Founder & Editor-in-Chief:
Executive Editor:
This is a companion discussion topic for the original entry at https://talkingpointsmemo.com/?p=1437094
This is a companion discussion topic for the original entry at https://talkingpointsmemo.com/?p=1437094
I know I’m not paying much attention to polls lately. My guess is that they’re probably all over the map right now. I haven’t clicked on 538 or RCP for over a year now.
3 Likes
Polls have been wrong, but the key information for the article is missing: WHY are they wrong? There’s no discussion, no real analysis.
So, too long and not much worth reading and could be just 1 paragraph:
Polls have been wrong and nobody really knows why, so nobody has really figured out how to make them more accurate. Networks and Media will continue to publish “results,” but we recommend that if you care about the results of an election you ignore any stories based on polls and make sure you vote.
Even shorter: Ignore polls and stories based on polls, and get your ballot in!
1 Like
The Kansas abortion polling showed my why they are no longer reliable.
Poll shows tight race on Kansas abortion amendment, with supporters narrowly ahead as Election Day nears
Andrew Bahl | Topeka Capital-Journal | July 23rd 2022
A new poll released Wednesday shows a close race in the passage of a proposed anti-abortion amendment to the Kansas Constitution…
The actual results:
Kansas voters defeat abortion amendment in unexpected landslide
BY: SHERMAN SMITH AND LILY O’SHEA BECKER - AUGUST 2, 2022
OVERLAND PARK — Kansas voters in a landslide Tuesday defeated a constitutional amendment that would have stripped residents of abortion rights, defying polling and political observers who expected a close result.
5 Likes
Back in 1792, anyone who could vote pretty much did vote. Even if you were a farmer, a days ride away from the polling place, you could vote. [Apparently, Tuesday was chosen so you could go to church Sunday, and still make it to the polling place in time to vote]. And people probably weren’t inclined to lie about who they would vote for. If there had been any polling people in those days, they probably would have got things pretty accurate. Today, there are all sorts of obstacles in the way of voting, and people do lie. No surprise it is tough to figure out who will vote, and how.
1 Like
I imagine that the only folks still reliably participating in any kind of poll are the same guys who buttonhole the ladies who work the sample carts at CostCo and lecture them on why they shouldn’t have to wear a mask.
1 Like
It’s strange that this particular polling fail was left of the article. Not only is it significant, but it’s the most recent example and it shows that polling can fail in either direction.
5 Likes
I know why!
It’s because it’s no longer possible to reliably reach a representative sample of the people who are going to vote in the election, so the pollsters are attempting to compensate by constructing models of who and what they are predicting the turnout will be and then sampling respondents until they have enough representative persons to fill out their model.
But the models are inevitably wrong because we live in a completely unpredictable electorate and who the hell knows who’s going to show up anymore? So it’s just pollsters fudging the data because they don’t know either.
7 Likes
There’s probably a lot of confirmation bias going on, too. Most polls are sponsored by either the media or politicians and they have a vested interest in their narratives. On purpose or not, any poll that confirms what they want to believe is going to be more attractive than one that does not. Kansas is probably a good example: what people outside Kansas believe about Kansas is not necessarily what is true about Kansas.
It’s not like Florida where they’re all irrational crazies or Texas where everybody wears a big hat, drives a Cadillac with steer horns on the grill and owns an oil well.
1 Like
Two main issues that polling faces:
-
Predicting turnout in non-Prez elections is difficult and often impossible.
-
Not doing follow-up calls in telephone polls leads to bias. First responders are different from later responders.
WaPo: “The sovereignty that tech giants are carving out in international digital space is getting increasingly entangled in conflicts among sovereign nations.”
Indeed, this sovereignty must be defeated.
How?
The why is simpler. Russia’s partnership with Saudi Arabia is obvious. And Musk’s partnership with Russia is also clear.
It’s Russia again, stupid. Apologies to JC, James Carville.
1 Like
Personally, I’m starting to get the feeling that political strife is going to center more around exit polls than actual votes at this point. And that’s scary.
The GOP has essentially corrupted the polling aggregators this month the way they did in 2012 during Obama’s re-election. Media have bought into that narrative.
Despite all of that, the Dems still lead in the raw average of polls having October dates on 538 from 0.5% - 1.5% depending if you exclude GOP polls and likely voter polls. If you did a likely voter polls only, the GOP would lead by about 0.9%, which is far narrower than the hype.
In the House, one thing that has always caused me to be skeptical of a dominant GOP position is that there really aren’t any data to show that front line Dems in swing or lean Trump districts (e.g., Golden, Kaptur, Cartwright, Axne etc) are behind. I don’t see how the GOP mathematically gets a 20-40 seat gain without beating those front line Dems. In addition, the notion that Dems would hold tight/swing districts but lose D+7 to D+13 districts in an election where Dems are turning out to vote doesn’t seem that credible to me.
