National Reuters/Ipsos Poll - Cankles lead cut to 4

RCP/Reuters ^ | 10/19/2016 | Reuters

[rb22982]

When in the last 100 years has the the election had more than 2% difference in the national lead and the electoral college not going to the winner (hint: never). And even though EV are what matters, the direction of states WILL MIRROR national polls. It is not possible for Trump to be up 5 in florida and down 10% nationally (bloomberg, cbs, etc) when we lost Florida by 1-3% in the last two elections and only lost nationally by 4% or win Nevada where we lost the last 2 elections.

[Nero Germanicus]

Latest Florida Polling averaged: Clinton up +3.6
http://www.realclearpolitics.com/epolls/2016/president/fl/florida_trump_vs_clinton_vs_johnson_vs_stein-5963.html

Averaging of 13 most recent national polls: Clinton up 6.5
http://www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton-5491.html

There are always outlier polls. Personally I discard the widest and narrowest and average the rest.

[rb22982]

As 2012, 2014, and be exit showed, not all polls are created equal. The two most accurate 2012 polls show trump +1 and trump tied. Rcp is anti trump and still hasn’t added the wapo state polls showing trump up in those states. 2 way is irrelevant anyway. Only reason to look at two way is la times (rand) is only in the two way. But 3 polls show trumped tied and 3 have trump down around 10. One of those groups methodologies is off massively.

[CraigEsq]

This poll is not D+14. It is D+ 4.6.

1750 total respondents. 723 democrats (41.3%), 643 republicans (36.7%), 210 independents (12%). That’s D+4.6%.

[Nero Germanicus]

I use the realclearpolitics.com 4 way average, only I alter it by excludng the widest margin outlier and the narrowest margin outlier from the average.
Currently that means averagng 11 national polls.
http://www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton_vs_johnson_vs_stein-5952.html
That means I excluded Monmouth College’s Clinton +12 and Investor Business Daily’s Trump +1

[rb22982]

Doing that for brexit would have made it +6% off actual results, 2014 was 4 Pts off actual results and 2012 was a bit over 3 Pts off. Polls were off even more in 96 (not winner though) and 80. IBD has constantly been accurate on presidential elections and I believe 538 rated them most accurate in each of the last three.

[Nero Germanicus]

For the current election cycle, 538 has analyzed 17 polls from IBD/TIPP and they have given that pollster an A minus rating. They have an average error rate of 2.3, they have called 78% of races correctly and they have shown a 0.7% Republican bias. All those scores are excellent.
http://projects.fivethirtyeight.com/pollster-ratings/

By way of contrast, Selzer & Company has a A plus rating. 37 polls analyzed, simple average error 4.6, 84% of races called correctly and 0.0 partisan bias.
Selzer is currently polling for Bloomberg News and they have the race with Clinton up by 9 in the 4-way.
http://www.bloomberg.com/politics/articles/2016-10-19/national-poll

[rb22982]

I don't care about 538's ratings TBH. He had Brexit at 15% chance and blew a ton of races badly in 2014.

Bloomberg is massively anti-trump and wikileaks show he talked HRC people about cabinet position. We'll see what their poll for this race looks like on election day but I bet it won't be +9.