This is my beef with some of you scoffing at the polls

[Kamala's side hoe]

If polls didn't exist, we would all think that Biden was a strong favorite for re-election.

What? I would’ve guessed the opposite, since no one seems to be pleased with Biden’s job performance (or his age) right now.

This would be a more convincing talking point were Trump and RFK not also senile and kooky in different and arguably worse ways.

[Absentee Voting Ghost of Ruin]

You guys are essentially arguing that polling doesn’t have predictive power if it defies some arbitrary, pre-defined baseline.

Speaking for myself, I'm arguing that polling doesn't have predictive power if it fails to demonstrate predictive power. Look at the recent Maryland  Democratic Senate primary: Only 1 poll (the final poll last week) got the winner right, and it was still off by nine points. That's a margin of error that makes Boeing quality control look superb by comparison.

Such a view of polling completely nullifies its very intent: the measure of the views of a sample of people and their shifts.

This has profound implications even in survey science outside politics. If a snack company took regular surveys monthly on their products and, for the first time ever, started to notice that a given snack began to decline in popularity, would the company ignore the survey, or see what’s going on? Under the logic of many on this board, the company should ignore the survey, because it defied the old results.

Polling means nothing without empirical data telling us whether it was right or not. When polling was reliably precise and accurate it made sense to use it as a tool. It currently does not appear to be either.

If a snack company takes regularly monthly surveys among consumers, and the surveys do not correspond to their sales numbers for the same timeframe, then yes, they should absolutely ignore the survey, because it at best utterly inadequate for their purposes.

Organizations, firms, and presidential campaigns spend large amounts of money on polling and survey data, not to pop-poo the data when it comes in, but to take it seriously and respond to it.

People spend money on a great deal of complete hogwash (see the likely Republican nominee's business history as a case in point). Either polling produces useful results or it does not - that people spend money on something that was previously believed to useful does not mean it still is.

Polling is either capable of capturing shifts in behavior or it isn’t. To argue that polling is unreliable if it captures shifts is to argue against the usefulness of polling itself.

If polling cannot predict results within a useful margin of error, there is no reason to believe such "captured shifts" are relevant, nor that they even exist as reported.

Personally, my take on this election is that eventual Trump voters have overwhelmingly already made up their minds, and that Biden will win or lose based on what people who have no interest in making up their minds before Labor Day (at the earliest) decide to do in November. And I've yet to see any convincing evidence that current polling can usefully predict how they will decide, or what will inform their decisions. Hell, most polling I see explicitly avoids such voters in favor of pushing the "horserace" narrative, which is a fundamentally broken (and arguably dishonest) framing device.

[Zenobiyl]

You guys are essentially arguing that polling doesn’t have predictive power if it defies some arbitrary, pre-defined baseline.

Speaking for myself, I'm arguing that polling doesn't have predictive power if it fails to demonstrate predictive power. Look at the recent Maryland  Democratic Senate primary: Only 1 poll (the final poll last week) got the winner right, and it was still off by nine points. That's a margin of error that makes Boeing quality control look superb by comparison.

Such a view of polling completely nullifies its very intent: the measure of the views of a sample of people and their shifts.

This has profound implications even in survey science outside politics. If a snack company took regular surveys monthly on their products and, for the first time ever, started to notice that a given snack began to decline in popularity, would the company ignore the survey, or see what’s going on? Under the logic of many on this board, the company should ignore the survey, because it defied the old results.

Polling means nothing without empirical data telling us whether it was right or not. When polling was reliably precise and accurate it made sense to use it as a tool. It currently does not appear to be either.

If a snack company takes regularly monthly surveys among consumers, and the surveys do not correspond to their sales numbers for the same timeframe, then yes, they should absolutely ignore the survey, because it at best utterly inadequate for their purposes.

Organizations, firms, and presidential campaigns spend large amounts of money on polling and survey data, not to pop-poo the data when it comes in, but to take it seriously and respond to it.