Where Dems could lose the election is for the GOP to pick up 10 gains through gerrymandering and some open seats and the Dems failing to offset by winning their fair share of open seat targets. In other words, the House is a true toss up.
The Senate is Lean D. Cortez Masto looks to be in a stronger position from 2 weeks ago. Hassan, Kelly and Warnock all lead in credible polling but the sheer avalanche of garbage GOP polls from the likes of Wick, Insider Advantage, Rasmussen et al have weighed down the averages and scared legit pollsters off. The truth is none of that stuff matters if we vote, and Dems are voting.
In addition to the incumbent Dem seats, the Dems seem to be ahead in PA (despite an avalanche of trash GOP polling) and are also right there in NC/OH/WI. Ryan and Barnes have looked particularly strong this month. Beasley seems to be polling well. She led +4 in a Survey USA poll of NC-13, a key House district that Biden won by +2. If she has that kind of incremental improvement over Biden, she’ll be the next Senator from NC.
Keep voting, folks. The Dems are in a position to win both the House and Senate.
13 Likes
Absolutely. My hope is that the Dems hold the house and senate, my fear is that the gerrymandering in the house loses that, and one or two lunatics win in the Senate (probably Walker and Oz, I think Ron Jon is beyond reach unfortunately)
I wouldn’t be at all surprised if it ends up with an overall +/-10% swing from the polling midpoint. Its a coinflip as to which side it goes, gas prices, the fear of crime vs Dobbs. Who knows?
2 Likes
Kyber, You hit on some important points, but I would also add when you try to correct a problem by addressing the wrong contributing factors, you are going to compound the original problem. Pollsters seam to be trying to correct/address the undercount of GOP in 2016 & 2020, but that was the symptom, not the underlying problem.
All polls are based on models, and models are based on past voting history plus assumptions. 2016 and 2020 saw a new (never before seen) phenomena injected into the political system - Trumpism. The problem pollsters should be looking at is how do we identify and account for something like that. While scientifically gibberish, Trafalgar’s process of adding 2 points for “shy Trump Voters” was the best solution to that. I don’t know the answer, but that is for polling science to figure out if the industry wants to survive. Rather than just adding more rural voters into their models, pollsters should be trying to model the impact of another phenoma that could be as large - the first time in modern history that Americans have had a right taken away from them.
As we sit here today, most of the high quality live polls (particularly in the Senate) show a large dispersion between registered voters and likely voters (5 - 6 point R lean). Yet 2018 broke records for midterm turnout, and the 2022 early vote so far is slightly exceeding that. However, registered D’s are meaningfully outpacing their 2018 total, and R’s are underperforming theirs. And in 2018, D’s effectively won the congressional ballot by 8.6 points (+ or - a point either way due to uncontested races). So to get a 14-15 point swing toward the R’s from 2018 when most polls are showing both D’s and R’s retaining 90% party support, you must be assuming that 1) D support will dry up, 2) R’s are going to blow away their 2018 numbers on election day, 3) Independents are breaking hard towards toward the R’s. I don’t see 1 happening, so how about some hard data on the independents and republican intensity as compared to 2018.
2 Likes
I don’t even read about the polls. Last week the headlines were that the Dems have an advantage as per the polls. This week suddenly the gop has the advantage. It goes back and forth until election day and then comes in the exit polls. Why should I tell some stranger who I voted for as I walk out of the polling place?
One troubling factor in polls is how they word their questions. Word placement and word choice has a big impact on your outcome. So does the choices you are given for an answer. And the respondent can only use those select choices. Nothing in between or none of the above.
Now the only people who care about the polls are the politicians.
Yes to the pollsters fudging data. But also I think the American electorate is finally not as static as expected and unfortunately the swings can happen in either direction in any location. I await Micheal Moore’s prediction but suspect several states will swing right.
1 Like
It may just be proof that lots of Dead Republicans always vote.
It’s that old “projection is confession” rule that Trump established early on in his rogue administration.
Whatever they accuse “us” of, they are probably guilty as hell of committing themselves.
Lots more Dead Republicans these days, too.
It has been hard on mean old, crotchety conservative types over the past couple years. Were they purged from the rolls when they expired, by whatever hoax or plague?
Or are their survivors still voting for them?
Seriously, that would explain everything.
And Republicans now peer-lie to the pollsters. Pretty simple.
1 Like
Lets hope is carries over to the November elections. If it does, better write a new list of poll questions and throw the old ones away.