Trump poll average by state

State: July / Present

PA: 47.9/47.8
WI: 47.1/47.3
MI: 45.5/47.6
AZ: 47.7/48.8
NV: 47.9/47.5
GA: 46.6/48.4
NC: 47.2/48.2

Notice anything interesting? In July, when Biden was still in the race, undecideds were well into the double digits

Now, with idks hovering around 3-5, Trump hasn't improved his position much anywhere

And other than maybe AZ, he hasn't pushed his poll avg to anything unexpected

Ceiling = 🚩 (for Rs)

Here was my Senate 2020 forecast vs 538.

Our forecasts were similar in shape but I had much, much higher percentages in the middle.

They, as usual, have indefensible Astrapia tails which say, basically, there's about 1/5 chance that events of the 1/50 magnitude will happen.

https://twitter.com/RealCarlAllen/status/1363899677662445569

And here is 2020 vs 2022.

This time, not only are their tails crazy wide (1/5 chance highly unlikely events happen) the shapes of our forecasts are very different

In simplest terms, they thought Rs should be favored bc of PA and GA, and Ds only slightly favored in AZ and NH

Election Eve final analysis, pt 1 of x. See part 2 for great chart

Biden's lead is historic. Since June I've said it, and the "but TiGHTENInG" crowd has - perhaps rightfully - said uncertainty was too high to declare Biden overwhelming favorite.

Now? Ppl afraid to get hopes up. (2/x) Here's every Battleground race, Prez and Senate, since 04 with 2020 races highlighted.

I'm going to do analysis that gets a lot deeper in pt 3 & beyond. But if you go no further, understand this:

Biden isn't just leading, he's ahead by more than ppl think (if we all vote)

First, something I meant to post about last week re @FiveThirtyEight pollster ratings:

You see this "called correctly" line?

Calling races correctly is hard in swing states (we'd expect it to be) and many pollsters *only* do swing states

Why give incentive to poll +15 states? I tell you all about the overlap in Sports Data and Election Data, many don't believe me. But it's like this:

If I forecasted a bunch of games with a team favored by 10 points, my expected "called correctly percentage" would be much higher than if I only forecasted close games!

NFL stuff. As always in order of stuff I like most.

Carr o250.5 passyds +110
Pascal o3.5 recs +125
Mayfield o1.5 passtds -127
Green o4.5 recs +105
Mayfield o232.5 passyds +115
Hunt u3.5 recs -115
Burrow o26.5 completions +115
J Jefferson u4.5 recs -115
B Scott u3.5 recs -112

1/ Rodgers u282.5 passyds -112
Stafford u23.5 comps -104
D Johnson (Steelers) TD +265
Tannehill u1.5 passtds +195
Stafford u274.5 passyds -112
Witten TD +1600
Kittle o6.5 recs +115
E Elliott u3.5 recs -150
Dinucci o0.5 INT -240
Tannehill u22.5 comps +101
T Taylor (Browns) +2200

2/

Senate. Forecast.

The Presidential race has been relatively consistent, moving in Biden's favor as we approach the Election.

Senate has been everywhere.

Both GA seats are big time in play now. Last month ~35%. Now near coinflips.

NC was up to 70% D at one point, now 58%. A little analysis. First off - the odds the Republicans keep the Senate (~11%) are roughly in line with odds they keep the White House plus the small chance Biden wins but Collins, Ernst, Tillis, and/or GA folks hold their seats.

Far more likely Dems pick up most of those.

1/2 Margin or Vote%? Ideally both 😁

But if you can only choose one?

Vote share. As you saw from previous data: only ONCE has a candidate polling 49%+ lost that state (FL12ROMNEY)

Even 3-4 pt leads are VERY strong at 49-50%.

@FiveThirtyEight @RealClearNews @SeanTrende 2/2 Biden's current polls + leads compared w results back to 04

Trump's only lead is TX (IA, OH, GA, AZ are close)

What are your takeaways?

I was surprised at the cluster of 48-49 tiny leads that mostly won

AMA. #ElectionTwitter

@RealClearNews @FiveThirtyEight @SeanTrende

1/2 Since 2004, a candidate has polled at or above 49% in a battleground state and gone on to lose that state once. One time. #ElectionTwitter

Never in recent history (curious if ever?) - has a candidate polled over 50% and lost.

@FiveThirtyEight
@RealClearNews Sidenote before pt 2 (which I think is even crazier than pt 1):

Battleground states are, by definition, supposed to be close.

That doesn't always mean Tossup - it includes many lean blue or lean red states.

HOWEVER.

