Ego And Bias: Pollster Explains How National Polls Got The Election Wrong

Most political pollsters aren’t receiving compliments from the public at large about their predictions leading up to the 2024 presidential-election. In fact, many people are questioning why the presidential polls were so off again this year after similar dismal performances in the 2020 and 2016 elections.

In 2020, the polls called for President Biden winning by a landslide, but instead Biden won a close race. In 2016, the polls had arguably their most embarrassing performance in modern history, calling for Democratic candidate H.Clinton to win the race by a comfortable margin when in fact this was the case for President-elect Donald Trump when it came to the electoral college.

Dustin Olson, a managing partner with the Republican polling firm American Pulse who has done polling in Wyoming, said the biggest problem within the polling industry these days is that many pollsters could use a little more humility.

He also said pollsters must also put aside their own internal biases and the pressures of the organization they work for to seek out the most accurate results.

“You have to constantly question,” he said. “You have to be looking for what do you know, you know? What do you know, you don’t know? You can’t look for what you don’t know, you don’t know. You might just screw it up and just own it.”

Olson believes the polls were closer to 2016 predictions than 2020 this go-around but to be fair, the polling wasn’t completely wrong this election season. Most polls showed a neck-and-neck race between Trump and VP Harris entering the last few weeks of the campaign. Trump ended up winning the electoral college by a comfortable margin, and the popular vote by a closer result, the first time a Republican candidate had done so for the latter since 2004.

Do as I do for many years now. (Since the George HW Bush/Clinton election.) IGNORE polls.

You know how all the polling for the Trump/Kamala race swing states showed it neck and neck?

Well, in the Bush Clinton race, the polling at RCP showed many states leaning Bush. It looked like an easy win for Bush.

Then we had the election and Bush LOST EVERY LEANING Bush state.

How could their “scientific” polling be so WRONG!

Easy, its NOT scientific. Its meant as an influencer.

To make some rally and others stay home.

IGNORE POLLS, most are FAKE!

Small sample size—the margin of error on small sample sizes is so large as to make those polls useless noise.

I gave an example here of a poll with a sample size that was too small. It wasn't because the MOE was too large, it was because there weren't enough raw responses to select a sample that covered all the necessary demographics. Instead, they "scaled" the responses up or down based on a national demographic standard that magnified the erroneous results that the over- or under-polled population indicated.

They either didn't have the money to canvas a larger population, or they couldn't get people to answer their calls.

See my post from October 12:

Let me point out a few things in their methodology that they point out (and a few that they don't).
Their poll is of 400 likely voters. From this, they look at the distribution of actual responses and then attempt to weight them to conform to national trends. The weighting can throw off the results if it is applied to actuals that are too far away from the national trends.
Activote says this about their weighting for Blacks:
Please note that the unweighted sample of Black voters (6%) results in a possibly large Margin-of-Error. In our (small) sample of Black voters, every single one preferred Vice-President Harris, leading to the inflated number of 100% backing her.
Six percent of a sample of 400 is 24 people, all of whom voted for Harris. Their target weight for African-Americans was 10% or 40 people. The manufactured 16 people will also all vote for Harris. Their sample makes no attempt to assume disaffected black voters who switch to Trump.
Now lets take a look at women, the other voting block that leans heavily towards Harris.
The Activote sample of women was only 28% or 112 people. Most polls have women making up between 52%-54% of their sample. Activote targeted 54% which is a 93% inflation rate from the actual sample. This is too extreme of a miss in their polling of women to simply weight higher by nearly double, and is resulting in 63% of women preferring Harris.
Now lets look at seniors, which are trending towards Trump.
The Activote sample of age 65+ was 14%; their target was 29%, an increase of 109%. This yields only a 53% support for Trump. Age 50-64 was sampled at 16% and targeted for 27% (52% for Trump).
The reverse is the case for the youth vote: 18-29 went from actual 31% to 15% (64% for Harris); 30-49 went from actual 39% to 29% (68% for Harris).
You can see all of this in their adjustment table.
They over-polled men and adjusted it down to only 52% support for Trump.
They over-polled the youth vote and scaled it down with still having 64% and 68% for Harris, respectively.
They under-polled a unanimous black vote yielding 100% support for Harris.
I'd have preferred to see them over-sample sufficiently that they could naturally assemble a sample population that met their desired conformity to trends without having to weight the sample data that they had.

-PJ

“Are we sure the pollsters actually talk to people to create their results? It seems the last couple of election cycles are more AI or opinion shaping than real research.”

I agree. I don’t think pollsters ever actually poll anyone now. I think they just “imagine” what 40% of black women think; they then “derive” what 30% of Asian men believe based on their interactions at conferences; and they assume the majority of white educated women are going to vote for Democrats. They then mish and mash a few numbers and spew out something to show the Democrats in the lead (by only a point or two) to give the media a narrative to push.

Who answers unknown numbers on their phone? Who responds to polls? The answer: Very few people on both sides of the aisle, but probably more leftists than conservatives. So, pollsters just make things up.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.