What could be the problem with polling?

[Pototan]

First off, I'm one who generally believes in polls. I realize that they are a snapshot in time, therefore not predictive. I also think they can be used as a trend but can be subject to error if used as an absolute.

Having said that, for the past 2- or 3-years Democrats have vastly overperformed in elections in relation to the polling data. In the 2024 primaries, Trump has underperformed. Before that, especially in 2016 and a little bit in 2020, Trump overperformed.

Here's my theory. In the Presidential elections of 2016 and 2020, voters were reluctant to tell strangers they were for Trump. They knew their opinions were socially unacceptable. Since then, it's been my experience that Trump supporters can't wait to confront you. Therefore, they are not reluctant to answer polls to show support for the asshole. This includes exit polls.

Now, the anti-Trump people I know are really worried about the loss of our Democracy. They firmly believe Trump wants to be a dictator. Hell, he openly threatens retribution. We are constantly bombarded with opinions that Trump is ahead and has a very good chance of winning.

As a result, anti-Trump people are very reluctant to tell pollsters that they hate Trump in phone polls, text polls or online polls or exit polls. Sure, it might not be everyone, but 10% would not be an unreasonable estimate.

In the Republican primaries, even the closed Republican only ones, the results had twice as many anti-Trump votes than the polls indicated. In 2022 and in every special election, as well as all the referenda votes, our side exceeded the polls by double digits.

In closing, let's hope this is exactly the way things go on November 5th.

[DemocracyWorks]

1. Polls have their place

but its just the media trying to use them as an oracle for November is the problem.

[Think. Again.]

2. For-profit polling or media companies must...

...sell their product to survive.

We all know that is the first priority of any for-profit concern.

When money gets into the mix, polling companies will do what they must do to make sure their polls get as much attention as possible and sell as much advertising space for the media sources that buy the rights to use their polls.

Doing what they must do includes putting straight-foward honest polling aside for the much more lucrative "designer" polls that are created to ellicit a response desired by whoever is buying those rights, or in a worst case scenario, whoever is outright paying to slant the results in one way or another.

[Pototan]

4. I understand what you say

but I'm just not quite that cynical.

I lean toward my human nature explanation.

[Think. Again.]

8. It's amazing what can be seen...

...when our vision isn't clouded by rose-colored glasses.

[pecosbob]

3. Who gets sampled in these polls?

Typically people with land lines and time on their hands. You might as well just poll the Villages...IMO poor sampling methodology is the primary cause of skewed polls.

[Wounded Bear]

5. I'm sure there is some of that hidden in the numbers...

I'm also not sure they have fully compensated for the landline/cell phone divide.

As was said above as well, for profit polling is bound to lead to some degree of push-polling, loading the questions to steer the outcomes.

I take it all with a grain of salt, to be sure. It's kind of like sports, where things like stats and scores and whatnot tell you some things about what's going on, but it's never the full picture.

[unblock]

6. More than in the past, it's more about turnout than approval.

Nowadays, there's not much that can change who you say you approve or not. Apparently even 34 criminal convictions can't change many minds.

But these things do affect how eager people are to actually vote or drive other people to the polls or knock on doors or make calls on or near Election Day.

Polls can't really get at turnout easily.

[bullimiami]

7. Problems with polls.

1) Some polls are intentionally skewed toward a specific outcome. These screw up any averaging.

2) Wildly unreliable data. Polls are too cheap to get a large enough pool. Like 100,000. 5k is not going to cut it. And then they take unreliable data and generate an unreliable moe.

3) Polling environment has changed and they have not figured out how to deal with it.

Some polls are better than others but as a whole methodology is sorely in need of an overhaul.

[Fiendish Thingy]

11. You can get reliable results with a sample size of 2-3k

But getting a good, representative sample of that size is very time consuming and expensive (it can take 25,000 calls or more), so most pollsters lower their standards, take what they can get, even if that means surveying a higher percentage of republicans and rural voters than make up the actual population. Then they use unscientific “reweighting” methodology, the formula for which is not disclosed, to “unskew” the results so they can claim their numbers are accurate, with a reasonable margin of error.

Then they trumpet the significance of the data of the subgroups, like young voters or POC, which, because of their smaller size, have a much higher MOE, sometimes so much higher it isn’t reported. Despite the screaming headlines, this data is often meaningless.

The alternative, more scientific approach, would be to randomly discard responses from the oversampled demographics, but that would result in a higher MOE as well.

[SocialDemocrat61]

9. Polls are effective in gathering data from existing voters

they don’t survey new voters. And the last few elections have had a lot of new voters.

[Fiendish Thingy]

10. You're overthinking it

Polling has become an increasingly unscientific business, as the expense and effort required to obtain a reliable sample has increased dramatically, so pollsters cut corners with Oversampling and non-transparent, proprietary “reweighting” methodology.

The media giants who pay for many of these polls use the often deeply flawed results not to inform the public, but to increase views and clicks and shape the horse race narrative.

Because so many pollsters got it so wrong in 2016, they changed their methodology to favor Republicans so they wouldn’t get caught being wrong again, and yet…they were, and continue to be, compared to actual election results.

[dawg]

12. Women with Trumper husbands could be part of it.

With him listening over their shoulder, they might say one thing to a pollster, but then do the opposite in the voting booth.

[Polybius]

13. Maybe this election no one over and under preforms

The last polls in early November may be correct, however they will be.