular vote margin but it's still a polling miss and quite a fair sized one.

Nov 9, 2020 -- 10:26PM, politicspunter wrote:

Florida is a little bit unique because of the Cuban American vote Trump attracted.

How could that be an unknown factor ? Hardly like the Cubans showed up in Florida this summer. They've been a big factor there for 50+ years.

It's a miss on head to head %

Put they seem to have got Biden votes correct.

Obviously they could have lucked that
But if they havnt lucked Biden votes
then it's trump votes they need
to explain... As in why did polls
not pick up these new votes or
if they did then why did they
not believe them enough to report them!

https://www.pewresearch.org/fact-tank/2016/11/15/unlike-other-latinos-about-half-of-cuban-voters-in-florida-backed-trump/

This is an analysis of the 2016 election. Have to wait for full details of the 2020 one.

Hillary won Miami-Dade Florida in 2016 63.68%-34.07%. Currently Biden leads there 53.3%-46%.

I also think Florida could be in
part folk
moving there, or switching vote
to holiday home like the trumps.

I think I read there's something
like 1 million 2nd homes in florida

Could easily be. If anything it's becoming a more republican state every election.

Who are the real Shy Trumpers?
Political correctness has left a cadre of white graduates unwilling to reveal their voting intention
BY ERIC KAUFMANN (Professor of Politics at Birkbeck, University of London, and author of Whiteshift: Immigration, Populism and the Future of White Majorities)

https://unherd.com/2020/11/meet-the-shy-trumpers/         

-  Some key points\;



Data from a recent Cato Institute survey shows that 88% of Trump-voting graduates compared to just 44% of Clinton-supporting graduates agreed that “The political climate these days prevents me from saying things I believe because others might find them offensive.”

As figure 1 illustrates, 45% of Republicans with degrees, compared to 23% of Democrats with degrees, said they feared that their careers could be at risk if their views became known.

Republican pollster Frank Luntz told Emily Maitlis that Trump voters were over twice as likely as Biden voters — by a 19 to 9 margin — to conceal their intended vote from others. I would expect this ratio to be considerably higher among university graduates, which would, accordingly, skew predictions the most among graduates.

According to a Pew survey on October 9, Trump was leading Biden by 21 points among white non-graduates but trailing him by 26 points among white graduates. Likewise, a Politico/ABC poll on October 11 found that ‘Trump leads by 26 points among white voters without four-year college degrees, but Biden holds a 31-point lead with white college graduates.’

The exit polls, however, show that Trump ran even among white college graduates 49-49, and even had an edge among white female graduates of 50-49! This puts pre-election surveys out by a whopping 26-31 points among white graduates. By contrast, among whites without degrees, the actual tilt in the election was 64-35, a 29-point gap, which the polls basically got right.

The exit poll I have has white female college graduate (14% of voters) 54-45 Biden and white male college graduates (17% of voters) 51-48 Trump.

The link in the document points to the same NYTimes exit poll. I do notice though:
The numbers on this page are preliminary estimates from exit polls conducted by Edison Research for the National Election Pool. These surveys interviewed voters outside of polling places or early voting sites, or by phone (to account for mail-in voters).

Results from interviews with 15,590 voters are shown below. These numbers will be updated as more data becomes available, and they will eventually be adjusted to match the actual vote count.

While exit polls offer an initial indication of how groups voted on election night, they are not perfect. The polls are not precise enough to distinguish between, say, 53 percent support or 50 percent support from a certain group. Like any survey, they are subject to sampling error, and they rely on estimates of how many people voted in each group.

They assert - white college graduates 49-49 -  now it's - White college graduate 32% of voters 48 - 51 - to Biden
They may well have written the article based in the numbers at the time, which have changed with added samples.

Look like they wrote the article before the numbers had settled (may yet settle further).

That said, the 3 point difference (48 - 51 - to Biden) is a lot less than the ABC " Biden holds a 31-point lead with white college graduates.’" and Pew 26 point lead for Biden with white college graduates.

I suppose the question we should be asking is that both the Pre-election polls and the exit-polls are "ALL POLLS"
- why suddenly are people apparently answering polling questions differently. i.e. if you were a shy-Trump voter before the election why are you not still a shy-Trump voter after the election?

Maybe people were "lying" to themselves (and the pollsters) before the election, but can't do that any more once they voted ?

Yes, this is the one that has been puzzling me too.

538's Senate Forecast:

Republican Seats	45 or less	46	47	48	49	50	51+
Democrat Seats (Inc 2 Ind)	55+	54	53	52	51	50	49 or less
Senate Majority	D	D	D	NO	NO	NO	R
Probabilty Lite	0.278	0.097	0.105	0.11	0.106	0.083	0.212
Probabilty Classic	0.198	0.102	0.122	0.136	0.127	0.097	0.204
Probability Deluxe	0.142	0.092	0.116	0.139	0.141	0.118	0.24

Notes (for the table above):

Democrat Seats - Includes the two Independent senators that caucus with the Democrats (King & Sanders)
Senate Majority - Refers to Democrat Majority (D), Republican Majority (R) and No Overall Majority (NO)
Senate Majority - Excludes independent senators that caucus with the Democrats (as applies under BF's rules)
Probability Lite/Classic/Deluxe - Refers to probability of that number of seats forecast under 538's 3 different models.
Due to rounding the forecast probability totals do not necessarily add to 1 (100%).

