Author Topic: US Elections 2020 thread  (Read 28851 times)

Re: US Elections 2020 thread
« Reply #810 on: May 16, 2019, 04:27:24 PM »
Poll after poll has Biden up. Are you really saying that the Quinnipac poll in PA from yesterday didn't sample anybody under 50? Or all the other polls? Or are you just basing this on that one CNN poll which is in line with all the others?

https://poll.qu.edu/pennsylvania/release-detail?ReleaseID=2620

Now, it's still far out, but you're doing yourself no favours trying to unspin the polls. And as it turns out, the over-50 Democrats are the ones who reliably show up in the primaries…

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #811 on: May 16, 2019, 04:37:38 PM »
And as it turns out, the over-50 Democrats are the ones who reliably show up in the primaries…
That's what I said.

Telephone survey polls are naturally weighted to old fuckers with all the time in the world, whether they "call cellphones" or not. Just as every poll on social media is massively biased towards progressives by their nature, so are telephone polls massively biased towards conservatives. To be broadcasting that, what, eight/nine months from the first primary that polls that match ones that have always been wrong are an accurate representation of reality is a choice to spread FUD.

Just assume that polls are biased and bias tends to work in the favour of establishment candidates. There's no other way to reconcile any polling from the last five years with reality if you don't. You may as well start believing that corporations donate to politicians because they just want to help out, too.

Re: US Elections 2020 thread
« Reply #812 on: May 16, 2019, 04:48:08 PM »
Twed, he's over 30 points up in most polls. "Bias" doesn't cover that. He's almost at 50 in SC (and I imagine a lot of other Southern states that haven't been polled).


(I'm voting for Warren in the primaries, for what it's worth)

Mr_Simnock

  • BREXIT? what brexit
Re: US Elections 2020 thread
« Reply #813 on: May 16, 2019, 05:02:37 PM »
Long way to go yet. Does anyone know how Hilary was polling at a similar point in time against Sanders during the last ellection?

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #814 on: May 16, 2019, 05:06:37 PM »
Twed, he's over 30 points up in most polls. "Bias" doesn't cover that. He's almost at 50 in SC (and I imagine a lot of other Southern states that haven't been polled).


(I'm voting for Warren in the primaries, for what it's worth)
Of course bias covers that. Just saying "Which candidate do you think..." and putting one particular candidate's name first in a list of 22 massively changes an outcome. Even if they were trying to be unbiased polls wouldn't manage it. I could show you ten other polls that would show Sanders and Warren ahead of Biden. Why? Because those are biased. Would you argue otherwise? You're saying nothing other than "well the establishment polls, those are the good, correct polls" despite their motivation, despite their methodology and despite their history of being absolutely fucking useless.

Four years after Trump and people are still taking the Nate Silver approach to political analysis.

Long way to go yet. Does anyone know how Hilary was polling at a similar point in time against Sanders during the last ellection?
Massively ahead.

You can look at the Hillary versus Obama stats for the 2008 primary around May 2007 to get another insight of the utterly inconsequential nature of poll at this point in time: https://en.wikipedia.org/wiki/Nationwide_opinion_polling_for_the_2008_Democratic_Party_presidential_primaries

Re: US Elections 2020 thread
« Reply #815 on: May 16, 2019, 05:06:57 PM »
Long way to go yet. Does anyone know how Hilary was polling at a similar point in time against Sanders during the last ellection?

Now that argument has a lot more basis to it. Sanders had not even declared at this point, I believe, and nobody seriously believed Obama could beat Clinton in May 2007. So there's that. The upcoming debates will hopefully sort a few things out.

But - if Biden keeps up his strong showing in the Southern states and places like PA, he's in a commanding position, for better or worse (mostly worse).

Re: US Elections 2020 thread
« Reply #816 on: May 16, 2019, 05:08:40 PM »
Quote
I could show you ten other polls that would show Sanders and Warren ahead of Biden

Show me two national polls from the past week that do so.

