Muh Polls

788,334 Views | 5744 Replies | Last: 15 hrs ago by ts5641
Captn_Ag05
How long do you want to ignore this user?
AG
Captn_Ag05 said:




This poll has Harris up 6 in the battleground states which is strange for a trump plus 2 national poll.
SA68AG
How long do you want to ignore this user?
AG
The Fox poll has always been crap. There's no way she's up 6 in the battleground states.
will25u
How long do you want to ignore this user?

will25u
How long do you want to ignore this user?
will25u
How long do you want to ignore this user?
Don't know if someone posted this or not.

Captn_Ag05
How long do you want to ignore this user?
AG
Yes both polls you posted have been on the prior page.
will25u
How long do you want to ignore this user?
Are we in the polls doom loop with Harris? She panicked when the polls were showing her slipping.

So she had to get out of the basement and do interviews. But has done horribly in those interviews. Which has made her polls suffer more. Rinse repeat...

The queen has no clothes?

Trump meanwhile has been killing the interviews and is in his element.

[This is relevant to this thread's discussion on polls but is also ripe for multiple potential derails. Keep the discussion focused on polls. Any responses that get into Kamala's or Trump's campaign strategy will be removed, and potentially this post if it generates persistent derails -- Staff]

4
How long do you want to ignore this user?
AG
Captn_Ag05 said:

Captn_Ag05 said:




This poll has Harris up 6 in the battleground states which is strange for a trump plus 2 national poll.

And it's an outlier.

Every other swing state poll out there shows Trump ahead in all but one or two states, and those are essentially even.

Don't know who Fox was polling on this one...
nortex97
How long do you want to ignore this user?
AG
Those who approve of Trump far outnumber those who approve of Harris' job performance.





nortex97
How long do you want to ignore this user?
AG


Pretty tough to materially change this election at this point with this data set.
texagbeliever
How long do you want to ignore this user?
Just your friendly reminder that the right direction / wrong track has correctly predicted the presidential winner since 1980.
Currently the democrats are -35 or so points.
Quo Vadis?
How long do you want to ignore this user?
Volume about to hit a ridiculous $2,000,000,000 as Trump skyrockets to a +23.4 over Kamala as of right now.


Captn_Ag05
How long do you want to ignore this user?
AG
Captn_Ag05
How long do you want to ignore this user?
AG
Captn_Ag05
How long do you want to ignore this user?
AG
Captn_Ag05
How long do you want to ignore this user?
AG
1836er
How long do you want to ignore this user?
AG
Captn_Ag05 said:


Friendly neighborhood reminder.

Many of these polls - such as YouGov, ABC, Reuters, Yahoo, and the Economist polls - utilize the same data sets, which heavily utilize opt-in panels of voters that skew overeducated and over-credentialled (meaning more Democratic) compared to the electorate.

Rather than doing their own actually polling, they buy same data sets (like IPSOS), and have their own "pollsters" weight the polls ever so slightly differently, and then publish them as their own; pretending that they are different polls.

They really should be treated like a single poll, therefore, rather than viewed as evidence of a bunch of polls trending or "herding" in the same direction.
Vance in '28
nortex97
How long do you want to ignore this user?
AG
LOL, Trump's not winning Indies by 5 and still losing PA. That's genuinely hilarious.
1836er
How long do you want to ignore this user?
AG
nortex97 said:

LOL, Trump's not winning Indies by 5 and still losing PA. That's genuinely hilarious.
The only ways they could get to that are by way overweighting Democrats and having a Republican sample that is not representative of the actual Republican electorate.
Vance in '28
txags92
How long do you want to ignore this user?
AG
1836er said:

Captn_Ag05 said:


Friendly neighborhood reminder.

Many of these polls - such as YouGov, ABC, Reuters, Yahoo, and the Economist polls - utilize the same data sets, which heavily utilize opt-in panels of voters that skew overeducated and over-credentialled (meaning more Democratic) compared to the electorate.

Rather than doing their own actually polling, they buy same data sets (like IPSOS), and have their own "pollsters" weight the polls ever so slightly differently, and then publish them as their own; pretending that they are different polls.

They really should be treated like a single poll, therefore, rather than viewed as evidence of a bunch of polls trending or "herding" in the same direction.
I have a question since you seem quite knowledgeable about the way these are done. We keep hearing about how the polls were skewed D in 2016 and 2020 and people are questioning whether that has been fixed in this go round. How exactly would the pollsters go about correcting for that bias and still be true to the data? I am assuming they can't just say "we used the same methodology as previous and then applied a +4% R fudge factor".

So where would the adjustment be made if the flaw is in the raw data they are(n't) receiving from the population responding to the polls? Shouldn't it be possible to look at how they are weighting their various demos and see if they made an adjustment from 2020 and 2016? Or is everything so screwy now that it is impossible to tell? I keep seeing things that don't make sense and wonder if their adjustments are actually working the opposite of the way they think.

As was pointed out previously, if Trump is getting 15-20% of the black vote and +5 with independents, it just seems hard to see how Kamala would still be in the lead with what we have always previously assumed about the influence of those demos.
Captn_Ag05
How long do you want to ignore this user?
AG


Philip J Fry
How long do you want to ignore this user?
AG
I had zero idea they did that. Sounds unethical as all get out
1836er
How long do you want to ignore this user?
AG
I know less about this stuff than I'd like to, but if I were one of the media/university polls that consistently skews 4-8 points too Democratic in the age of Trump, and I was genuinely interested in more accurate polling... I'd need to figure out ways to more effectively reach the lower propensity working class voters in the rust belt and swing states that we keep missing.

