Polling can be a useful tool. We need to look at it differently

[ProgressiveModerate]

Whenever we discuss polls, folks always talk about the margin. "Fetterman is up 8 over Oz". However, time and time again we see Republicans consistently outperform the margins in well polled states. In 2020, Trump and Senate Republicans in key races outperformed polls by anywhere from a negligible amount to 10 points. That's a huge range and means that just because a D is up 9 points in a polling average, someone can rightfully make an argument that the polling is off 10 points and hence Rs will win just based off precendent.

For this reason, looking at the topline margin IMO isn't helpful unless someone is consistently up double-digits in a polling average.

However, looking at the % candidates are polling at is very useful. One theme I've found across looking back at polling errors in 2016, 2018, and 2020, is that in any well polled state it's extremely rare for a candidate to get much lower than % they're polling at. Let's use 2020 Pres as an example cause it's recent, well polled, and many folks considered the polling errors to be masive in Trump's favor (Polling average are 538 final):

National: 51.8% polling/51.3% actual
Georgia: 48.5%/49.5%
Arizona: 48.7%/49.4%
Wisconsin: 52.1%/49.5%
Pennsylvania: 50.2%/50.0%
North Carolina: 48.9%/48.6%
Nevada: 49.7%/50.0%
Michigan: 51.2%/50.6%
Florida: 49.1%/47.9%
Texas: 47.4%/46.5%
Minnesota: 51.8%/52.4%
New Hamster: 53.9%/52.7%
Ohio: 46.8%/45.2%
Iowa: 46.3%/44.9%
Maine: 53.3%/53.1%
Virginia: 53.7%/54.1%
New Mexico: 53.9%/54.3%
Colorado: 53.6%/55.4%

ect.

As you can see in basically every state Biden nearly matched the share of the votes he was expected to get, but "undecides" broke heavily for Trump. This is true even in states where the margin was off by high single digits such as PA and MI. The worst offender on this list is WI where Biden underperformed his share by 2.5%, but that's still significantly better than the nearly 10 point margin error in the average.

This general logic works for 2016 and 2018 as well. You'd be extremely hard pressed to find a state where a candidate got a significantly lower share of the vote than they were polling at, even if the topline margin was way off.

So a better rule of thumb than just swinging the margins "x" points to Rs may be to give the Republicans all the "undecides" and assume the Dem will roughly match the % they're polling at. Even in a race like PA-Sen where Fetterman holds a wide 9 point lead over Oz, he's still only polling at 48% meaning that if the election were held today, there'd still be adequate chance for Oz to catch up. Generally, as the election nears the number of "undecides" will shrink and Fetterman's % may go up, and if he's at 51 or 52% on e-day, that's a really solid position to be in whereas 49% is iffy.

Thoughts?

[ctherainbow]

New Hamster?   xD

But yes, this seems a relatively reasonable way of interpreting polling numbers.

[Darthpi – Anti-Florida Activist]

Yeah, I went through like a day or two after the election in 2020 and found the same thing. Anyone who says they are undecided in, like, the last week or two before the election in our modern, hyper-polarized system is probably lying.

[soundchaser]

Petition for a "New Hamster" state avatar to be added to the board.