Election models megathread

[GeorgiaModerate]

What happens in this timeline?

Trump told all his followers to vote on November 9.

[GeorgiaModerate]

Current models, with change from last update (Oct. 22).  The Economist models have been frozen since Oct. 16, but G. Elliott Morris said there will be more updates coming.


Senate D chances (most to least):

Economist 80 (not updated since Oct. 16)
538 Lite 66 (-3)
538 Classic 63 (-2)
JHK 59.2 (+0.6)
538 Deluxe 55 (-1)
DDHQ 52.3 (-3.6)


House R chances (most to least):

JHK 86.9 (+0.4)
538 Deluxe 81 (nc)
538 Classic 80 (nc)
DDHQ 78.5 (-0.6)
538 Lite 73 (nc)
Economist 65 (not updated since Oct. 16)

Continued movement (although at a slower rate) toward Republicans in the Senate, while the House was quite stable.

[GeorgiaModerate]

538 polling averages for key Senate races (change since Oct. 22):

AZ D+4.5 (-0.4)
CO D+8.8 (nc)
FL R+4.8 (+0.4)
GA D+3.0 (-0.7)
IA R+6.9 (nc) - added for OC
NH D+5.5 (-1.9)
NV R+0.1 (-0.1)
NC R+2.6 (+0.6)
OH R+1.7 (+0.
PA D+2.3 (-1.3)
UT R+7.8 (nc)
WI R+2.6 (-0.3)

Mostly small movements except in NH and PA, favoring R's more than D's.

[GeorgiaModerate]

I know that 538's Pollster rating are subjective, but what does it mean when the ratings are outlined by dots instead of a solid line (e.g. Wick under the PA-SEN race).

Provisional rating because they don't have enough qualifying polls since the last time 538 did the full ratings calculation.

[GeorgiaModerate]

Current models with change from last update (Oct. 26).  The Economist models have been frozen since Oct. 16, but G. Elliott Morris said there will be more updates coming.


Senate D chances (most to least):

Economist 80 (not updated since Oct. 16)
538 Lite 63 (-3)
538 Classic 61 (-2)
538 Deluxe 53 (-2)
JHK 51.2 (-8.0)
DDHQ 48.3 (-4.0)


House R chances (most to least):

JHK 88.4 (+1.5)
538 Deluxe 80 (-1)
538 Classic 80 (nc)
DDHQ 77.5 (-1.0)
538 Lite 71 (-2)
Economist 65 (not updated since Oct. 16)

Similar to the last update, continuing erosion for D chances in the Senate while the House remained mostly stable.

[GeorgiaModerate]

538 polling averages for key Senate races (change since Oct. 26):

AZ D+3.6 (-0.9)
CO D+10.8 (+2.0)
FL R+7.0 (+2.2)
GA D+1.2 (-1.8 )
IA R+7.1 (+0.2)
NH D+4.1 (-1.4)
NV D+0.4 (+0.5) - flipped back to D
NC R+2.6 (nc)
OH R+1.6 (-0.1)
PA D+1.0 (-1.3)
UT R+7.8 (nc)
WA D+7.6 (-0.3) - new to the list
WI R+3.4 (-0.8 )

A mixed bag with D leads in GA & PA getting uncomfortably tight, while they regained the lead in NV.  CO, FL, IA, UT, and WA are likely out of reach for the trailing candidate at this point.

[GeorgiaModerate]

https://twitter.com/UrsulaPerano/status/1586508454759260160

[GeorgiaModerate]

A few interesting tweets from G. Elliott Morris tonight:

https://twitter.com/gelliottmorris/status/1586800011148513281
https://twitter.com/gelliottmorris/status/1586848750521155586

[GeorgiaModerate]

https://twitter.com/SplitTicket_/status/1586749499350822913

[GeorgiaModerate]

The Economist models have finally updated, with their new probabilities at 52% D for the Senate and 74% R for the House.  G. Elliott Morris has a long technical explanation of why they paused the models on Oct. 16 and what they've done to fix them.

[GeorgiaModerate]

Did they? It's back up to 51-49 Dems. 

It was 50-50 a minute ago.

Minor fluctuations occur from one model run to another. 

[GeorgiaModerate]

Does anyone know what "adjusted poll average" means?

It's the RCP average plus the average poll underestimation.

But why are NV and AZ projected as GOP pick-ups when Dems lead the adjusted average?

I don't understand what "average poll underestimation" means.

It's another way for RCP to fudge the numbers.

[GeorgiaModerate]

Current models with change from last update (Oct. 30). 


Senate D chances (most to least):

538 Lite 56 (-7)
538 Classic 54 (-7)
DDHQ 47.6 (-0.7)
538 Deluxe 47 (-6)
JHK 46.8 (-4.4)
Economist 46 (-6)


House R chances (most to least):

JHK 92.9 (+4.5)
538 Deluxe 85 (+5)
538 Classic 84 (+4)
DDHQ 80.4 (+2.9)
Economist 78 (+4)
538 Lite 77 (+6)


Democrats continue to slide as the consensus now looks like a Likely R House and Tossup Senate.

