The US Election Polling / Data Thread

[Acey]

I don't even understand why they bother publishing popular vote polls... cause that's not how the election works.

[Senator Clay Davis]

Dave Weigel ‏@daveweigel 1h1 hour ago
According to 538, there is now a 24.4% chance of a post-election column titled "I wrote in Harambe because I didn't think Trump could win."

https://twitter.com/daveweigel/statu...30824188280832

[GGG]

Originally Posted by Acey

I don't even understand why they bother publishing popular vote polls... cause that's not how the election works.

Because the states especially states near each other in general move together. Each state election is not an independent event so in smaller areas where you lack polling due to cost or difficulty reaching minorities the national poll allows for larger sample sizes especially in minority Groups.

So in states like Ohio and Florida the national poll is likely of little predictive value but in less polled states that are part of the 12-15 states in play this election they offer a lot of value.

Also in a 3+ point race the state polls don't matter. In a 0-2 point race they do. So right now given the race the state polls are not required to predict the outcome of an election today. As it tightens they are.

[troutman]

Trump has to win every battleground state, and flip one blue state to red, to get to 270.

[Senator Clay Davis]

Trump doesn't have to win every battleground state, FL, NC, OH and PA makes him President, rendering battlegrounds NH, NV, CO, VA all meaningless. I know people want to think it's still a real long shot, but it's not. At this point one big polling error in one state might be enough to give it to Trump. It doesn't even have to be a national polling error.

[troutman]

I'm not counting PA as a battleground state.

http://www.cnn.com/election/interact...al-college-map

I predict 323 to 215 for Clinton.

[Acey]

Originally Posted by GGG

they offer a lot of value.

I'll rephrase. Because Trump is 46-45 in a poll this morning does not mean he's 46-45 to win the election... obviously that metric is still in Clinton's favour despite dwindling. That's all I was really getting at; there is clearly information to be derived from national polls.

[Senator Clay Davis]

Originally Posted by troutman

I'm not counting PA as a battleground state.

http://www.cnn.com/election/interact...al-college-map

I predict 323 to 215 for Clinton.

But that's the thing, if any state will have a massive polling error it will be Pennsylvania since it's demographically similar to Ohio, where Trump has been ahead for almost two months. Hillary's campaign is also spending money again in PA (and Wisconsin) which suggests it's closer than the polls indicate.

[mrkajz44]

Originally Posted by CorsiHockeyLeague

A 25% chance of anything happening is a significant chance. Of course it's worth being concerned about, especially when it's a 25% chance of... this.

Yeah, I thought we were headed to landslide territory a few weeks ago, but a 25% chance is quite high when you think about it. His odds are roughly the same odds as flipping heads on a coin on two consecutive flips. That is anything but rare.

Since the national data still leans Hillary, I think she'll still win by a comfortable margin, but I think a lot of people are considering this a done deal when all Trump needs is a few percentage points swing to make it a really tight race.

[Bunk]

Which currently blue state does he flip (in addition to winning every swing state) to win the election though? Pennsylvania? Virginia? Michigan? Colorado?

He can win Florida, Ohio, Nevada, Iowa, NC, Utah, Arizona and still lose! He has to win ALL of those and still one more.

It just doesn't seem all that plausible.

[Senator Clay Davis]

Trump getting to this point in the first place didn't see all that plausible. At this point I wouldn't rule anything out. Hillary possibly swiping North Carolina is looking to be huge though, if Trump could just hold onto it he would be in a true 50/50 race.

[Iowa_Flames_Fan]

Originally Posted by troutman

I'm not counting PA as a battleground state.

http://www.cnn.com/election/interact...al-college-map

I predict 323 to 215 for Clinton.

Although I agree, I think SCD's point is valid. A Trump victory just needs polls in PA to either be off, or to shift at the last minute. And unlike most of the other battlegrounds, PA has no early voting to speak of, so he won't be playing catchup on election day in the same way he is in North Carolina.

[troutman]

Some early voting states allow you to change your vote 1 or 2 more times?

[GGG]

Originally Posted by mrkajz44

Yeah, I thought we were headed to landslide territory a few weeks ago, but a 25% chance is quite high when you think about it. His odds are roughly the same odds as flipping heads on a coin on two consecutive flips. That is anything but rare.

