The absentee/early vote thread

[Erich Maria Remarque]

BTW, is there usually any/big difference in AV/EV between rural compared and non-rural areas? If not, poor Trump

Anybody?

And the same applies to education. We clearly see a widining education/rural–urban gap in this election. If different groups have different patterns of early voting, it could explain Trump's lagging. At least partly.

That's why EV sample might be non-representetive of all voters, as 538 mentioned.

[Erich Maria Remarque]

Retweeted (only the first one) by Nate Cohn, Upshot.

https://twitter.com/electionsmith/status/791836459506495488

You must be logged in to read this quote.

You must be logged in to read this quote.

[Erich Maria Remarque]

NC 10/28:

https://twitter.com/ElectProject/status/791980869191733249

You must be logged in to read this quote.

You must be logged in to read this quote.

[Erich Maria Remarque]

All signs point to a very high turnout election. Good news for Democrats, of course.

EV is becoming more popular, so probably no. Hard to say now.

[Erich Maria Remarque]

Then you should show some facts instead of mumbling

Nate Cohn ‏@Nate_Cohn  2h2 hours ago Manhattan, NY
"Clinton has an 18 point lead in NC early voting with more than 1 million votes cast, according to our estimates"

This isn't like other years when you could just use Rep and Dem registration ID to safely infer voting leads-- a significant number of registered Republicans in both NC and Florida will be voting for Clinton, as indicated by the Upshot's model extrapolating from their polling.

You're going to come out very disappointed with the conversation you just started.

I'm not saying he is wrong, but their latest poll was conducted when Trump was 7% down nationally efter a terrible one-month-long news cycle. His favorability among R was ~60% back then, before Obamacare "gate" it was about 70%, right now it might be even higher. He is even doing again better among Indys (so it is not enough data/evidence for that).

[Erich Maria Remarque]

This thread is strangely quiet today. Throughout the past week I've heard people say that the weekend results would be crucial, so how is it going?

Souls to the Polls happens on Sunday, you paranoid Chicken Littleing twat.

[Erich Maria Remarque]

do not-registered UAF matter in any way?

they are part of the dem/rep-numbers anyway, not?

Cohn is pointing out that there's a difference in how people are registered with the state, and how they identify themselves when a pollster calls.  I believe there are more voters who are registered as unaffiliated (because they didn't bother to enroll) but then identify with their current lean when they are called.  My recollection (maybe I'm wrong) is that a lot of young people are registered unaffiliated, but identify as Democratic.

Yep. And, OTOH, many current R's/conservatives are self-IDing as independents, even though they aren't registered as such. Both just point to a need for a different lens according to whether you're looking at EV data or poll #s.

Yep!  This is the big problem with re-weighing to party ID.  Imagine both things are true.  You have young, independent-registered voters identifying as Democrats.  You also have Republican-registered voters identifying as independent, because they don't like the party.

That means your poll will show as having too many Democrats, and too few Republicans.  It also means your self-identified Independent sample is more Republican than actual, registered Independents.  That's because it's lacking a lot of Democratic-leaning Independents (those young voters who are IDing as Democrats), and includes a lot of registered Republicans (those registered Republicans identifying as Independents).

If this happens, your sample should be more Democratic, and less Republican, than registered voters.  Re-weighting to party ID will mess this up.  Also, because your self-IDed Independents are a lot more Republican than actual Independents, re-weighting to party ID will further skew the sample Republican.

(Sorry, I'm sure y'all know this, just my daily Bill Mitchell antidote)

In Florida it might be the other way round (OBS. it is based just on one poll, might be an outlier).

http://www.nytimes.com/interactive/2016/10/30/upshot/florida-poll.html

You must be logged in to read this quote.

Though

You must be logged in to read this quote.

[Erich Maria Remarque]

This isn't surprising. Plenty of Panhandle whites register as Dem despite never voting that way.

As I understood, Trump is extra strong (comparing to other Republicans) in this segment. The point is that one should be extra cautious with "extrapolating" early voting data.

[Erich Maria Remarque]

http://www.nytimes.com/interactive/2016/10/31/us/elections/earlyvoters.html?hp&action=click&pgtype=Homepage&clickSource=story-heading&module=b-lede-package-region&region=top-news&WT.nav=top-news

You must be logged in to read this quote.

[Erich Maria Remarque]

Michael McDonald ‏@ElectProject  23m23 minutes ago
North Carolina demographic characteristics of unaffiliated (no party reg) #earlyvote (as of 10/30)

But how does it compare to overall voters? Unaffiliated voter are younger, less white and more female I guess?

