Nate Cohn: ABC Tracking Poll Assumes Very White and Very Old Electorate

[Doctor Jack]

In fact, it appears that a few polls are assuming that white voters and old voters will be up quite a bit this year and that black, hispanic, and millennial voters are going to be way down. In the case of ABC's tracking poll, they actually had closer to what the presumed electorate was going to look like until last weekend. That was when Clinton was up by double digits. Then they change their assumption on how the demographics were going to look on Friday or Saturday (that isn't speculation or "unskewing" ABC said they did this) and now suddenly its a very small Clinton lead. Non-Doom and Gloom Nate says it is extremely unlikely that old white voters are going to turn out in numbers we haven't seen in many cycles, while at the same time black and especially hispanic votes actually decreasing compared to 2012. This despite hispanic votes in early voting in Florida, for example, increasing by nearly 100% over 2012. Hmmm...hmmmmm...hmmmmmmmmm.

I like this new Nate, as opposed to Nate Silver who says the polls may be extremely unreliable this year but only if they show Clinton ahead. If they show Trump ahead then they have to be put into the forecast ASAP.

https://twitter.com/Nate_Cohn/status/794343177553592324

[Dawson Leery]

1. That's Nate Silver. Any good polls for our Hillary are discounted right away.

Charlatan.

[redstateblues]

6. Nate Silver always hedges his bets. His last paragraph is the exact opposite

of what comes before it. Nate was a one trick pony. No more.

[Doctor Jack]

12. I hesitate to question Nate Silver but...

...I do wonder if his day has passed. The Republicans underestimated him 2012 and I don't want that to be me this year but I do see some differences. In 2012 the republicans were saying that 538 was wrong and that romney was way ahead, not behind by a few points. And the GOP in 2012 made a lot of really dumb arguments to try to prove their point. Of course this year, he still has clinton ahead so we aren't starting from a place of desperation but people saying that Silver is wrong this year are making good points that they can back up with numbers and facts. Not "well my neighbor has a Romney lawn sign therefore romney will win Michigan."

Nate Silver did well in 2008, 2010, and 2012 but he didn't do great in 2014, he absolutely bombed the last British election, and his primary projections were somewhat hit or miss. You can definitely see cracks forming over at 538.

[Blue_Roses]

24. I think you might be right, because the demographics

have definitely changed this election cycle. What was usually the norm is now a free-for-all.

[Koinos]

2. I think many of the polls underestimated the importance of newly motivated "unlikely voters."

[Doctor Jack]

7. I think they are so focused on "millennial" enthusiasm being kind of flat

that they are missing the massive number of new hispanic voters flooding the polls. The go to, undisputed expert on elections in Nevada is saying everyday now that polls are massively underestimating hispanic turnout in the state but most people, including Nate Silver, don't seem to care. 538 still says Nevada is a toss up or even slightly leaning trump, even though the experts in Nevada are saying that democratic early voting in Nevada is so overwhelming that Trump likely has already lost. We are getting similar stories in North Carolina and Florida. We are seeing historic hispanic turnout in early voting and in voter registration in general but polls keep showing hispanic turnout as equal to 2012 or down roughly 20%. Latino Decisions, which tracks this kind of thing, says that polls like ABC assume that hispanic turnout will be only 8% this year when it was 10% in 2012. The good news is even assuming a much whiter and older electorate, we are still winning.

Again, Nate Cohn seems to have his finger on the pulse with these polling issues. He has been pointing out that 538 is becoming bloated with extremely questionable internet only polls showing odd results like Clinton ahead in Kansas and a tie in Maryland. 538 assumes all data is useful, even when it is clearly garbage. The NYT and Huffington Post projections throw out the unproven, internet polls and only stick with trustworthy, reputable polls with a long track record and that is why thy have clinton with significantly higher odds of winning.

[Democat]

3. ABC changed the poll to bring Clinton down?

Liberal media.

[Doctor Jack]

4. They changed the demographics of who they were polling

because they said it seemed to be too generous to hillary. That is why we saw a 13 point drop in 2 days. Of course said liberal media made a huge fuss about that, saying there was a massive spike in support for Trump, without reading the press release. It seems if I, a random nobody on the internet, can find information about that poll (again from the pollster readily admitting they changed their methodology) that some fancy journalist could too but what do I know, I don't own a media conglomerate.

[MyNameIsKhan]

20. Yup media is invested in horse race

[Quixote1818]

19. News orginizations do this every electon cycle to make it look like a horse race

That is what you get when you have news organizations that thrive off of the number of viewers. If you look at Obama's internal poll from 2012 there was zero tightening but all the national polls showed him dropping several points. Don't think it has anything to do with the media being liberal or conservative but more about keeping people tuned in which translates to more cash.

[Chasing Dreams]

5. I've been saying CLINTON LANDSLIDE for a while now...

And nothing I've seen since the Comey Hurricane suggests otherwise. Look at the real turnout data from Nevada, Florida, Texas, and Arizona. Ipsos +8 pretty close to what will happen:

Clinton 50
Asshole 42
Johnson 5
Stein 2
Others 1

Hillary 350+ EVs
52 D Senate seats
Within 15 of House majority

[MyNameIsKhan]

8. More than this look at cross tabs my god

[Charles Bukowski]

10. And STILL Clinton is up 3 points.

Unreal.

[Dawson Leery]

16. More Republicans are not voting vs. Democrats.

[MyNameIsKhan]

21. How can you change the LV just based on Comey letter? I now think everything is rigged...

[Charles Bukowski]

9. A number of pollsters are assuming

an increase in white voters. This isn't likely, as it would buck a 36 year trend, and our electorate has only grown more diverse, but whatever. HRC's got this.

[Doctor Jack]

15. The good news is, we are still winning

Despite some questionable polling crosstabs in the last few days, with demographics being far more pro-trump that what is likely, we are still winning by 3 or 4 points. She is beating him, even with both arms tied around her back and with a blindfold on. Worst case, the polls are 100% correct and...we still win.

[Warren DeMontague]

17. Dollars to donuts none of these polling geniuses account for pot legalization in several states

Bringing larger numbers of energized young voters to the polls this time around.

Nevada and Arizona it might actually be a factor.

[Mz Pip]

18. That may be

But old white people get out and vote. They are reliable. They show up.

[Wounded Bear]

22. But they don't grow in numbers that fast....

I know, I am one.

[RAFisher]

23. ABC didn't just pull this out of their ass. They aren't just "assuming" turnout, it's calculated.

http://abcnews.go.com/US/PollVault/abc-news-polling-methodology-standards/story?id=145373

Our practice for ABC News is to develop a range of “likely voter” models, employing elements such as self-reported voter registration, intention to vote, attention to the race, past voting, age, respondents’ knowledge of their polling places and political party identification. We evaluate the level of voter turnout produced by these models and diagnose differences across models when they occur.

Nate just said they seem off, which is possible. I don't know why people keep insisting that voter turnout models are just assumed. Read the methodology. This poll uses self reported data to calculate voter turnout. This isn't complicated. If Clinton supporters are telling ABC posters that they aren't registered to vote, don't plan to vote, etc. then I think ABC has good reason to believe that they aren't actually going to vote. It could just be that the Comey news is making less Clinton supporters want to vote or other factors.

This isn't some conspiracy. ABC states it clearly in the article that Nate liked to. They are finding Trump supports across the board more enthusiastic than Clinton supports. That translates to voter model that is much more favorable to Trump.