Lite

Classic

Deluxe

Republican Majority    ||    0.212    ||    0.204    ||    0.24    ||
No Overall Majority    ||    0.299    ||    0.36    ||    0.398    ||
Democrat Majority    ||    0.48    ||    0.422    ||    0.35    ||

Ooops, 538 overall senate forecast:

	Lite	Classic	Deluxe
Republican Majority	0.212	0.204	0.24
No Overall Majority	0.299	0.36	0.398
Democrat Majority	0.48	0.422	0.35

Senate Majority - Excludes independent senators that caucus with the Democrats (as applies under BF's rules)
Probability Lite/Classic/Deluxe - Refers to probability of that number of seats forecast under 538's 3 different models.
Due to rounding the forecast probability totals do not necessarily add to 1 (100%).

Dave, I am not sure of this. In the tables you have "Democrat Seats (Inc 2 Ind)" but in the next post you have "Excludes independent senators that caucus with the Democrats (as applies under BF's rules)" ?

Do 538 mention Betfair in their summary? Smarkets have completely different rules btw regarding the independent senators.

It does not seem to have been a very good forecast with the two Georgia Senate races yet to be determined (run-offs in January). This is hardly surprising as the polling that 538 relies on to make its projections was rather poor (or 'Not Great' according to Nate).

Dave, I am not clear on your tables. Can you have a look at my post of 2:47 pm please?

No politicspunter, 538 did not mention Betfair in their Senate forecast. However, as I have largely been betting on the Betfair markets and I'm currently posting on their forum, I considered it reasonable to base the 'Senate Majority' according to their rules. I believe that I did state this clearly in my notes.

All of the figures have been extrapolated directly from 538's site so, barring any errors on my part, it's very easy to obtain the Democrat Senate Majority or Control figures inclusive of the two independent senators; this can be done using the seat numbers or by adding by adding the No Overall Majority numbers to the Democrat Majority numbers.

Sorry, in one table you have included the independent senators but in in the next you have excluded them?

politicspunter, the Total Seat numbers are inclusive of independent senators that caucus with the Democrats (working on the assumption no additional independent senators would be elected, as transpired). The Senate Majority numbers exclude these independent senators from the Democrats. I stated this in the notes.

If your argument is that my tables are not very reader-friendly, as opposed to disputing its accuracy, I would agree with you. I designed them for my own use prior to the election and have made minimal effort to make them any more reader-friendly prior to posting them here. However, I maintain that the figures are accurate and specified in the table notes I included.

Btw, Betfair "rules" are often unique in the betting market as this election has proved yet again.

If you apply those 538 figures to let's say Smarkets eg, they make more sense.

I'll agree to differ. I again emphasise that the Democrat Seat Total include the two Independent Senators (the 'Number of Democrat Seats Betfair Market' does not!), the Senate Majority numbers exclude the Independent Senators from the Democrats (as do the Betfair relevant markets).

I'm happy to post the 538 Senate Forecasts with the Democrats inclusive of the 2 Independent Senators, although that can quite easily be calculated from the tables above or directly obtained from the 538 site.

https://projects.fivethirtyeight.com/2020-election-forecast/senate/
.

I'm sorry Dave, I have no idea why you think Betfair should even be mentioned regarding the 538 forecast. It's a completely different market with a completely different set of rules. You are not comparing like for like.

I think that the Betfair market rules should be mentioned as most of the Betfair forum readers that are betting on these markets are probably largely doing so on Betfair.

However, I do accept that the tables above could, despite my notes, provide a false impression of 538's Senate forecast. Therefore, I'm happy to post the tables again with the Democrats inclusive of the two Independent Senators. In the event of Republicans obtaining 50 exactly seats and the Democrats obtaining 50 seats (inclusive of the 2 Independents), I have assigned control of the senate to the Democrats on the assumption they are like to obtain the Presidency/Vice Presidency and thus have a casting vote in the event of a tie. Tables:

Republican Seats	45 or less	46	47	48	49	50	51+
Democrat Seats (Inc 2 Ind)	55+	54	53	52	51	50	49 or less
Senate Control (Dems Inc 2 Indpts)	D	D	D	D	D	D	R
Probabilty Lite	0.278	0.097	0.105	0.11	0.106	0.083	0.212
Probabilty Classic	0.198	0.102	0.122	0.136	0.127	0.097	0.204
Probability Deluxe	0.142	0.092	0.116	0.139	0.141	0.118	0.24

Republican Majority	0.212	0.204	0.24
Democrat Control	0.779    0.779    0.782    0.779    0.748

	Lite	Classic	Deluxe
Republican Majority	0.212	0.204	0.24
Democrat Control	0.779	0.782	0.748