This is a strong piece of data to show 49%+ is DOMINANT

I have a really cool analysis I'm working on for #ElectionTwitter. Idk if it'll have cool charts or just numbers. But I'm counting on you guys for some retweets

Really hammers home the point that I (& we) have been sharing for the past several months

Share tonight or tomorrow? I'm not exaggerating this is one of the coolest pieces of research I've done. I really wish I had a graphics team to make it look cool but you all are gonna get some shitty excel chart and it'll still blow your mind.

I'm trying my best. I'll get it up tonight

Here's my Election Forecast and accompanying map.

I think IA (lean D) GA (lean D) along with TX (lean R) and OH (lean R) will be the closest states. Maybe too close to call on election night

But Biden wins FL, PA, MI and WI by comfortable enough margins that election is called. So some of you know but for those who don't, my background is in sports data.

When it comes to sports betting, it's not good enough to pick the winner, you need to be able to predict the spread too. So let's compare my final vote predictions to those of the top forecasters.

Just a reminder for my friends in forecasting - @jhkersting @gelliottmorris @NateSilver538 @ReedForecasts and I'm really sorry if I missed anyone -

That 2012 is the best comp for 2020.

2016 being most recent, most memorable, makes it *feel* more comparable, but that's a bias.

https://twitter.com/RealCarlAllen/status/1305624347110002688

And while I'm here, I need to issue an apology to @gelliottmorris.

At the beginning of my deep, deep dive into politics/polling, I had a big problem with @TheEconomist Forecast.

He may or may not even remember (and I wasn't *that* big of an a-hole..) but I should elaborate

Biden's polling at 49.1% in FL with ~4% undecided.

Trump won FL in 2016 with 49.02%.

Biden's polling at 49.4% in NC with ~4% undecided.

Trump won NC at 49.8% in 2016.

People are seriously underestimating how strong Biden is doing, and how much he's outperforming Hillary. Not just FL and NC.

Hillary's poll average wasn't above 47% in ANY SWING STATE. None. Do you know how crazy that is?

Biden is polling at 47.6% *IN GEORGIA*
Biden is polling at 47.5% *IN IOWA*

Hillary was polling at 45.1% *IN MAINE*
Hillary was polling at 44.9% *IN MINNESOTA*

The problem with 538's definition of "polling error" goes back to experimental design

That is, does this thing measure what we're saying it does?

In the case of "poll margin - election margin = poll error" they're assuming poll attempts to measure the final outcome. It doesn't! If you're not a science person, and you're like "what does this mean?" Here's how I break it down

Polls measure *preferences* of decided voters, and *how many* undecided voters. That's it!

Elections, unlike polls, don't include undecideds. This means variables have changed!

So here I'm going to start a Forecast thread

First up, the general election forecast. This is basically a blended model of my lean-Trump and lean-Biden undecided models, with weight to the lean-Biden because there's evidence to suggest the undecideds - while fewer - will break D Notably, in this forecast, Biden relatively easily wins the major swing states.

The most contentious states are Iowa, Ohio, Texas, and Georgia with Biden narrowly winning Iowa and Georgia and narrowly losing Iowa and Texas.

Remember, these are just probabilities, not concrete

1/x

Who's ready for an election thread and a (statistically literate) poll analysis & update?

There's an OBSESSION with "what the polls missed" in 2016. If you follow me, you'd know: the polls weren't wrong - people just read them wrong. 2/x

That's part of why @FiveThirtyEight's statistically invalid analysis of

"poll margin - election margin = poll error"

Is so damaging. Not only is it logically and statistically invalid, it leads them/the public to believe the POLLS were wrong when that's likely not true.

2012, not 2016, is the best comp for the 2020 Election.

Why? Undecideds.

2016 had an unprecedentedly high number of undecided voters, and polls don't capture whims of undecideds.

Currently, only ~10% of voters fit this category, roughly in line with 2012 numbers. In 2012, Obama polled +/- 1% of 50% in SEVEN swing states (@RealClearNews)

In 2016, Clinton polled +/- 1% of 50% in ZERO swing states

In 2020, Biden is polling +/- 1% if 50% in SIX swing states (@FiveThirtyEight)

Given this & comparable % of undecideds, 2012 is far better comp

1/x The polls weren't wrong, you just read them wrong.

The spread in a poll matters some. But a 50-47 lead (+3) is FAR MORE ROBUST than a 46-40 lead (+6)

People see a poll avg at 46-40, and when the result is 47-49, they say "pOlL wAS WrONg"

From +6 to -2?!? Off by 8!!!

Nope. 2/x

If the polls were "off" by 3 or 4 but the result of the election didn't swing because of it, no one would care, except pollsters. We wouldn't be talking about it.

Here's the most important question: what do polls measure?