Fivethirtyeight has Biden:Trump at 85:15

[wing2000]

Texas is going to be very interesting. Just under 2 weeks left until election day and they've already gotten roughly 2/3rds of the total votes in 2016.

Yes, early voting is strong. in Arizona, 1/3 of the 3 million early ballots mailed out 2 weeks ago have already been returned and processed.

 

[trunks2k]

remember 2016...polls are meaningless.

*sigh* Do we need to go through all this again?

2016 polls weren't wrong, they were mostly within MoE with a large number of undecides. People analyzing them just assumed the odd chance of undecides breaking hard to Trump wouldn't happen, there also was an under-weighting of uneducated white males (that's the real criticism of polling from 2016).

This year ain't the same thing. Pollsters have added weight to uneducated white males (which helps out Trump in polls), there are waaaaay less undecided voters, and the the polling is much more stable, indicating that people aren't changing their minds. All indications are that Republicans this year are falling into the trap that the Democrats did in 2016 and Republicans did in 2012 - not looking at what polls are really saying.

 

[Desk trauma]

*sigh* Do we need to go through all this again?

No, it’s a waste of effort at this point.

 

[98cwitr]

*sigh* Do we need to go through all this again?

2016 polls weren't wrong, they were mostly within MoE with a large number of undecides. People analyzing them just assumed the odd chance of undecides breaking hard to Trump wouldn't happen, there also was an under-weighting of uneducated white males (that's the real criticism of polling from 2016).

This year ain't the same thing. Pollsters have added weight to uneducated white males (which helps out Trump in polls), there are waaaaay less undecided voters, and the the polling is much more stable, indicating that people aren't changing their minds. All indications are that Republicans this year are falling into the trap that the Democrats did in 2016 and Republicans did in 2012 - not looking at what polls are really saying.

If I didn't know better, you're calling white male conservatives "uneducated"

The fact of the matter is that the polls were extremely errant, misguided and false, so why trust them for 2020?

 

[trunks2k]

If I didn't know better, you're calling white male conservatives "uneducated"

Then you need to work on your reading comprehension. I'm saying that white uneducated males tend to vote for Trump. It's a demographic that he does very well in and they showed up in larger numbers than expected in 2016, hence their underweighting in 2016 polls and why polls are giving them more weight this year. That's not really in dispute.

The fact of the matter is that the polls were extremely errant, misguided and false, so why trust them for 2020?

They weren't errant, misguided, and false. National polls were pretty spot-on. State polls were, generally, within MoE. What analysts got wrong was primarily underestimating how strongly undecided voters would break for Trump. That big chunk of undecided voters don't exist this year.

 

[98cwitr]

Then you need to work on your reading comprehension. I'm saying that white uneducated males tend to vote for Trump. It's a demographic that he does very well in and they showed up in larger numbers than expected in 2016, hence their underweighting in 2016 polls and why polls are giving them more weight this year. That's not really in dispute.

Do you have data to back this claim?

They weren't errant, misguided, and false. National polls were pretty spot-on. State polls were, generally, within MoE.

They had Hillary at >90% chance...she lost. If that's not errant idk what is.

 

[trunks2k]

They had Hillary at >90% chance...she lost. If that's not errant idk what is.

The POLLS did not give her a 90% chance. ANALYSTS did. The POLLS were pretty correct, as I said - nationally they were correct, state wise they were mostly within MoE. Analysts just didn't understand what they were looking at.

 

[wing2000]

The fact of the matter is that the polls were extremely errant, misguided and false, so why trust them for 2020?

I guarantee the respective campaign managers are definitely tuned into each poll...and thousands of other data points. Polls are indicators of the electorate at any given point in time.

Ultimately, what counts is how effectively each candidate gets out the vote - nowand through election day. The voters are doing so in record numbers in a number of early voting states....

 

[essentialsaltes]

Do you have data to back this claim?

Pew Research

non-college-educated whites went 64/28 Trump/Clinton (this group made up 44% of all voters)
college-educated whites went 38/55 Trump/Clinton (this group made up 30% of all voters)

They had Hillary at >90% chance...she lost. If that's not errant idk what is.

Polls don't do that.
538 (source in OP) had Hillary at a 71% chance.

 

[98cwitr]

The POLLS did not give her a 90% chance. ANALYSTS did. The POLLS were pretty correct, as I said - nationally they were correct, state wise they were mostly within MoE. Analysts just didn't understand what they were looking at.

Yet we see those with the largest sample sizes and smallest MoEs were those who were wrong (see Oct-Nov 2016 data): Presidential election, 2016/General election polls - Ballotpedia

Then we had the same source as the OP pegging Clinton at 71.4%: https://projects.fivethirtyeight.com/2016-election-forecast/

So why trust this? It's clearly wrong.

 

[trunks2k]

Do you have data to back this claim?

Back up the claim that uneducated whites have a preference for Trump? He won that demographic by huge amounts in 2016:

But there's a massive fissure between college- and non-college-educated whites. Romney won non-college-educated whites by 26 points, according to polling data provided to NPR by the Pew Research Center. Currently, exit polling shows Trump's margin among that group to be roughly one-and-a-half times that.

