538 Model Megathread

[BoAtlantis]

Emerson has a Republican bias because they only call landlines but still has Hillary up by 2 in AZ.

I doubt AZ will go blue in the end but she seems to be making progress.

[BoAtlantis]

Emerson has a Republican bias because they only call landlines but still has Hillary up by 2 in AZ.

I doubt AZ will go blue in the end but she seems to be making progress.

We could see movement happen in either direction.  The range of possible outcomes is narrowing, however.  IN, MO, TX, SC, and MS were on the radar two months ago, but they've since fallen off.  I think GA may still be in the cards, but we haven't had post-debate (non-junk) polling.

Right now Clinton's ceiling is 2012 + NC, AZ, NE2, GA

I dare say Clinton's floor is the 272.  Though I can see 272-CO-PA-MI happening in an utter disaster scenario.  The window on that is closing rapidly, and might be closed permanently after the second debate.

The 538 model is of dubious use as long as it's using the junk Google surveys.  I'd like to see Nate defend them in an article.

Google Survey was actually very good in 2012 and got Obama's margin over Romney correctly in popular vote. Their strength is that they use a huge number of samples.

Their only problem is that their state to state results aren't reliable.

Their NC 9/27 - 10/3 survey shows +11 Trump.
Their FL 9/27 - 10/3 survey shows +8 Trump.
Their PA 9/27 - 10/3 survey shows a tie.
Their GA 9/27 - 10/3 survey shows +15 Trump.
Their ME 9/27 - 10/3 survey shows +28 Hillary.
Their KS 9/27 - 10/3 survey shows +8 Hillary.

[BoAtlantis]

Emerson has a Republican bias because they only call landlines but still has Hillary up by 2 in AZ.

I doubt AZ will go blue in the end but she seems to be making progress.

AZ and GA are not going to go blue. i think her resources are better spent in shoring up the firewall and trying to flip Ohio.

Just like how Republicans think PA is in play every four years.

AZ is just a noise. It's Democrat's wet dream though I'd love to be proven wrong.

[BoAtlantis]

The 538 model is of dubious use as long as it's using the junk Google surveys.  I'd like to see Nate defend them in an article.

If the surveys aren't faked and if they provide numbers that are better than what you would obtain from random chance (they do), then it has use for a statistical model if given a proper weighting. This is true even if they are very inaccurate. He doesn't have to defend using them because it's just statistics. You could quibble with him about the weighting.

The data quality is poor, the likely voter screening is poor (it's self-reporting, not human-guided or at all scientific), and it's based on IP-location, which is deeply flawed.  Trump is up 8 points in Florida and Clinton is up 9 in Missouri?  Clinton up 29 in Maine and 3 in Indiana?  And only 1 nationally?  Meanwhile Clinton's +9 in Kansas is completely accounted for by the fact that IP geolocation defaults to the geographic center of the United States if no precise location can be found.

There's no justification for including this garbage.  At least not that I can see.  If Nate wants to defend it, I'm all ears.

He has addressed this before.

http://fivethirtyeight.com/features/is-a-50-state-poll-as-good-as-50-state-polls/

"Is a 500-person subsample of Colorado voters from a 20,000-person national poll the same thing as a 500-person poll that was dedicated to Colorado, specifically?

After thinking and researching my way through the problem, my answer is that these polls aren’t quite the same. The Colorado-specific poll is likely to provide a more reliable estimate of what’s going on in that particular state. And it deserves a higher weight in our model as a result."

"One potential source of error has to do with demographic weighting. Polls of all kinds engage in extensive demographic weighting because people aren’t equally likely to respond to polls Typically, for instance, white voters are more likely to respond to telephone polls than black voters. Pollsters attempt to counteract this by giving extra weight to the black voters they reach until the demographics of their poll matches that of Census data or other reliable sources."

"With online polls, the problem is that IP addresses aren’t 100 percent reliable — for instance, a website would think I’m in Connecticut right now because that’s where ESPN’s internet connection is based, even though I’m writing from the FiveThirtyEight office in New York City. "

"..........we’ll continue to use the state-by-state breakouts from Ipsos and other pollsters, but we will significantly lower the weight assigned to them in our polling averages. These polls will continue to have some influence around the margin — we don’t want to discard their data entirely — but state-specific polls will have more influence on the forecast. "

[BoAtlantis]

Emerson has a Republican bias because they only call landlines but still has Hillary up by 2 in AZ.

I doubt AZ will go blue in the end but she seems to be making progress.

AZ and GA are not going to go blue. i think her resources are better spent in shoring up the firewall and trying to flip Ohio.

Just like how Republicans think PA is in play every four years.

AZ is just a noise. It's Democrat's wet dream though I'd love to be proven wrong.

AZ and GA went for Bill Clinton.  IN went for Obama.  These things aren't set in stone.

The Sunbelt will be unusually good for Hillary this year, offset by the Rust Belt being unusually bad.  This isn't a trend, just issues specific to this election cycle.  Where the new economy thrives, Hillary will do well.  Where the old economy sputters, Trump will find traction.  This is why Trump is doing relatively well in Ohio and Iowa, and Clinton is doing relatively well in Arizona and Georgia.

It's not a trend, it's new economy vs old.

I thought AZ was only in play for Clinton because of Perot.

I hope the poll is up to something but until I see another poll, I doubt she is truly ahead.

[BoAtlantis]

God bless the freiwal! Such a beautiful freiwal, folks, like none you've ever seen. And it's going to grow BIGGER in the next few cycles, BELIEVE me.

So what are you saying ..... that the Firewall just got two feet higher ?

BUILD THAT WALL... BUILD THAT WALL!

We're gonna build a 322 EV freiwal.  AND WE'RE GONNA MAKE REINCE PRIEBUS PAY FOR IT!!

I'm dying after reading this. lol

[BoAtlantis]

Trump's chances keep going up, exactly what I was kinda freaking out about a few days ago. First time we've seen a state flip from blue to red since the debate I think.

Also, Ohio is to the right of Arizona. That is so strange

His models has been going towards Trump for a while now even with very positive polls for HRC.

AZ: C+2 [increased for Trump]
GA: Tie [increased for Trump]
Breitbart Nat: C+1 (+1 from their last poll) [increased for Trump]
Pew Nat: C+7 [increased for Trump]

It's doing very questionable movements. I'm not buying it.

Yea it's interesting because Drew Linzer at Daily Kos and Sam Wang at Princeton Consortium did better than Silver in 2012, margin-wise, and they're more optimistic at 95% and 99%, respectively.

With new batch of polls, Hillary's chance of winning is almost exactly the same for them.

[BoAtlantis]

Do we generally view Nate's model as more reliable than the others?

Silver gets more publicity because his site is more decorated. Drew Linzer of Daily Kos and Sam Wang of Princeton Consortium were more accurate in 2012.

http://rationality.org/2012/11/09/was-nate-silver-the-most-accurate-2012-election-pundit/

Drew Linzer   0.00384
Wang & Ferguson   0.00761
Nate Silver   0.00911
Simon Jackman   0.00971
DeSart & Holbrook   0.01605
Intrade   0.02812
2008 Repeat   0.03922
Margin of Error   0.05075
Coin Flip   0.25000

That doesn't translate to future performance. Sam Wang did way worse in 2014 but that was midterm in which Democrats don't tend to turn out.

Sam Wang also only uses polls and mean-revert whereas Silver uses his own special sauce like the economy, endorsements etc.