These people, especially white voters who fit this profile, are the hardest to reach. From what I've heard, these voters are 50 times less likely to respond to a pollster than high propensity/highly educated voters, regardless of ideological or party preference.

Unfortunately, the only truly effective way I can think of to reach enough of these voters to include them as an accurate representation of the electorate in your sample, is to spend 50 times as much time, 50 times as much effort, and most importantly 50 times as much money to survey them, as it takes to survey your high propensity/highly educated voters.

And for these hardest-to-reach voters the most effective way to reach them (more so than online panels, for example) is still a good old fashioned person to person phone call, which is, of course, the most expensive way to survey voters, at a cost of around $35 or more for a single completed survey. If these people are 50 times harder to reach, imagine how much more money it costs to poll a comparable number of them compared to your leftwing professor who always votes and actually likes sharing his opinions with a pollster.

For most of the media/university polls, even if they wanted to do this, they don't have the funds to spend on it like an internal pollster financed by a campaign might.

And even if i do manage to survey enough of them, because they are so much harder to reach and to survey, I would have to overcome my natural inclination (based on the way they used to be pre-Trump) to dismiss them as unlikely voters anyway.
Vance in '28
txags92
How long do you want to ignore this user?
AG

Yeah, that is what I was kind of getting at. If the problem is that they are just not getting a representative data set because there is a portion of the electorate that just won't respond to attempts to survey them, how would they go about "correcting for" that portion they are unable to survey if they assume that missing data is responsible for the skew in their previous polls? If you want to be true to the data you have, then you can't just throw a fudge factor into everything to make up for it. If it is too expensive and time consuming to go chase down those reluctant responders, how else do you try to make your data reflect reality?

People keep saying the polls skewed hard left in 2016 and 2020 relative to the actual election results...and then speculate on whether the pollsters have "fixed the glitch" in the current polls. I would think if they were changing the weighting or adding a factor to account for it, it should be apparent when you compare 2016 and 2020 weighting against 2024. I haven't seen anybody showing any evidence of them doing so in this round of polls, so my assumption is that they haven't done it. But maybe that is just me hoping for a blowout win.

[Removed previous quote to avoid cluttering the page. No edits to this post -- Staff]
1836er
How long do you want to ignore this user?
AG
Yes, I think your assumptions are basically right. IF they aren't able or willing to spend the money on reaching those voters there just isn't a really good way to make up for it... just less bad ways (like better weighting compared to worse weighting).

Once of the things that Baris does (I don't think he weights by party) is really focus on polling down to the most granular level demographically - compared the actual voting electorate - and in like elections.

In polling Wisconsin, for example, of course you have to have make sure you survey enough white, rural, working class voters, but that's not good enough. To be more accurate, you must survey the right demographic mix of white working class rural voters who happen to be of English, Norwegian, Swedish, German, Dutch, or Finnish ancestry, etc., because even amongst the white, rural, working class voters in Wisconsin, each of these subgroups has their own peculiar voting characteristics.

Just because you poll the right number of white people in Sheboygan doesn't mean you've accurately polled Sheboygan, if your sample of ethnic subgroups amongst that subset of the electorate are off.
Vance in '28
aggiehawg
How long do you want to ignore this user?
AG
Wait, Morning Consult now can't skew everything in Kamala's favor anymore?

agsalaska
How long do you want to ignore this user?
AG
Wow.
The trouble with quotes on the internet is that you never know if they are genuine. -- Abraham Lincoln.



1836er
How long do you want to ignore this user?
AG
I'm an genuinely surprised by these Morning Consult polls.

Don't want to put the cart before the horse here, but I suppose it's possible Morning Consult has concluded the "jig is up" in terms of publishing their unrealistically bullish-for-Harris outliers, and is herding back to the pack somewhat, in an effort to preserve a shred of credibility.

I would still expect them, of course, even if they publish polls with modest Trump leads, to be at the head of the bearish tranche that polls stronger for Harris compared to the emerging consensus.
Vance in '28
aggiehawg
How long do you want to ignore this user?
AG
Billy Moose
How long do you want to ignore this user?
AG


Silver trending towards Trump now
1836er
How long do you want to ignore this user?
AG
Another way of looking at these polls.

Morning Consult, one of the most favorable pollsters for Harris and the Democrats, just put out a set of polls that has Donald Trump one state way from winning the presidency, and in two of the other states that would put him over the top, they only have Harris +1.

One of the most bullish pollsters for Harris has her just barely winning the presidency by a sliver (one state).
Vance in '28
Captn_Ag05
How long do you want to ignore this user?
AG
Captn_Ag05
How long do you want to ignore this user?
AG
1836er
How long do you want to ignore this user?
AG
LOL. It's October 17 and they're publishing their FINAL Michigan poll.

I wonder why many of these pollsters are deciding to make mid-October polls their Final polls, rather than continuing to poll for the next three weeks?

What could it be... hmm... what could it be...?
Vance in '28
SwigAg11
How long do you want to ignore this user?
AG
1836er said:

LOL. It's October 17 and they're publishing their FINAL Michigan poll.

I wonder why many of these pollsters are deciding to make mid-October polls their Final polls, rather than continuing to poll for the next three weeks?

What could it be... hmm... what could it be...?

When have these polling firms typically published their final polls in previous elections?
First Page Last Page
Page 121 of 165
 
×
subscribe Verify your student status
See Subscription Benefits
Trial only available to users who have never subscribed or participated in a previous trial.