538 Model Megathread

[Xing]

Anyone hazard a guess if and when Clinton will finally gain another state in the polls-plus forecast?

We'd probably need to see polls showing a consistent lead for her in Arizona and Georgia, starting in September or so.

[Crumpets]

A couple polls came out today that, while having Clinton in the lead, were less than glorious for her, yet the model doesn't seem to have budged since yesterday. Anyone know why?

[Gass3268]

A couple polls came out today that, while having Clinton in the lead, were less than glorious for her, yet the model doesn't seem to have budged since yesterday. Anyone know why?

They haven't put in the Pew or Nevada poll yet. It should also be noted that the last Pew poll was in early June, so even though it had Clinton up 12, it won't be as awful as it would be if the poll would have been out in late July.

[Mehmentum]

A couple polls came out today that, while having Clinton in the lead, were less than glorious for her, yet the model doesn't seem to have budged since yesterday. Anyone know why?

538 takes into account where each individual poll previously had the candidate.

For example, Rasmussen had Clinton's margin down 1 point from their last poll.  Yougov's poll has the same margin as earlier in the month. The Reuters poll only has a 1 point drop from the previous poll.

[riceowl]

Somewhat off-topic, but 538 cites this website throughout the attached article. Have they done that before?

http://fivethirtyeight.com/features/election-update-why-clinton-doesnt-have-this-race-locked-up/

[Crumpets]

By the way, since I'm bored, here's the 538 maps done Atlas-style.

Polls-plus probability:

Polls-plus winning percent:

Maybe I'll do another one minus third parties if I'm still bored later.

[Mehmentum]

Somewhat off-topic, but 538 cites this website throughout the attached article. Have they done that before?

http://fivethirtyeight.com/features/election-update-why-clinton-doesnt-have-this-race-locked-up/

Yep, they've done it several times before.

[Gass3268]

They put the new numbers in, Clinton dropped about 2% in all three models.

[Wiz in Wis]

They put the new numbers in, Clinton dropped about 2% in all three models.

All things considered, the past 15 days have been very stable in their model.

[Xing]

Clinton's still consistently winning 347-191 in the polls-plus forecast, so I can't complain.

[Gass3268]

They put the new numbers in, Clinton dropped about 2% in all three models.

All things considered, the past 15 days have been very stable in their model.

This is good news for Clinton. We are almost to the one month historical precedent of whoever leads a month after the conventions wins the election.

[elcorazon]

polls only has clinton winning Arizona, but outside of arizona, all of the Clinton states including NC, FL, OH, IA, NH and NV (PA, too) are over 70%.

[dspNY]

Starting today, there are no artificial bounce effects in 538's models, so what you see is what you get

[heatcharger]

They really just added all those Reuters polls... Montana is a lovely light red though.

[Xing]

They're putting some stock in Reuters. That somehow flipped NC back on the polls-plus map.

[heatcharger]

They're putting some stock in Reuters. That somehow flipped NC back on the polls-plus map.

Reuters/Ipsos used to be a live phone pollster, and they were generally good, but then they switched to being an internet only firm, and now they're all over the place. If Nate and Co. are smart they'll change their weight in the model.

[Erich Maria Remarque]

They're putting some stock in Reuters. That somehow flipped NC back on the polls-plus map.

Reuters/Ipsos used to be a live phone pollster, and they were generally good, but then they switched to being an internet only firm, and now they're all over the place. If Nate and Co. are smart they'll change their weight in the model.

Are you sure? I thougt they've always been using internet polling only.

[heatcharger]

Reuters/Ipsos used to be a live phone pollster, and they were generally good, but then they switched to being an internet only firm, and now they're all over the place. If Nate and Co. are smart they'll change their weight in the model.

Are you sure? I thougt they've always been using internet polling only.

This poll from 2012 says "live telephone interviewing". They had Obama +3 which was pretty accurate.

[Erich Maria Remarque]

Reuters/Ipsos used to be a live phone pollster, and they were generally good, but then they switched to being an internet only firm, and now they're all over the place. If Nate and Co. are smart they'll change their weight in the model.

Are you sure? I thougt they've always been using internet polling only.

This poll from 2012 says "live telephone interviewing". They had Obama +3 which was pretty accurate.

If 538's rating is based on their "live telephone interviewing", it is really odd

Someone can probably ask them.

