538 model & poll tracker thread

[Dr. Frankenstein]

And we've gotten Biden +12, Biden +13, and Biden +9 this morning, and it's still at Biden +8.0.

Once again, for whatever reason, it's as if good Biden polls don't help all that much, but a bad one hurts way more than the good ones help.

There was two new PA polls today with Biden +9. There was also a Biden +7 poll, Nate has Biden only up  4.7 in PA.

So, in spite of some favorable polls, Biden's lead went from 5.4 on Sunday to 4.6 today, mainly due to the national lead shrinking from 8 to 6.9. Shouldn't state polls cancel that out?

[wbrocks67]

And we've gotten Biden +12, Biden +13, and Biden +9 this morning, and it's still at Biden +8.0.

Once again, for whatever reason, it's as if good Biden polls don't help all that much, but a bad one hurts way more than the good ones help.

There was two new PA polls today with Biden +9. There was also a Biden +7 poll, Nate has Biden only up  4.7 in PA.

So, in spite of some favorable polls, Biden's lead went from 5.4 on Sunday to 4.6 today, mainly due to the national lead shrinking from 8 to 6.9. Shouldn't state polls cancel that out?

Also, Nate tried to say before that national polls don't have that much weight, but they clearly do. Even so, shouldn't that effect the *forecast* though? His state averages model should literally just be *averages* without any other f**kery going on. I want an RCP aggregator but just not sh**tty.

[Landslide Lyndon]

And we've gotten Biden +12, Biden +13, and Biden +9 this morning, and it's still at Biden +8.0.

Once again, for whatever reason, it's as if good Biden polls don't help all that much, but a bad one hurts way more than the good ones help.

There was two new PA polls today with Biden +9. There was also a Biden +7 poll, Nate has Biden only up  4.7 in PA.

So, in spite of some favorable polls, Biden's lead went from 5.4 on Sunday to 4.6 today, mainly due to the national lead shrinking from 8 to 6.9. Shouldn't state polls cancel that out?

Also, Nate tried to say before that national polls don't have that much weight, but they clearly do. Even so, shouldn't that effect the *forecast* though? His state averages model should literally just be *averages* without any other f**kery going on. I want an RCP aggregator but just not sh**tty.

It really sucks that Pollster.com went bankrupt. They were my go-to aggregator.

[Gustaf]

And we've gotten Biden +12, Biden +13, and Biden +9 this morning, and it's still at Biden +8.0.

Once again, for whatever reason, it's as if good Biden polls don't help all that much, but a bad one hurts way more than the good ones help.

There was two new PA polls today with Biden +9. There was also a Biden +7 poll, Nate has Biden only up  4.7 in PA.

So, in spite of some favorable polls, Biden's lead went from 5.4 on Sunday to 4.6 today, mainly due to the national lead shrinking from 8 to 6.9. Shouldn't state polls cancel that out?

Also, Nate tried to say before that national polls don't have that much weight, but they clearly do. Even so, shouldn't that effect the *forecast* though? His state averages model should literally just be *averages* without any other f**kery going on. I want an RCP aggregator but just not sh**tty.

As far as I've understood there are no state averages.

There is a national polling average which is kind of its own thing. Then there are state projections based on polls and fundamentals for that state. There is also a national forecast that is created by aggregating the state forecasts.

The only role the national average plays is that it affects the impact of state polls. So Biden's current lead in Pennsylvania is weighted by polls there from say the last month and some fundamentals factor. If his national lead has gone down compared to 1 month ago his lead in a poll from a month ago is also adjusted down.

How big of an impact that is depends a bit, obviously. In general I'd guess a month old poll won't have much impact anyway, but now that we don't have that many polls in general I'm guessing it's more sensitive to those things.

[ProgressiveModerate]

And we've gotten Biden +12, Biden +13, and Biden +9 this morning, and it's still at Biden +8.0.

Once again, for whatever reason, it's as if good Biden polls don't help all that much, but a bad one hurts way more than the good ones help.

There was two new PA polls today with Biden +9. There was also a Biden +7 poll, Nate has Biden only up  4.7 in PA.

So, in spite of some favorable polls, Biden's lead went from 5.4 on Sunday to 4.6 today, mainly due to the national lead shrinking from 8 to 6.9. Shouldn't state polls cancel that out?

