I am referring to the one at polling. reuters.com/
Many of us had a suspicion that pollsters abuse sample distributions by oversampling Democrats. However, I decided to take it a bit further and analyze Reuters data itself. Now getting the sample distribution wasn't straight-forward, they make it hard to compare. Doing it manually will discourage anyone from digging. However, I automated it and got the data which you can find here pastebin.com/36bWBs5q
Anyway as you can see, at one point there were 38.55% more Democrats than Republicans in their sample back in early August. What's very interesting is the trend from 9-18 to 9-26, you can see that it goes from there being less than 5% more Democrats to there being over 20% more Democrats. What's also interesting is that Cred Forums post that was right on the money.
>We've stopped oversampling dems as much so that after the debate we can hugely over sample them again, and create a narrative of "Trump's campaign is in free-fall! He's going to lose!"
You guys are fucking idiots. They call random people in the phone book and try to figure out how likely they are vote. There are more democrats because there are more democrats in the general population. Obviously different samples will vary slightly by nature of being SAMPLES.
Josiah Torres
That's fucked up.
Do liberals even have a conscience?
Jaxson Hernandez
Here is another graph demonstrating what Reuters is doing
Noah Evans
...
Landon Morales
okay bill mitchell
Sebastian Williams
Uhh you are the one that is the idiot. There are only 2% more democrats than republicans in the US.
Yet the poll samples 38.55% more democrats!
A complete fraud.
Luke King
In the picture I stated that there are at most 15% more Democrats than Republicans in Gallup, yet in the Reuters sample they have a point where there are 38% more Democrats than Republicans.
Michael Kelly
thank you for going to the trouble of confirming this my autist friend
Jacob Cooper
...
John Rodriguez
DEAD PEOPLE ARE RISING FROM THE GRAVE TO VOTE FOR HILLARY RIGHT NOW. IGNORE BEAUTY PAGEANT AND CUBA FALSE FLAGS. POLLS ARE RIGGED.
Cred Forums has always been about conspiracies. Some real, some nonsense.
How new are you?
Ryan Fisher
mods please delete this nice work user
Carter Morris
There is 30% more Democrat.
Asher Cook
>random >oversampling one side repeatedly That's not how statistics works.
Elijah Hughes
there are more dems than republicans, silly.
Dylan Martinez
Push polls are a real thing, and the Clintons are notorious for using this technique going back to when Bill was trying to become the governor. It's old hat.
Christopher Peterson
bumping for number crunching
Xavier Sanchez
>One particular Hispanic worker of mine even said, with tears in his eyes, "Trump is going to send us all to Mexico." >mfw
Colton Jones
If that was true, the ratio of D:R would be random. It's not. It rises and falls over the course of several days.
Easton Foster
Actually, Dems and Repubs both hover in the vicinity of 30% or less as per the most recent figures. Independents are 40% and rising. So fuck off with your misinformation.
Lucas Harris
>more democrats in the general population Thanks for Correcting The Record
Justin Ramirez
>There are more democrats because there are more democrats in the general population.
The polls are rigged to discourage you and make you accept a rigged election. Fight the fight and don't ever lose hope!
Ayden Evans
Really now I think we need to look at methodology. Do they ask about party affiliation? Or do they know party affiliation before calling (and hence the survey is not double blind) from some other data?
Isaiah Mitchell
Thank you for attempting to polish the turd. Your sleeve will get shitty from your efforts!
Samuel Perry
another user wrote the following:
The purpose of the fraudulent polling data isn't to influence the outcome of the election, it exists to make you not question the outcome.
Jaxson Smith
Look on 9/06
The gap is basically nonexistant and yet she is still winning
what does this mean?
Joseph Rivera
Anti slide bump
Carry on
Benjamin Evans
where can i get this data? in a spreadsheet or something. i want to take a shot at it.
Owen Bennett
B-b-but I thought Mexico wasn't bad at all??? Isn't Trump a racist for disliking Mexico???
In the OP, you can paste that into excel and make it into a CSV and then anlayze it with python, stata, or R. Or just do it in excel if you don't know how to use those. I'll try and confirm if those values are correct from their website to make sure they weren't altered with before, because this is really suspicious
Sebastian Kelly
Check the paste bin link in the OP
Camden Wilson
Thanks for this, this is a really good redpill. I was going to run the same data but just forgot to do it.
And hey, uh.... nice dubs!
Jose Sullivan
These people use the flag of the country they're protesting being sent to, they're living in two separate realities.
Robert Rogers
whoops im a moron. yeah it's a little fishy.
