Politics
How accurate/inaccurate were they?
|
By:
As far as I can make out, on an initial run through, most were in the ball park for each state. The main exception, funnily enough as in 2016, was Wisconsin.
|
|
By:
Cheers, I'll take a look at that later, but yeah there was some weird Wisconsin stuff this year. Wasn't there a 17% poll? I know there's outliers, but how do you get that?
|
|
By:
All I know is that Trump was closing the gap everywhere in the last week of campaigning. He appears to have made a lot of rapid progress in the key battlegrounds late on.
|
|
By:
Our analysis suggests the polls reported in the media were slightly different. We will be going back over this.
It will certainly be interesting to pick over all these. For instance RCP had these for Florida: https://www.realclearpolitics.com/epolls/2020/president/fl/florida_trump_vs_biden-6841.html Suggesting in October that Trump was well behind. It is the same on the more detailed wiki page https://en.wikipedia.org/wiki/Statewide_opinion_polling_for_the_2020_United_States_presidential_election#Florida I think the big thing we will be looking at is the margin of error numbers. I suspect a poll in October underestimates the margin of error for the actual results. Regardless of who one supports the data suggests the Democrats have under performed the central points of the polls for two elections now - although in many cases within the margin of error of the final polls. There needs to be honest analysis of this to correct it (or at least allow for it in prediction models) going forward. |
|
By:
There was a huge vote increase, so lots
of low percentage voters turned out. I reckon they got Biden increase, as he's right where they predicted him, overall. so it's the, extra trump votes they missed |
|
By:
I think Trump was behind in Florida in October but his support from Cuban Americans in Dade County may have been greatly underestimated. His support from other more standard groups was not dissimilar to other battleground states. You would have to look through A/B registered pollsters tables for Florida and check where the differences were from his polling figure to his actual figure once we have it.
|
|
By:
easypeasy made a strong case for Trump in Florida and he was proved correct.
|
|
By:
Looking at the post rooms in florida, there was worry postal votes were being
indexed and not delivered. Dade results were a surprise, and not near any poll I saw. |
|
By:
Just had a look at a Florida exit poll and they have Biden winning Hispanics (19% of voters) by 52-47. Now that is massive for Trump I believe.
|
|
By:
Overall it might not look too bad.
But that is when you include the more accurate ones, notably Trafalgar and Susquehenna, which the Biden backers told us were nonsense polls, or not even real polls at all. If you just go with the 'trustworthy' polls then obviously it is a complete bust. And of course RCP assessed it much better than 538 this year. |
|
By:
https://www.nytimes.com/interactive/2020/11/03/us/elections/results-florida-president.html
This is the exit poll info I was looking at. |
|
By:
|
|
By:
A lot of it is down to perception.
We had people calling up from the coastal states saying Trump had no chance. Polls were being reported that Texas was in play! In reality Biden was always way behind in Texas. Similar thing in Ohio. As more polling was done Biden was behind from October on wards. If the media is going to be honest going forward that must ignore national polls and report only on the battleground polls. People must not pick and choose the polls they state whether they are the President at a rally or CNN in a studio. |
|
By:
One they happened in 2016 and 2020 is that Trump closed the gap late. With more early voting this time that could have cost him.
|
|
By:
Texas was in play. Ohio was in play. So were many others like Georgia, Arizona, Michigan, Wisconsin, Pennsylvania and the like.
|
|
By:
Biden's team did well to counter the Trump surge by getting voters out early.
|
|
By:
Texas... Biden never led..
https://www.realclearpolitics.com/epolls/2020/president/tx/texas_trump_vs_biden-6818.html |
|
By:
https://en.wikipedia.org/wiki/Statewide_opinion_polling_for_the_2020_United_States_presidential_election#Texas
He certainly did lead in some polls. |
|
By:
Btw, Real Clear Politics is a modeller, not a pollster.
|
|
By:
Total Hokum. Trafalgar were miles out in many states.
You think they were out more than the 538 average in more states ?  |
|
By:
Which counties in Texas were you thinking would flip enough to turn it?
|
|
By:
|
|
By:
So do all modellers.
|
|
By:
I know it is a modeller, which is why I compared it to 538.
Nate Silver had a go at RCP on the eve of the election for including Trafalgar. He was very smug at the time. |
|
By:
|
|
By:
The poll results for each state are in the header at the top of the page.
|
|
By:
|
|
By:
day before election
Nate Cohn, the New York Times’ polling expert, described RealClearPolitics as “a small casualty of this election”, and said they were no longer credible. A second polling expert said they were unabashedly pro-Republican. “They've been doing this all year. Arbitrary date ranges and arbitrary exclusion of some polls, even within range, which just happen to be unfavorable to the GOP,” he said. Nate Silver, founder of polling aggregator FiveThirtyEight, agreed. “Endorse this, even though there are many things I like about RCP,” he said. |
|
By:
These folks are modellers not pollsters. All of them weight polls differently. If the polls are wrong to start off with, they are on a hiding to nothing.
|
|
By:
If you say Trump is going to win a state by 2 points, the other guy says Biden is going to win it by 9 and Biden wins it by 1....
Are you seriously saying the 9pt prediction is the better call  One is 3pts out and the other has missed by 8. |
|
By:
I think we have to be honest about it.
The averages have Trump lower than actually happened in a number of cases. If we look at Pennsylvania the polling average had Biden up around 5%. It appears that the gap is going to be less or around 1%. For a betting point of view something is not quite right about these poll numbers. |
|
By:
Hold on, the modeller is working by what the polls are. If the polls are wrong, then obviously the modeller is also going to be wrong.
|
|
By:
I feel it is useful to have an honest conversation about if/why this is the case and if it will continue for future elections.
|
|
By:
|
|
By:
The modeller decides which polls to include and how much weight to give them etc, hence 538 and RCP having different numbers despite having all the same info
|
|
By:
I agree.
|
|
By:
it was clear Trump was rapidly closing the gap in the last week
Which would be another factor Biden backers had kicked into touch as 'everyone had made their minds up' |
|
By:
|
|
By:
|