People spend money on a great deal of complete hogwash (see the likely Republican nominee's business history as a case in point). Either polling produces useful results or it does not - that people spend money on something that was previously believed to useful does not mean it still is.

Polling is either capable of capturing shifts in behavior or it isn’t. To argue that polling is unreliable if it captures shifts is to argue against the usefulness of polling itself.

If polling cannot predict results within a useful margin of error, there is no reason to believe such "captured shifts" are relevant, nor that they even exist as reported.

Personally, my take on this election is that eventual Trump voters have overwhelmingly already made up their minds, and that Biden will win or lose based on what people who have no interest in making up their minds before Labor Day (at the earliest) decide to do in November. And I've yet to see any convincing evidence that current polling can usefully predict how they will decide, or what will inform their decisions. Hell, most polling I see explicitly avoids such voters in favor of pushing the "horserace" narrative, which is a fundamentally broken (and arguably dishonest) framing device.

Exactly this.

[cherry mandarin]

Are you basing that on polls or personal experience? I would definitely think the election was likely Biden if not for polls and just going on things like the economy, special elections, primaries, and Trump's trial + lackluster campaign compared to his previous runs.

I’m trying my best to deduce how I would handicap this election if we had no horse race polling right now. It’s true that Trump fatigue, his scandals and criminal trials would weigh him down relative to 2020, but in my personal experience, swing voters seem to care about that a lot less than Atlas Dems seem to believe.

Voter attitudes on the state of the economy are bleak, though I suppose we wouldn’t know that if there were no polls. In the absence of polling, I’m mostly going by vibes to make my prediction here.

Sure, special election results have been atrocious for Republican candidates lately, but they usually aren’t very predictive of the result of the subsequent presidential election. Personally I’d place more of an emphasis on the demographic trends of the past decade or so, which generally seems to favour the GOP right now.

Lol give it a rest Vaccinated Bear we know who you are

😭😭😭😭

Turnout should drop but it's pretty unlikely to drop this much. In particular, black turnout collapsing this much is unlikely.

The crosstabs for most national polls right now don’t seem to assume black turnout levels dropping much from 2016 or 2020 levels.

[DaleCooper]

My gripe is the same people downplaying the polls here were worshipping them 4 years ago. If you came to Atlas in June 2020, when NYT/Siena had Biden+14 and him up double digits in MI/PA/WI all the red avatars were worshipping them. "Polls are so much better this time, Trump is indeed finished", It was an absolute sh*thow last election cycle and those of us who doubted the polls were mocked for it. Now when Biden is trailing behind (even just narrowly) they are starting to bring out the nuance and hand waving. And I agree with that to an extent, but where was that in 2020? Let's not kid ourselves, if the polls showed Biden up 5 in the national average and him leading in every swing state, Atlas would be taking them at face value like they did four years ago. And this isn't just Atlas, it applies to Twitter, and other pundits.

Sure the polls aren't a great indicator and not worth a lot of attention. But why wasn't this being said in 2020? Why only now when Biden is down?

Apparently they learned the right lesson from 2020.

[Amenhotep Bakari-Sellers]

My gripe is the same people downplaying the polls here were worshipping them 4 years ago. If you came to Atlas in June 2020, when NYT/Siena had Biden+14 and him up double digits in MI/PA/WI all the red avatars were worshipping them. "Polls are so much better this time, Trump is indeed finished", It was an absolute sh*thow last election cycle and those of us who doubted the polls were mocked for it. Now when Biden is trailing behind (even just narrowly) they are starting to bring out the nuance and hand waving. And I agree with that to an extent, but where was that in 2020? Let's not kid ourselves, if the polls showed Biden up 5 in the national average and him leading in every swing state, Atlas would be taking them at face value like they did four years ago. And this isn't just Atlas, it applies to Twitter, and other pundits.

Sure the polls aren't a great indicator and not worth a lot of attention. But why wasn't this being said in 2020? Why only now when Biden is down?

Apparently they learned the right lesson from 2020.

