I thought this would be a great example to use in order for me to point out some Dos and Don'ts with weighting and polling in general.
First, notice that the unweighted sample for 18-24 is N=53, or just 6.3%. However, they weighted UP to nearly double at 11.5% to attempt to compensate for falling considerably shy of the 12.0% benchmark.
That alone will almost assuredly reduce the weighting efficiency to nearly 90%, which alone would be fine, but it's just one variable. Weighting up by a large factor significantly increases the error rate for the subgroup.
But in fact, 18-44 were all weighted UP, significantly in two age groups.
The raw N=77 for 25-34, or 9.2% of the total sample, was weighted UP to 16.8%, still shy of their 17.6% benchmark.
Given the percentage of independent voters in that age group, they are highly likely to misrepresent vote preference for independents.
What should they have done? Well, given that "70% of completed surveys on cell phones" and "sampling frame was designed so that nearly every telephone number in Wyoming
had an equal probability of selection for this project", they should've simply trusted that turnout among certain age groups would be lower than the benchmark.
That's what voters are telling them, that is, if that particular Likely Voter Model is worth a pot to piss in, which is another question altogether.
Regardless, considering they also weighted for gender, age, and county, I'm willing to bet their weighting efficiency is in the 80s, and that is not good. While gender was lightly weighted and unlikely to significantly impact efficiency, Albany County was weighted up nearly two-fold and Laramie County was the only large county to be weighted down.
Coupled with the heavy weights on age, I'd be concerned about a big miss if I released this poll.
You cannot weight and extrapolate to compensate. A little rhyme to help us remember that in polling, bigger really is better.
https://wysac.uwyo.edu/wysac/reports/View/7723
The percentage of registered voters who identify as Republican has fallen from ≈40% in November 2024 to ≈29% now as July comes to an end.
And no, it isn't a response bias.
In July 2024, Republicans held an identification edge over Democrats, 37.4% to 34.1%.
Why are Republicans losing identification?
Read #BurnItDown.
Pre-Order "Burn It Down: What the Polls Say Young Americans Really Want" https://amazon.com/dp/B0GV4PT4LB
Interview with THE PEOPLE'S PUNDIT on the Matt Gaetz Show at One America News Network for July 20, 2026.
Matt and Rich discuss the White House and President Donald Trump using his latest poll on the peace deal with Iran to potentially set the stage for fresh negotiations to end the Iran War.
Watch Live 3:00 PM EST — Robert Barnes and Rich Baris discuss in detail bombshell results within the Public Polling Project for Early Spring 2021, and more civil unrest amid the trial of Derek Chauvin for the death of George Floyd.
Support the Public Polling Project
https://www.bigdatapoll.com/public-polling-project/
Locals Communities
https://peoplespundit.locals.com/
https://vivabarneslaw.locals.com/
Like on Facebook
https://www.facebook.com/PeoplesPunditDaily
Follow on Twitter
https://twitter.com/Peoples_Pundit
https://twitter.com/Barnes_Law
Theme Song: "Highway" By Grammy Considered Bowen Band
https://www.chrisbowenband.com/
This was a GREAT conversation.
We're going to do a press release on it tomorrow. But interestingly, the Registered Voter Economic Confidence Index actually improved marginally by 2 pts for the quarter.
However, it's noteworthy it all came from improvements in the indicators for the 6-month outlook. The Present Situation Index effectively canceled each other out.
https://www.bigdatapoll.com/project/registered-voter-economic-confidence-index/
Watching people attempt to "unskew" polls conducted by all walks of this industry—ranging from Nate Cohn at The New York Times to Spencer Kimball at Emerson College to Tim Malloy at Quinnipiac—all to deny Donald Trump's gains against Joe Biden with various voting blocs, is more than a little sad.
The slew of recent polls over two weeks—to include no less than four today alone—have simply confirmed prior findings published from other pollsters who have previously been "unskewed". That includes your's truly and our work at BIG DATA POLL, Mark Penn at Harvard University, Patrick Ruffini at Echelon Insights, and many others.
I'm temped to equate this with an Occam's razor-like situation. But this debate is more about likelihood than simplicity.
Here's the Presidential Vote Preference Trend for Biden v. Trump going back to August 2020. The Public Polling Project did not begin asking the Rematch Question for 2024 until September 2021. However, we can still make some pretty important and interesting observations.