Zachary Hill
Hi, I've volunteered for two campaigns so far. You're full of shit They use voter registration info.
Luke Allen
a good OP post on Cred Forums, with dubs? you're awesome OP
Ethan Howard
If you browse Cred Forums you should fucking kill yourself you underage faggot. You can't vote
Liam Peterson
Just ran a little regression to see what is the predicted effect of the difference in % dem - repub on difference on Hill % vote - Trump % vote and the results were staggering. I'll post the results, but it appears to to meet the conditions for normality so the model is pretty decent, probably would have to expand on it, but it is very very clear that oversampling is the major reason for difference in voting.
Bentley Wilson
>different samples vary by 30% r u a retard ?
Josiah Nelson
Btw, I have that row added next to the residual data just to see if it was correlated with the residuals because if they do, it violates one of the conditions for normality and the model would be useless
Zachary Perez
Dunno what the fuck any of that means, but I know we're making America great again.
Levi King
This is really good dude, thanks.
Julian Allen
I'm still questioning the data in the pastebin link, I can't figure out how to pull it directly
Grayson Ross
Reuters has it's own Api which you can access through this URL..
I manually copied the json output and used then used javascript to deduce the sample distributions. Next set is creating a script that just grabs it automatically.
Charles Lee
I will post my code (I'm home from the office btw, its OP) shortly
Brayden Diaz
bumpin
we all know it's rigged
record status: corrected
Hunter Powell
I know this sounds fucking crazy but we'll have to keep a track of all the deceased dem voters in this election
Hunter Powell
Thanks for the help!
Connor Foster
Look at the gender balance, especially over the last few days. 65% women, 34% men, Hillary +6. Really makes you think.
Brandon Walker
So can you use this data to unskew the polls?
>we 538 now
Joseph Hughes
>>We've stopped oversampling dems as much so that after the debate we can hugely over sample them again, and create a narrative of "Trump's campaign is in free-fall! He's going to lose!"
Where did you see this post OP?
Connor Morris
It's literally in OP's image.
Ayden Gonzalez
Basically it is model determining how big the effect of one variable is on the outcome of another. So I took the difference between polling percentages (Hill % - Trump %) call that y, or the dependent variable. Then, I took the difference in sampling % of republicans and % democrats to get (democrat % - republic % polled), call that x the independent variable. So through a process known as OLS regression, it creates a line that minimizes the square errors (basically fitting a line that best predicts y based on the relationship of y and x). You will end up with an equation like:
Predicted Difference in Hill - Trump = A + 0.769047036(x)
A = constant from the intercept coefficient, and constant in front of x is the predicted effect of a change in the sampling % difference.
An interpretation would be, when democrats are oversampled by 1%, the difference between Hillary and Trump should be about 0.77% higher.
Gabriel King
>wt:18.4375 lbs that is an impressive scale
Elijah Nguyen
The fact that they're constantly changing the sample size proves that it's intentional deception.
Levi Jones
Thinking back, this does seem to correspond with narratives about how each candidate is doing.
Christopher Wright
>There are more democrats because there are more democrats in the general population
If that's the case then why do they change the percentage of sampled democrats? Why not keep it the same value constantly?
>The poll was 15% rigged in her favor >So of course Hillary won by 12% Hillary won by 35% though
Grayson Wright
They don't make the oversampling obvious either, You have to look at the sample size with Democrats and then Republicans to compare. They obscure the trend.
Ian Bailey
>Pay $250 dollars to adopt a mentally retarded dog. No I'll just find a stray puppy, or pick from a litter.
Lucas Young
>They call random people in the phone book and try to figure out how likely they are vote
Sociologist here You're more retarded than my life choices
Wyatt Gray
>$250 for retard dogs Makes more sense than whatever it costs us to take in these "refugees"
Nathaniel Stewart
>on his balcony
he will announce he has purchased a new dress and he will be going to the ball after all
Jackson Gonzalez
how do you actually determine if a dog is retarded
If you say aleppo and the dog tilts its head, its retarded.
Levi Murphy
Also >this proves they knowingly selected the above percentages It's a sample of debate viewers, dipshit, not Americans. More Democrats watched the first debate due to the friendly host and their candidate not saying literally everything is rigged and everyone in media is a son of a whore feeding them lies and to trust him and only him for only he can lead you to the promised land.
Connor Flores
So what do the polls look like once we correct the record?
David Taylor
...
Noah Young
Considering we are in the "Trump free-fall" phase of their plan, then the %DemLargerThanRep should increase in the next batch of data they release.