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #817 on: May 16, 2019, 05:10:13 PM »
Now that argument has a lot more basis to it. Sanders had not even declared at this point, I believe, and nobody seriously believed Obama could beat Clinton in May 2007. So there's that. The upcoming debates will hopefully sort a few things out.
You are cherry-picking which factors you think are relevant. I quite like the look of "oh Sanders is hated by the establishment so that's going to hurt his figures in current polls" if I want to bias things my way. Or fuck it, moon phases. Whatever it is, everybody is unscientifically fudging things by factors they've picked to support their favourite narrative.

Cherry-picking your metrics is bias.

The choice to blast the airwaves with deductions that serve your interests as a universal truth is a choice to broadcast propaganda.

Re: US Elections 2020 thread
« Reply #818 on: May 16, 2019, 05:23:20 PM »
Quote
There's no other way to reconcile any polling from the last five years with reality if you don't. You may as well start believing that corporations donate to politicians because they just want to help out, too.

If you're talking about our current national nightmare, then this is bollocks. Trump was leading the polling though most of 2015 into 2016 - people like Nate Silver were twisting themselves into knots saying it'd never happen…and then the primaries started. And the national 2016 result did line up with most poll's margin of error. It's just that some state level polling didn't. With hilarious results. And from 2016 onwards, there hasn't been that much surprise in poll v. election results from special elections or the midterms.

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #819 on: May 16, 2019, 05:25:18 PM »
Poll after poll has Biden up. Are you really saying that the Quinnipac poll in PA from yesterday didn't sample anybody under 50? Or all the other polls? Or are you just basing this on that one CNN poll which is in line with all the others?

https://poll.qu.edu/pennsylvania/release-detail?ReleaseID=2620

Quote
1. (If Registered Democrat) If the Democratic primary for president were being held today, and the candidates were: Joe Biden, Bernie Sanders, Beto O'Rourke...

Why did they ask the questions about the primaries to registered Dems only? What a sample set. Registered Democrats who answer voice polls. It's the same shit. I am emboldened that a poll of what is basically Biden's entire demographic is 13% for Sanders.

Re: US Elections 2020 thread
« Reply #820 on: May 16, 2019, 05:26:30 PM »
Why did they ask the questions about the primaries to registered Dems only? What a sample set. Registered Democrats who answer voice polls.

Perhaps because PA is a closed primary?

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #821 on: May 16, 2019, 05:34:19 PM »
Perhaps because PA is a closed primary?
And your idea of data science is that this is a representative sample set for the 2020 PA primary that people can register for a month before?

You don't understand this. This is how it works. At every step there's an opportunity to introduce bias. The sample, the time of the polling, the area polled, which polls are amplified, polling closed primaries versus open ones. How can you look at all of this and decide "yes, this metric above all others at this time is useful in gauging the outcome of the 2020 primary"? How can you even think that anybody involved in pushing it is trustworthy?

Re: US Elections 2020 thread
« Reply #822 on: May 16, 2019, 05:38:39 PM »
And yet. And yet. Each of these polls is being dismissed by you for different reasons. The reasons differ in each poll. And yet, the polls themselves are hitting around the same level, despite being carried out by different companies with different interests. Given that, if you take a broad look at the polls, you can at least see that that there is a trend. And the trend is that Biden currently has a wide lead.*

Now - you are completely right to say that it's very early! But I'd focus on that rather than claiming that every poll that doesn't have Sanders 1,000% in front is garbage.


(* and hey, funnily enough, I actually do work in data science, so…)

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #823 on: May 16, 2019, 05:49:00 PM »
And yet. And yet. Each of these polls is being dismissed by you for different reasons. The reasons differ in each poll. And yet, the polls themselves are hitting around the same level, despite being carried out by different companies with different interests. Given that, if you take a broad look at the polls, you can at least see that that there is a trend. And the trend is that Biden currently has a wide lead.*
My dismissal is dismissing that it's accurate, not that the opposite of what it says is true. You should understand the difference between these two things, as a data scientist. I am not drawing conclusions from the data, I am stating that it's invalid to draw conclusions and a hostile act to broadcast deductions based on the data.