[Erich Maria Remarque]

It's getting really hard to see a path for Trump in CO/NV.  He's going to need to break through in PA or WI, as expected.

But is it some evidence that EV is so predictive and better than polls?

For example Obama won Nevada by 12% and 6% in 2008 and 2012, respectively. Was his EV margin in 2008 much better than in 2012? Specially, EV at the middle/start of the campaign (there are still 7 days to go).

[Erich Maria Remarque]

https://twitter.com/ElectProject/status/793457148323737600

[Erich Maria Remarque]

Guys, do you remember how you criticised/unskewed pollster's for having too white and not black enough crosstabs? Well, guess *what


*it is though a bit too early, but still.

[Erich Maria Remarque]

What was the Dem lead in CO at this point in 2012?

On November 5th, Dems trailed by 2%.

http://www.huffingtonpost.com/2012/11/05/early-voting-results-2012_n_2076029.html

EDIT: misstake...

In 2008, Obama won EV by 9 4% (correct me if I'm wrong), he won NVCO 9%.
In 2012, Obama    lost EV by  2%                                     , he won by  5%.

EV is not that pretty predictive. With reservation that laws didn't change too much.

[Erich Maria Remarque]

What was the Dem lead in CO at this point in 2012?

On November 5th, Dems trailed by 2%.

http://www.huffingtonpost.com/2012/11/05/early-voting-results-2012_n_2076029.html

EDIT: misstake...

In 2008, Obama won EV by 9 4% (correct me if I'm wrong), he won NV 9%.
In 2012, Obama    lost EV by  2%                                     , he won by  5%.

EV is not that pretty predictive. With reservation that laws didn't change too much.

Hmmm....do you have a link for that '08 number? Electproject has the party difference in '08 as D+1.

I took if from here
http://www.theatlantic.com/politics/archive/2012/11/whos-really-winning-early-voting/264436/

You must be logged in to read this quote.

Oh f**k. I mixed CO and NV

[Erich Maria Remarque]

Nate Cohn ‏@Nate_Cohn Nate Cohn Retweeted Greg Sargent
Key is that it's based on a poll that had Clinton up 7. If race tightened/or poll too D, these estimates will be too

I don't than really understand what then the model is? As I understand this tweet, it is just a [Siena] poll, but instead of LV model, thay use actual results, right? Intresting and, actually, a good news for Trump, if it is true.

EDIT:
Good news for Trump, if this poll was wrong, the race changed etc..

[Erich Maria Remarque]

Nate Cohn ‏@Nate_Cohn Nate Cohn Retweeted Greg Sargent
Key is that it's based on a poll that had Clinton up 7. If race tightened/or poll too D, these estimates will be too

I don't than really understand what then the model is? As I understand this tweet, it is just a poll, but instead of LV model, thay use actual results, right? Intresting and, actually, a good news for Trump, if it is true.

It's at least not bad news. The model is pinned entirely to Upshot's NC+7, and they have her at +5.5, which she certainly shouldn't love. That said, if she leads polls by more than 1.5 on ED, this model says she wins.

All of that said, if there's anything wrong with the underpinnings of Upshot's demos, the whole thing is worthless.

Great news for me

I gave almost up on NC, cause I thought that they used their "average" from Upshot rather than Siena poll only. Siena has a D house effect and 538 adjusted it from C+7 to C+4 (IDK, if this includes last week trend line, though). And Upshot's turnout model already shows 1.3% points swing (if I understood it right). So even if NC is lean D, it is doable for Trump

[Erich Maria Remarque]

You must be logged in to read this quote.

[Erich Maria Remarque]

The Early Vote In Nevada Suggests Clinton Might Beat Her Polls There, 538

[Erich Maria Remarque]

You must be logged in to read this quote.

[Erich Maria Remarque]

How many have already voted in Florida? In mln or  in % of total of 2012.

[Erich Maria Remarque]

I start to think, that Upshot/Sienna was a pretty big outlier. NC is clearly a tossup. Probably slightly towards Trump.
[with all information for now]

From https://twitter.com/ElectProject/status/793792941127004160

[Erich Maria Remarque]

2. Polls consistently indicate Clinton is leading and leading big among early voters.

It depends on how well Obama was doing in polls among EV. For instance if he was doing ~better, then it is nothing strange at all. And MOE should be taken into account.

Otherwise, it could indicate some sort of realigment (that would not shoke me).

It looks like the FL/NC polls are backwards, with NC being less diverse, Florida being more diverse (Latinos more than offsetting lower Black share).

What is raw numbers of EV by race in Florida 2012 vs 2016?