Since the national data still leans Hillary, I think she'll still win by a comfortable margin, but I think a lot of people are considering this a done deal when all Trump needs is a few percentage points swing to make it a really tight race.

I think the 25% though overstates his chances somewhat.

Because the large bulk of that chance is counting on a polling error. And we don't really know the possibility of a polling error that well given the small sample size of presidential elections.

So in one universe you have a 50/50 chance of flipping heads and in the other universe you have 95% chance of flipping heads. And you only have a few clues to which universe you are in.

I think that Hillary is more likely to out perform polling then not for a few reasons.
- ground game matters
- Trump consistently underperformed polling in primaries (no shy trump affect)
- Hillary out performed polling outside of Michigan in primaries.

So these items are not baked into the model, the model assumes that a polling error in each direction is just as likely. I would disagree with that assumption.

[Parallex]

Originally Posted by Senator Clay Davis

But that's the thing, if any state will have a massive polling error it will be Pennsylvania

Trump hasn't led in any poll of Penn since the middle of September... in that time 23 different polling companies have conducted 1 or more polls of the state. To assume a massive polling error is to assume that 23 seperate entities all with their own methodology have all been blowing it. Doubtful

Originally Posted by Senator Clay Davis

since it's demographically similar to Ohio, where Trump has been ahead for almost two months.

That's not true. In the past two months Clinton has been ahead several times (and behind several times). It has not been a steady hold for Trump. I'd say he's ahead right now... but if I were to say how long he's been ahead I'd say two weeks (not two months).

Honestly, if Trumps hopes are pinned on there being a massive polling error (in his favor) in some places (but not in others) and no corresponding error in Clinton's favor anywhere else and Clinton's money advantage counting for naught and organizational advantage counting for naught...

I understand the nervious nellying... the thought of Donald Trump controlling nuclear codes is terrifying... but folk ought to chill, she got this.

[Senator Clay Davis]

The only time Hillary led in Ohio in the past two months was after the tapes came out, and she only briefly led those polls. Otherwise Trump has led every poll, and the Hillary campaign has more or less stopped spending money/time there. She might win it if there's a big polling error her way, but it's looked like a safe Trump state for a while now. And it's not about methodology for PA, it's about the "hidden Trump support". PA is demographically similar to OH, though PA has larger suburban populations which hurts Trump. But if she under performs Obama with blacks (as she appears to be doing so far in early voting), it'll be a lot closer than people want it to be.

I know people think she's got this, but it's not even close to in the bag. 60/40 is a realistic advantage for her, people assuming 80/20 are probably ignoring that her net favorability is the same as his. These are unquestionably the two most hated candidates in history.

[Parallex]

Originally Posted by Senator Clay Davis

The only time Hillary led in Ohio in the past two months was after the tapes came out, and she only briefly led those polls. Otherwise Trump has led every poll

I just took a look at every poll conducted entirely past Sept 1 (using the 538 updates of Ohio)...

Trump led in 29
Clinton led in 21
They were tied in 4

Trump has led in roughly half the polls. Half is not every.

Originally Posted by Senator Clay Davis

probably ignoring that her net favorability is the same as his.

http://www.realclearpolitics.com/epo...able-1131.html
http://www.realclearpolitics.com/epo...able-5493.html

Clinton: -9.3
Trump: -21

[Senator Clay Davis]

I use RCP, and I count 18 polls, with Hillary ahead in only 4 of them, all right at the time the tape came out.

[Parallex]

Originally Posted by Senator Clay Davis

I use RCP, and I count 18 polls, with Hillary ahead in only 4 of them, all right at the time the tape came out.

538 has more. My own view of the race in Ohio is basically that Trump has had a lead for roughly the past two weeks, Clinton for roughly the two weeks before, and before that no one had a clear advantage (no sustained polling lead just small spurts on both sides).

[Bunk]

Originally Posted by Iowa_Flames_Fan

Although I agree, I think SCD's point is valid. A Trump victory just needs polls in PA to either be off, or to shift at the last minute. And unlike most of the other battlegrounds, PA has no early voting to speak of, so he won't be playing catchup on election day in the same way he is in North Carolina.

Let's remember Pennsylvania hasn't gone Republican since 1988, which was a massive landslide election. It's barely a swing state.