What are their interests? Enumerate them, or what you're saying is just a vague cloud of misdirection. What pollster has a left-bias?

(* and hey, funnily enough, I actually do work in data science, so…)
Jesus.

Re: US Elections 2020 thread
« Reply #824 on: May 16, 2019, 05:54:29 PM »
PPP is regarded as a pollster with a left-bias (don't quite see it myself, but that's how they're treated in poll aggregators). I don't believe that Q-Polls are supposed to have a right-bias, but you seem determined to see devils everywhere.

Quote
Jesus

People have said nice things about my code, but I wouldn't go that far.

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #825 on: May 16, 2019, 06:04:59 PM »
but you seem determined to see devils everywhere
Or you have a boner for a shitty industry.

Re: US Elections 2020 thread
« Reply #826 on: May 16, 2019, 06:09:12 PM »
Or you have a boner for a shitty industry.

As I said earlier, they have been broadly right for the past few years, so spending so much of your time 'unskewing' seems a waste. But feel free, I guess.

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #827 on: May 16, 2019, 06:10:10 PM »
You have got to stop reading FiveThirtyEight.

I don't know how you could support this:

  • Reserving the question "who do you prefer outta Biden and Bernie" for members of the the sampled group isolated to 'registered Democrats'
  • Taking that and blasting it over the mainstream news with no massive caveat

It's blatant. And on top of all of that, it's beyond clear that the data is being promoted to change the outcome. Data science, indeed.
« Last Edit: May 16, 2019, 06:22:25 PM by Twed »

Re: US Elections 2020 thread
« Reply #828 on: May 16, 2019, 06:26:58 PM »
You did see the rest of that poll where out of all voters (GOP/Dem/Other), Biden leads Trump 53-42, and Sanders 50%-43%? Suggesting that indeed, Biden is more popular overall with all voters in PA?

Again, arguing that it's too early is fine! But you're going to have an aneurysm if you are attempting to unskew every poll in the next few months with a Biden lead.

And how do you think I feel? The only candidate with a brace of plans and understands that the Senate is a danger to democracy is languishing behind both of them...

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #829 on: May 16, 2019, 06:30:10 PM »
you're going to have an aneurysm if you are attempting to unskew every poll in the next few months with a Biden lead.
I'm not unskewing. Unskewing means trying to remove the bias from figures to come out with a corrected outcome, and has become a bit of a trope in (correctly) shutting down people who wanted to reinterpret polls to reflect their desired outcome. I am rejecting the data's validity entirely, and suggesting that the act of airing these polls at all is an intentionally dishonest act.

Re: US Elections 2020 thread
« Reply #830 on: May 16, 2019, 07:14:49 PM »
12 year old sockpuppet activated?

Re: US Elections 2020 thread
« Reply #831 on: May 16, 2019, 07:19:30 PM »
12 year old sockpuppet activated?

Long time lurker, honest.

Twed

  • "J" Joe Jeans and his jelly beans
Re: US Elections 2020 thread
« Reply #832 on: May 16, 2019, 07:19:49 PM »
He's genuine. That's the problem, reasonable people give so much credit to this PR-infested idea of data science. It's the same way that lobbyists don't believe themselves to be doing evil, and (some) warmongers think they are doing the right thing. People get caught up in the swampy dogma of pocket industries.

Re: US Elections 2020 thread
« Reply #833 on: Yesterday at 05:57:39 AM »
This is great from Bernie, responding to the incredible reports that there were "anti-American chants" from people opposing extreme right-wing US-backed death squads in their own country: https://twitter.com/Papapishu/status/1129788903727259649