[eric82oslo]

On August 25, the closest (and furthest away) states relative to the projected national popular vote margin (of +7.1% for Clinton) are, according to 538's Polls Only forecast, the following:

New Hampshire – 0.1% D
Pennsylvania – 0.3% R
Nevada – 1.4% R
Minnesota – 1.6% D
Virginia – 1.7% D
Colorado – 1.8% D
Florida - 2.1% R
Ohio – 2.5% R
Michigan – 2.6% D
Wisconsin – 2.6% D
Maine – 3.0% D
Iowa – 3.7% R
New Mexico – 4.0% D
North Carolina – 4.2% R
Oregon – 4.5% D

Connecticut – 6.0% D
New Jersey – 6.4% D

Georgia – 7.9% R
Washington – 8.9% D
Missouri – 8.9% R
Arizona – 9.1% R
Delaware – 9.6% D
South Carolina – 9.9% R
Montana – 10.9% R

Illinois – 12.4% D

New York – 13.9% D
Rhode Island – 14.6% D
Texas – 14.8% R
Indiana – 14.8% R
Kansas – 16.0% R
South Dakota – 16.0% R
Alaska – 16.1% R
Massachusetts – 16.6% D
Mississippi – 16.6% R
Tennessee – 17.0% R
Kentucky – 17.0% R
Arkansas – 17.2% R
North Dakota – 17.2% R
California – 17.5% D
Utah – 17.8% R
Vermont – 17.9% D
Louisiana – 18.5% R

Nebraska – 20.9% R
Maryland – 21.2% R

West Virginia – 23.5% R
Idaho – 24.2% R
Alabama – 25.3% R

Oklahoma – 31.7% R
Wyoming – 33.0% R
Hawaii – 33.6% D

D.C. – 67.8% D


Notice that no less than 15 states are currently considered to be within 5% either way of the national popular vote. This is a slightly higher number than in 2012, when only 12 states where considered to be true battlegrounds.

Kind of interesting to see states like Georgia, Missouri, Arizona and South Carolina only ranked as number 18, 20, 21 and 23. On the other hand, states that are hardly mentioned these days, like Maine, New Mexico and Oregon, are ranked higher, at places 11, 13 and 15.

Perhaps about time to get some new, fresh polls from Minnesota (3), Maine (4), New Mexico (3), Connecticut (3), Washington (4), Delaware (2) and Montana (2)? The number in brackets is how many times they've been polled since last fall, according to 538. Minnesota (with its 5.5 million citizens), which is currently ranked as the fourth most competitive state (relative to national popular vote) has only been polled 3 times, while Florida has been polled 38 times - and even New Hampshire (with only 1.3 million citizens) and North Carolina have been polled 26 times each.

Here's the number of times each state has been polled since last fall and up until August 25 (again according to 538):

38 - Florida
26 - New Hampshire and North Carolina
23 - Ohio
18 - Virginia
17 - Iowa, Wisconsin and New York
16 - Michigan and Georgia
15 - California
12 - Arizona
11 - Colorado
10 - Nevada, Utah and Missouri
9 - New Jersey and Illinois
7 - Indiana
6 - Kansas, South Carolina and Texas
5 - Oregon
4 - Maine, Idaho, Kentucky, Maryland, Massachusetts and Washington
3 - West Virginia, New Mexico, Mississippi, Connecticut, Oklahoma, Alabama, Minnesota and Tennessee
2 - Vermont, Alaska, Delaware, Montana, Arkansas and Louisiana
1 - Nebraska
0 - Wyoming, D.C., North Dakota, South Dakota, Rhode Island and Hawaii

If anyone are interested, I might update these statistics in a few weeks, for instance after the first debate.

[Doimper]

I've been informed by very good people that NH will only flip in a 45-state landslide scenario.

(Seriously, though - good thread)

[LLR]

something south carolina

[Lachi]

Good analysis, but I probably would have implied that it was based on the national average by putting something like that in the title, since it might confuse some people.

[Xing]

LOL, Maine's going to suddenly be a bellwether and Nevada's going to suddenly trend sharply R? Cute, Nate.

[Dr. Arch]

The models are now beginning to converge. Now-cast and Polls-only have been slowly descending to the Polls-plus model numbers, which have been averaging between 75 and 76%.