Wrong if you look at the 22 Database it's full of Inside Advantage polls and Trafalgar polls that got races wrong

[cherry mandarin]

There is a lot of good reason to think Biden is in good shape if the polls didn't say otherwise.

Yes but there’s also a number of compelling arguments that Trump could win, some of which I just listed above.

This would be a more convincing talking point were Trump and RFK not also senile and kooky in different and arguably worse ways.

Yes but my point was about the sitting president and not his challengers. As POTUS Biden gets the bulk of the media attention and rightfully so. As such, the criticisms from the public will follow.

None of this is to say this will continue until November, since most people of course aren’t paying close attention to the race yet.

Literally the only indicator pushing for Trump winning by high single digits or anything more than a practical tie is polls. As far as I'm aware anyhow. And polls aren't so singularly important as to carry such a case by themselves.

And Biden has a history of being underestimated as well.

And the polls show a near-tie in the NPV right now, with either a statistical tie or Trump leading by 2-3 points at most in the tipping-point state on average. (Right now the tipping-point state is either MI or NV if you go by polling.)

So it’s not like the polls show an overwhelming Trump lead at all.

To your other point, Trump has a far more consistent track record of being underestimated (including in his one race against Biden) than Biden does.

I’ll admit I was one of those people who bought into the Biden landslide idea in 2020 and the red wave in 2022 based on polling. I learned my lesson, so why wouldn’t I be more skeptical of polls this time around?

This is kind of revisionist history at play here, because in summer/fall 2022 people were all debating between whether to believe the red wave narrative due to Biden’s unpopularity, or the polls which contradicted that narrative and showed Dems holding up well.

Those who bought into the red wave theory largely did so in spite of what the polls were showing, as they believed the GOP would beat their polling in the same way they did in 2016, 2020, and to a lesser extent 2018.

The truth is, it’s way far out still. This time in 1988, Michael Dukakis was leading by double digits but he lost in a landslide. This time in 92, Ross Perot was going to be president according to the polls. This time in 2008, McCain and Obama were tied and McCain actually led some polls. Kerry was leading in 04. Bush was far outpacing Gore this time in 2000 but won by a small handful of votes in Florida. Not to mention Hillary in 2016.

What’s accurate in May has little bearing on November.

It’s not 1988 or 1992 anymore. Polarization has fundamentally altered the state of the electorate.

Just look at how little fluctuation there was in the 2020 presidential GE polls, even those going back to late 2019 or earlier, and compare that to the wild roller coaster that the polls gave you for presidential elections a few decades ago by comparison.

And I do think 2020 is indicative not just because it was the most recent election but more importantly featured the same two candidates as this year. Not only that, they’re the two most recent presidents and very well-known as a result. Everyone has very strong opinions on Biden and certainly on Trump, and these perceptions are dug-in and won’t budge significantly between now and November.

When it comes to presidential campaigns these days, it tends to get late early.

[All Along The Watchtower]

Yes, it's bad news that Biden is losing in the polls of an election that is taking place in less than a month. It's mid-October right now, yes?

On a more serious note, I think it will be close (again) because we have a rather deeply and closely divided country, if both presidential and congressional election results in recent years/decades are anything to go by. Polls are useful snapshots, but they are snapshots, and neither side should take anything for granted.

[cherry mandarin]

American electoral polling on the whole (presidential, gubernatorial, senatorial, congressional / presidential, midterm, and off-year cycles / primary and general elections) has literally never been more accurate over a five-to-ten year span than it has been over the past decade or so.

It’s just that there’s a lot more people paying close attention now, and way more polls of way more races, so it’s far more common for people to pinpoint and cherry-pick individual races where big misses are made.

But if you look on the whole, American election polls actually remain highly accurate. People just need to remember that candidates can easily win races even when they’re trailing by a handful of points in the polling averages.

[All Along The Watchtower]

American electoral polling on the whole (presidential, gubernatorial, senatorial, congressional / presidential, midterm, and off-year cycles / primary and general elections) has literally never been more accurate over a five-to-ten year span than it has been over the past decade or so.

