Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Natalie Jackson Profile picture
@nataliemj10
Feb 8 • 4 tweets • 2 min read • Read on X
This is making a big splash. But there's a big problem, here.
@pewresearch has documented that opt-in online panels (the only way to get 100k+ respondents, so has to be the methodology here) have a lot of error in measuring young people's opinions - SPECIFICALLY on antisemitism.Image
Image
It's connected to the fact that opt-in online panels reward "bogus" respondents who aren't taking the surveys seriously and are just trying to get rewards for taking it. They also may be trolling.

Link to the Pew study: pewresearch.org/short-reads/20…
The best advice I can give anyone about polling is that if a result seems surprising or weird, you should check it out thoroughly before lighting the world on fire about it.
And, as always, thanks to @pewresearch for doing crucial work on methodology.

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Natalie Jackson

Natalie Jackson Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @nataliemj10

Natalie Jackson Profile picture
Nov 10, 2024
"The keys depend on a rational, pragmatic electorate deciding based on governance..."

I mean, literally every political scientist studying American political behavior could have told you that is the worst possible assumption.

Laying it on Elon Musk is laughable.
Also, I'm pretty sure most of us DID say that this was a bad way to do it prior to the election.
Credit to him for going out and discussing why he's wrong, though.
Read 4 tweets
Natalie Jackson Profile picture
Oct 26, 2024
On a different note, I do want to say more about why polls might converge around similar estimates.

The idea that poll estimates should follow a normal distribution is a flawed assumption for a number of reasons.

This will take multiple tweets to explain.
First, I don't dispute that low-quality pollsters might herd. But the originally quoted tweet is looking at higher quality pollsters.

Now on to the statistics of it:
The idea that poll results will be spread out along a normal distribution uses common statistical assumptions - if we take repeated random samples, they will fall along a normal distribution and be somewhat spread apart.

This is the Central Limit Theorem.
Read 13 tweets
Natalie Jackson Profile picture
Aug 18, 2024
Not saying this is the answer, but worth remembering that polls are really bad at measuring opinion at the very high and low levels.

It’s why the biggest “errors” occur in the reddest/bluest states.

Polls just don’t usually measure 90% support levels. Never have.
Part of it is undecideds/item nonresponse. If your margin is 90-7-3 (+83), then any deviation makes the margin difference look huge.

And if you’re weighing partisanship/past vote across the board rather than by subgroup, you will falsely inflate some group’s Trump margins.
I find that it’s white voters who need R/Trump upweighting for party ID/recall (although not a fan of recall weighting). You’re skewing the other groups to weight across the board.
Read 4 tweets
Natalie Jackson Profile picture
Jun 23, 2024
CBS/YouGov hasn't shown the same age depolarization that other polls have - I believe this to be due to their sampling and weighting practices that keep samples steadier at the subgroup level. Topline-only weighting that many others do can result in weird subgroups.
To be clear, I do not have inside info on how public pollsters are currently weighting, I just know YouGov's practices, and I know what happens when I weight polls in different ways (topline vs. subgroup). Topline only has different results for subgroups.
Also to be clear - we do not know which is "correct" and won't until November.
I have a strong opinion that we need to deal with subgroups in weighting if you're going to report subgroups, but we won't know until we know.
Read 6 tweets
Natalie Jackson Profile picture
May 2, 2023
When polls force respondents to pick one of two partisan jargon options without an effort to find out what *the respondents* really think, we're quite possibly misrepresenting public opinion.

this week's Leading Indicators @nationaljournal
(unlocked)
nationaljournal.com/s/721303/how-p…
@nationaljournal Now, if it's reported appropriately, great. But more often than not, both pollsters and headlines run away with the results as "this is what people want!!" (provided it aligns with their priors)
Or, "look how divided we are!"

Well, yes, if you ask questions that are literally designed to divide people between two options, you will get divided answers.
Read 5 tweets
Natalie Jackson Profile picture
May 1, 2023
The adult population of Texas is about 22.5 million people. Abbott received about 4.4m votes. Rural red outvoted urban blue, as it always does.

I've been saying for years that Dems have an organizing problem in Texas. Beto O'Rourke was never going to single-handedly solve that, nor was all the out-of-state support specific to Beto.
If there is a massive, and I mean MASSIVE blue wave in 2024, Ted Cruz could go down. He's the most vulnerable of the statewides. But the wave would have to be much bigger than in 2018.
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

Send Email!

:(