And here's some info on how polls are adjusting this year for it:

But one thing came up again and again in our interviews: Pollsters told us they were now weighting their samples by education, because one key takeaway from 2016 was just how important someone’s level of educational attainment was in predicting their vote. “In mid-2016, we changed our weights by education, moving the percentage of high school or less respondents up while dropping the college-plus down,”

 

[trunks2k]

Then we had the same source as the OP pegging Clinton at 71.4%: https://projects.fivethirtyeight.com/2016-election-forecast/

So why trust this? It's clearly wrong.

71% to win is between a 1-in-4 and 1-in-3 chance to lose. Those are pretty good odds for the underdog. This is basic statistics here, and what people have been saying in this thread, an UNLIKELY event happening doesn't mean the model that pegs it as an unlikely event was wrong. It just means the unlikely thing happened.

But we're still comparing 2016 to 2020 which is a completely different comparison. We can look at past elections and see that polling is generally accurate. 2016 was a slight aberration, but that was due to several factors (such as the high number of undecides) that does not exist today.

In every respect, Biden is doing better this year in polling than Clinton did in 2016. He is polling 50%+ in key places, his national lead polling is exceeding Clinton, his leads are consistent, and there are few undecided voters. Ignore all that to your own detriment.

 

[98cwitr]

Back up the claim that uneducated whites have a preference for Trump? He won that demographic by huge amounts in 2016:

And here's some info on how polls are adjusting this year for it:

Thanks for the source, I remember seeing this now.

The reciprocal claim could be made as well, so I'm failing to see how this is somehow relevant. Dividing people by race is not a great way to unify the country. If a division for the sake of gaining statistically information must be done, we should draw those lines around cultural labels, rather than racial ones. In the case of the Pew Research Data provided above, we can simply deduce that college educated people have a proclivity of being liberal and therefore voting for liberal or progressive candidates. Now addressing the "whys" around that is very nuanced and I think outside the scope of this thread, but worth broaching.

 

[trunks2k]

The reciprocal claim could be made as well, so I'm failing to see how this is somehow relevant. Dividing people by race is not a great way to unify the country. If a division for the sake of gaining statistically information must be done, we should draw those lines around cultural labels, rather than racial ones. In the case of the Pew Research Data provided above, we can simply deduce that college educated people have a proclivity of being liberal and therefore voting for liberal or progressive candidates. Now addressing the "whys" around that is very nuanced and I think outside the scope of this thread, but worth broaching.

None of that is relevant to polling and the adjustments made for 2020 to account for the trends in 2016 in order to make better polling. Good stats break down things by the categories they can break it down by. If race is a good indicator for how people will vote, then that's a factor that needs to be taken into account. Same with education. Same with sex. We can combine them to be even more accurate. A given unedcuated person is X likely to vote Republican, but X changes depending on whether that person is male/female, black/white, economic status, etc.

 

[iluvatar5150]

If I didn't know better, you're calling white male conservatives "uneducated"

The fact of the matter is that the polls were extremely errant, misguided and false, so why trust them for 2020?

Yet we see those with the largest sample sizes and smallest MoEs were those who were wrong (see Oct-Nov 2016 data): Presidential election, 2016/General election polls - Ballotpedia

Then we had the same source as the OP pegging Clinton at 71.4%: https://projects.fivethirtyeight.com/2016-election-forecast/

So why trust this? It's clearly wrong.

Dear Mr. Moderator,

Why are you rehashing the same arguments made over the last 9+ pages of this thread instead of, oh... I don't know... reading the thread first and progressing from what we've already discussed?

 

[98cwitr]

None of that is relevant to polling and the adjustments made for 2020 to account for the trends in 2016 in order to make better polling. Good stats break down things by the categories they can break it down by. If race is a good indicator for how people will vote, then that's a factor that needs to be taken into account. Same with education. Same with sex. We can combine them to be even more accurate. A given unedcuated person is X likely to vote Republican, but X changes depending on whether that person is male/female, black/white, economic status, etc.

I would like to see economic status and poll data. Now THAT would be interesting! Also would like to see religion added

http://cdn.static-economist.com/sit...11/blogs/graphic-detail/20161112_woc961_0.png

Seems if you make more than $50k a year you're more likely to vote Trump.

Just did a bit of googling and I am going to post this here for the sake of future reference and posterity: President—Forecasting the US 2020 elections

Says 93% Biden!

 

[essentialsaltes]

Seems if you make more than $50k a year you're more likely to vote Trump.

If you make 50K-100K, yes. The over 100K one looks almost dead even. That's also the demographic that jumped the most away from the Republican candidate, compared to 2012.

 

[98cwitr]

If you make 50K-100K, yes. The over 100K one looks almost dead even. That's also the demographic that jumped the most away from the Republican candidate, compared to 2012.

Looking forward to 2020 numbers being added to that chart. Will be interesting for sure.

 

[durangodawood]

If I didn't know better, you're calling white male conservatives "uneducated"

The fact of the matter is that the polls were extremely errant, misguided and false, so why trust them for 2020?

Those polls were correct that voters preferred Hil. At least they got that right.

 

[98cwitr]

Those polls were correct that voters preferred Hil. At least they got that right.

48.2% of voters. Thank God the electoral college prevented a "tyranny of the majority" situation, as it was designed to do. With Biden polling higher than Hillary did, it will be interesting to see the results. Will have to remember to come back to this thread once all the chips have fallen.