It’s just that there’s a lot more people paying close attention now, and way more polls of way more races, so it’s far more common for people to pinpoint and cherry-pick individual races where big misses are made.

But if you look on the whole, American election polls actually remain highly accurate. People just need to remember that candidates can easily win races even when they’re trailing by a handful of points in the polling averages.

The bolded is debatable, at least for this year.

[cherry mandarin]

This is debatable, at least for this year.

What I’m trying to say is that there are a lot more poll followers who scrutinize and even pick apart the crosstabs for seemingly every survey result that gets released, more so than back in the 1970s, 90s, or even the early aughts.

[Arizona Iced Tea]

My gripe is the same people downplaying the polls here were worshipping them 4 years ago. If you came to Atlas in June 2020, when NYT/Siena had Biden+14 and him up double digits in MI/PA/WI all the red avatars were worshipping them. "Polls are so much better this time, Trump is indeed finished", It was an absolute sh*thow last election cycle and those of us who doubted the polls were mocked for it. Now when Biden is trailing behind (even just narrowly) they are starting to bring out the nuance and hand waving. And I agree with that to an extent, but where was that in 2020? Let's not kid ourselves, if the polls showed Biden up 5 in the national average and him leading in every swing state, Atlas would be taking them at face value like they did four years ago. And this isn't just Atlas, it applies to Twitter, and other pundits.

Sure the polls aren't a great indicator and not worth a lot of attention. But why wasn't this being said in 2020? Why only now when Biden is down?

Sure, there’s bias at work when people decide whether to believe a poll.

I’ll admit I was one of those people who bought into the Biden landslide idea in 2020 and the red wave in 2022 based on polling. I learned my lesson, so why wouldn’t I be more skeptical of polls this time around?

In 2016 I expected a D landslide. In 2018 I expected fewer incumbents to lose than who actually did. In 2020 I assumed Biden would outperform polling by a few points. In all three cases I was off in some way.
The lesson I took from 2020 was that this nation was too polarized for a landslide election. In 2022 I never believed in a red wave and my prediction for Congressional topline numbers was pretty much identical to the results when everything was counted.
I see no evidence the country is significantly less polarized than in 2020 or 2022. I have since taken to assume very little net change from 2020 in most of the country. Regardless of what the polls say. Now we're past Labor Day I'm willing to consider taking into account polling, however the actual evidence we'd see mass drops in turnout or relatively huge scale of voter fleeing from Biden has to contend with special elections being great for Democrats.
Literally the only indicator pushing for Trump winning by high single digits or anything more than a practical tie is polls. As far as I'm aware anyhow. And polls aren't so singularly important as to carry such s case by themselves. And the thought Trump might actually be competitive among Black males is still unproven at best.
Biden's position isn't amazing but it's hardly awful. And Biden has a history of being underestimated as well.

While its still somewhat hilarious to laugh at the people who fell for 2016, I don't blame anyone at all for thinking Hillary would romp. Even I bought into them In 2012, polls weren't too bad and did underestimate Obama and it was unknown how Trump would actually be perceived at the ballot box. So it was still fair game imo. However in 2020, we knew the polls were off and there was a significant Trump supporter undersampling. A simple formula trick of "weighting for education" was never going to fix such massively internally flawed surveys. That's why I think the 2020 poll believers don't get the same pass the 2016 ones do.

[cherry mandarin]

While its still somewhat hilarious to laugh at the people who fell for 2016, I don't blame anyone at all for thinking Hillary would romp. Even I bought into them In 2012, polls weren't too bad and did underestimate Obama and it was unknown how Trump would actually be perceived at the ballot box. So it was still fair game imo. However in 2020, we knew the polls were off and there was a significant Trump supporter undersampling. A simple formula trick of "weighting for education" was never going to fix such massively internally flawed surveys. That's why I think the 2020 poll believers don't get the same pass the 2016 ones do.

What’s your take on the 2024 polls Minute Maid?