Also, Nate tried to say before that national polls don't have that much weight, but they clearly do. Even so, shouldn't that effect the *forecast* though? His state averages model should literally just be *averages* without any other f**kery going on. I want an RCP aggregator but just not sh**tty.

As far as I've understood there are no state averages.

There is a national polling average which is kind of its own thing. Then there are state projections based on polls and fundamentals for that state. There is also a national forecast that is created by aggregating the state forecasts.

The only role the national average plays is that it affects the impact of state polls. So Biden's current lead in Pennsylvania is weighted by polls there from say the last month and some fundamentals factor. If his national lead has gone down compared to 1 month ago his lead in a poll from a month ago is also adjusted down.

How big of an impact that is depends a bit, obviously. In general I'd guess a month old poll won't have much impact anyway, but now that we don't have that many polls in general I'm guessing it's more sensitive to those things.

Basically the lack of recent state polls will make the national environment have a larger influence because the polls were from longer ago.

[Sorenroy]

I was wondering about this:

https://twitter.com/micahcohen/status/1300499465518886914

Using the thresholds established in this Tweet, I decided to go through and look at what the different maps would be if Biden or Trump won every state at all the different levels.

Biden's 98th Percentile:

Joe Biden — 496 EV
Donald Trump — 42 EV

Biden's 90th Percentile:

Joe Biden — 434 EV
Donald Trump — 104 EV

Biden's 70th Percentile:

Joe Biden — 369 EV
Donald Trump — 169 EV

Biden's 55th Percentile:

Joe Biden — 334 EV
Donald Trump — 204 EV

Current Model Output:

Joe Biden — 319 EV (52.3%)
Donald Trump — 219 EV (46.4%)
Interestingly, the only true toss-up under Cohen's tweet is North Carolina, where Trump only has 53% chance of winning the state. Arizona is categorized on the 538 site as a toss-up as well, as Biden only has a 54.9% chance of winning the state, but going off the tweet, which uses rounded whole numbers, it falls into the category of slightly favored.

Trump's 55th Percentile:

Joe Biden — 319 EV
Donald Trump — 219 EV

Trump's 70th Percentile:

Donald Trump — 294 EV
Joe Biden — 244 EV

Trump's 90th Percentile:

Donald Trump — 337 EV
Joe Biden — 201 EV

Trump's 98th Percentile:

Donald Trump — 377 EV
Joe Biden — 161 EV

[Fmr. Gov. NickG]

Biden's odds have bumped back up to 70% in the model today, but for some reason he's still listed as only "slightly favored".

[Crumpets]

https://twitter.com/NateSilver538/status/1301190941110341632

[Fmr. Gov. NickG]

538 has added 13(!) distinct national polls to their database today.
Mean, median, and modal Biden lead are all 8 points.

[GeorgiaModerate]

The model has also moved back to Biden being "favored" from "slightly favored".

[Landslide Lyndon]

facepalm

[GeorgiaModerate]

https://twitter.com/NateSilver538/status/1301585415057608712

[TiltsAreUnderrated]

https://twitter.com/baseballot/status/1299525104842543104

[Sorenroy]

FiveThirtyEight released a "Model Talk" episode of their podcast for the first time since they originally published the model back in early/mid April. While I was listening to it, I had three key takeaways:


1. The model built in the effect of convention bounces. One of the most annoying things for me and the rest of the people on this forum has been the on-the-face confusing changes in the forecast over the past two weeks. The dip of Biden's odds to a low of 67% just didn't make sense when looking at how the polls, both nationwide and at a state level, were shifting. While Biden gained from around +8 to +9, his odds of winning the election shrunk from 72% to 67%. The reason for this is that the model was expecting Biden to go from +8 pre-Democratic convention to +10 post convention and then fall back to +7 post-Republican convention. Essentially, Biden's minor gains in the post-convention polling were not as much as the model expected, and thus Biden's chances went down. Now that the polling is basically in line with the forecasted estimate, the numbers are very close to what they were pre-either of the conventions.

2. There will be more information coming to the model. When trying to compare this year's model to 2016's, I was again frustrated by the lack of detail the 2020 model seems to have, especially at the state levels. In 2016, you could look at the polls, see how they were adjusted and fed into the model, and see the output as a series of additive elements at the bottom of each page. Now all that's been replaced by a series of unhelpful dots. While they did not say exactly what information they were going to add down the line, they did say that more of it was coming. This will be both in the form of information on the model itself as well as more detail in the spreadsheets downloadable at the bottom of the National page under "Model outputs."