[GeorgiaModerate]

538 polling averages for key Senate races (change since Oct. 30):

AZ D+2.6 (-1.0)
CO D+9.7 (-1.1)
FL R+7.1 (+0.1)
GA D+0.4 (-0.8 )
IA R+7.2 (+0.1)
NH D+2.8 (-1.3)
NV R+0.6 (+1.0) - flipped again
NC R+3.1 (+0.5)
OH R+2.3 (+0.7)
PA D+0.8 (-0.2)
UT R+9.8 (+2.0)
WA D+6.1 (-1.5)
WI R+4.0 (+0.6)

General movement toward Republicans, but I think the biggest takeaway is that there are a bunch of races within the margin of error.

[GeorgiaModerate]

https://fivethirtyeight.com/features/2022-polling-error/

Good article that starts from the following:

Since 1998, polls of U.S. Senate elections conducted within three weeks of Election Day have had a weighted-average error of 5.4 percentage points, and polls of U.S. House elections have had a weighted-average error of 6.3 points.

And examines what would happen with uniform errors of that magnitude in both directions.  Summary:

If current poll averages are exactly right: Senate 51R (+1), 49D; House 225R (+12), 210D

Polls underestimate R's by that amount: Senate 54R (+4), 46D; House 259R (+46), 176D

Polls underestimate D's by that amount: Senate 54D (+4), 46R; House 227D (+5), 208R

[GeorgiaModerate]

How come you didn’t just put this one in the model thread but did my article from Silver which was about more than the model in a way this isn’t?

TBH I didn't think about it, but this should reasonably be there too.  I'll merge it.

[GeorgiaModerate]

Here's the counterpoint to Nate Silver's case for a Republican wave that was published a couple days ago:

https://fivethirtyeight.com/features/the-case-for-a-democratic-surprise-on-election-night/

[GeorgiaModerate]

Current models with change from last update (Nov. 3). 


Senate D chances (most to least):

538 Lite 53 (-3)
538 Classic 51 (-3)
JHK 45.6 (-1.2)
538 Deluxe 45 (-2)
Economist 45 (-1)
DDHQ 44.5 (-3.1)


House R chances (most to least):

JHK 93.0 (+0.1)
538 Deluxe 84 (-1)
538 Classic 84 (nc)
DDHQ 79.1 (-1.3)
Economist 77 (-1)
538 Lite 76 (-1)


Democrats slide a bit further in the Senate, which now looks like a Tilt R consensus, while they improve by a hair in the House.

[GeorgiaModerate]

538 polling averages for key Senate races (change since Nov. 3):

AZ D+1.9 (-0.7)
CO D+9.7 (nc)
FL R+7.6 (+0.5)
GA R+0.1 (+0.5) - flipped
IA R+7.2 (nc)
NH D+3.2 (+0.4)
NV R+0.6 (nc)
NC R+3.8 (+0.7)
OH R+3.9 (+1.6)
PA D+0.4 (-0.4)
UT R+9.8 (nc)
WA D+6.1 (nc)
WI R+3.8 (+0.2)

No change or minor improvements for Republicans, except in NH.  The GA average strongly suggests a runoff coming.

[GeorgiaModerate]

https://twitter.com/Nate_Cohn/status/1589255493020979203

[GeorgiaModerate]

https://twitter.com/Nate_Cohn/status/1589255493020979203

Explain this for us dumb people, GM. 

Cohn, among others, was raising a warning flag in September when the Democrats were at their polling peak that Democrats were polling above expectations in precisely the areas where polls overestimated Democrats in 2020 -- so if the nonresponse problem still existed, that's exactly where you'd expect it to happen.  In short, the strong Democratic polling at that time was an illusion (in most places).

Now the disparity is no longer there, so the polls now look in line with expectations in most places; that is, Democrats are probably not overperforming expectations due to nonresponse now.  (They could of course be overperforming -- or underperforming -- for other reasons).

[GeorgiaModerate]

https://mobile.twitter.com/gelliottmorris/status/1589613622241947649

[GeorgiaModerate]

Final update!  Current models with change from last update (Nov. 5).


Senate D chances (most to least):

Race to the WH 50.3 (new)
538 Lite 50 (-3)
538 Classic 49 (-2)
JHK 46.3 (+0.7)
Economist 43 (-2)
538 Deluxe 41 (-4)
DDHQ 40.7 (-3.8 )


House R chances (most to least):

JHK 94.2 (+1.2)
538 Deluxe 84 (nc)
538 Classic 82 (-2)
DDHQ 79.7 (+0.6)
Economist 77 (nc)
538 Lite 75 (-1)
Race to the WH 66.9 (new)

Consensus: Tossup/Tilt R Senate, Likely R House

[GeorgiaModerate]

Final update! 538 polling averages for key Senate races (change since Nov. 5):

AZ D+1.5 (-0.4)
CO D+8.4 (-1.3)
FL R+8.8 (+1.2)
GA R+1.0 (+0.9)
IA R+9.6 (+2.4)
NH D+2.2 (+1.0)
NV R+1.4 (+0.8 )
NC R+4.3 (+0.5)
OH R+6.2 (+2.3)
PA R+0.5 (+0.9) - flipped
UT R+9.8 (nc)
WA D+4.7 (-1.4)
WI R+3.3 (+0.5)

Still a lot of races within the margin of error, but there's no question that Republicans are happier with these numbers than Democrats.

[GeorgiaModerate]

I know Kansas abortion referendum vote is likely not very predictive of today's elections for various reasons... But what was the difference in polling vs results?

There were only a few polls and they greatly underestimated "No", which ended up winning.