3. House and Senate models are in the works but there will be no governor's model. While Nate Silver did not give a time frame for when they would come out besides soon^TM, he did say that a House and Senate model were on there way and that they were even at a point for a first test spin that unsurprisingly showed that the House would be noncompetitive this year. Additionally, he said that there would not be a gubernatorial model for any state. He did mention that the 2022 model will likely have all three of the House, Senate, and governor's races, as well as possibly more information around all of them as the midterm crowd is a different audience.


TL;DR: Built in convention adjustments are to blame for some of the downward shift in Biden's odds, more information will be coming to the models over time, and House and Senate models are in the works while there will be no governor's model this year.

[wbrocks67]

I mean at this point I don't even care about a house or senate model, i just want aggregate polling pages because again, RCP doesnt stack up

[GeorgiaModerate]

https://twitter.com/micahcohen/status/1301895482902482944

[SnowLabrador]

Now we're talking!

[afleitch]

The economics/incumbency 'boost' is new to the model this year?

[Fmr. Gov. NickG]

The new info on the state pages is actually pretty helpful.

It looks like the “economics and incumbency” portion always constitutes 18% of the projections right now, and basically always favors Trump.  I believe this portion will gradually decline to 0 as we get closer to the election.

Of the remaining 82%, it look pretty evenly split betwwen polls versus demographics on average, though it tilts in favor of polls in swing states with more polling.  Unlike the “economics and incumbency” factor, it doesn’t seem like the demographics factors consistently favors on side or the other.

In also looks like the model always allocated undecided voters exactly evenly.

[Ferguson97]

The new info on the state pages is actually pretty helpful.

It looks like the “economics and incumbency” portion always constitutes 18% of the projections right now, and basically always favors Trump.  I believe this portion will gradually decline to 0 as we get closer to the election.

Of the remaining 82%, it look pretty evenly split betwwen polls versus demographics on average, though it tilts in favor of polls in swing states with more polling.  Unlike the “economics and incumbency” factor, it doesn’t seem like the demographics factors consistently favors on side or the other.

In also looks like the model always allocated undecided voters exactly evenly.

I'd sort of reject the premise that incumbency should always be considered beneficial.

Incumbency didn't help Herbert Hoover.

[wbrocks67]

538 national average: Biden +7.4

Polls added since last night: Biden +9, Biden +11, Biden +9, Biden +13, Biden +6

538 national average today: Biden +7.4

Doesn't seem to make much sense.

[GeorgiaModerate]

538 national average: Biden +7.4

Polls added since last night: Biden +9, Biden +11, Biden +9, Biden +13, Biden +6

538 national average today: Biden +7.4

Doesn't seem to make much sense.

Depends on what the house effects of those pollsters are; 538 adjusts for them.  Depends on what previous polls by those pollsters were included in that 7.4 average, and will now get less weight. For example, if one of the +9 polls supersedes a +12 by the same pollster, it won't help Biden's margin.

[Sorenroy]

https://twitter.com/micahcohen/status/1301895482902482944

You'll notice that one of those lines looks suspiciously like the nowcast from 2016. So I decided to make a map out of it. Not the most interesting given that it makes up over 80% of the current model, but there are a handful of slight differences. Since the nowcast is not an official thing anymore, we don't get state by state odd-based percentages, but we can still look at margins to find safe, lean, and tossup states.

Nowcast Output:

Joe Biden — 334 EV (53.4%)
Donald Trump — 204 EV (45.2%)

Safe — >10% Lead
Lean — >5% Lead
Tilt — ≤5% Lead


Current Model Output:

Joe Biden — 319 EV (52.9%)
Donald Trump — 219 EV (46.2%)

Safe — >10% Lead
Lean — >5% Lead
Tilt — ≤5% Lead

The only four states that are affected by a change from the Nowcast to the Current Model are Colorado (safe to lean), Maine (safe to lean), Missouri (lean to safe), and Pennsylvania (lean to tossup).

[American2020]

https://twitter.com/NilesGApol/status/1302672082657054727

[GeorgiaModerate]

For everyone overreacting to any of the polls today:

https://twitter.com/NateSilver538/status/1303482618709790722