Josiah Rivera
It was a part of their plan for Trump to horribly cock up the first debate?
Ryder Gomez
Regardless of how Trump did in the debate, in the last week the %DemLargerThanRep has been increasing.
Nicholas Carter
>allbodies.. Fat, Fat, Fat, Fat, Fat and black, Fat and old
Luis Lee
This is going to happen. The tears will taste so good.
Robert Brown
>1 post by this ID Really overclocks my RAM.
Eli Garcia
And likely voters in obama vs romney are vastly different ppl than likely voters in trump vs hillary. Fuck I threw my vote to Stein for the fuck of it. May as well mot have shown up
Nolan Gomez
nice try. 2/10 for effort.
Adam Anderson
I mean can we adjust the existing data to reflect what fair sampling would look like?
Ryder Taylor
With that graph you should have put in a "corrected numbers if sampling was even" or something like that.
Camden Ward
I see what you mean, yeah it can be done. I can start grabbing the other data (such as who Republicans and Democrats choose and in what amount). I'll try to get it done this weekend
Isaac Collins
>12=35 I can't even call that a nice try. It's just objectively wrong.
Henry Davis
As in the 4th of october?
Thomas Robinson
God speed user
Zachary Cox
I'm 100% behind Trump, but I recall in the last election (Romney), very detailed efforts to "unskew" the polls via various methods, usually involving adjusting the D/R/I turnout. In particular, mainstream pollsters based turnout on 2008 turnout, whereas the unskewers said 2010 was a better model (during the 2010 midterms, GOP turnout was huge).
In the final analysis, the unskewers were wrong, and the mainstream polls were correct.
Related: politico.com/news/stories/1012/82948.html (Oct 26): "220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008." --And yet Obama won Ohio in 2012.
The secondary media being 100% biased is not news but Reuters...that's another level of corruption, even honest media use their data, so they unknowingly propagate propaganda, Orwell would wet himself
the scale isn't 4 decimal point precision, it's just in pounds and ounces.
18.4375 == 18lbs 7ounces.
Sebastian Ramirez
Aw shit, if kek wills it then I really got to do it
Lucas Diaz
I wonder if people would've actually liked Romney as POTUS. From what I've heard he was also a gungrabber.
Evan Cook
What the fuck is Reuters doing?
Wyatt James
>431 Males, 702 Females WTF? Holy shit
Jace Parker
I don't know where liberals get the idea that there are so many of them around, or that they outnumber conservatives... here is the actual result of the 2012 election, where a lot of republicans just stayed home... Obama won by 4 points. interestingly liberals made up 25% of the voters and conservatives made up 35% of the voters, and moderates made up the largest block 41%... democratic turnout was high and outnumbered republicans 38 to 32, while independents slightly favored the republicans... A large number of Democrats actually consider themselves moderate rather than liberal, and can sway... A realistic poll shouldn't sample democrats with more than 4 point advantage... anything higher than that is fraud.
Jaxon Williams
>More republicans voted this year in the primaries than democrats.
One last bump from me. Thanks for the analysis, OP.
Chase Hill
Bumperino. This is good shit. It's too bad you decided to post it where the least amount of eyes would see it.
Nolan Moore
I'm still amazed Romney even got that much. I didn't even bother to vote in the general in 2012 after Ron Paul lost because Romney was just as appealing as a soggy potato chip. All he had was the typical Republican niche that's nothing compared to the MAGA anthem of Trump.
Charles Davis
Didn't we already do this with Long Room?
How'd that work out?
Jason Morris
It's weird that they would list a dog's weight with 4 decimal points of precision.
Joseph Garcia
I think you should use %difference instead of %more_than to smooth out the curves. The steep curves look dramatic, but it's too noisy to comprehend %DemLarger and %MoreFemale in context of Clinton Vs Trump
If you smooth the curves, it'll be easier to glean trends from your data
William Ward
It's better to just use OLS regression to give you correlation.
Hunter Baker
Good thread, have a bump.
Luis Wilson
good work
surprised they don't just make the numbers up though
I might do it later. Although my STATA license might be expired.
Easton Gutierrez
>Sociologist here >You're more retarded than my life choices And you chose to be a sociologist so...whoa.
Kevin Bell
This. How can you claim oversampling if everything that they do is ask about candidate they would vote for now? Then it is just "sampling", and outcome is determined by feelings people are having.
You make it look like they are purposefully calling people who are known to have democrat affiliation, which makes no sense to do.
Nathan Bell
This doesn't explain the gender anomalies.
Jack Jackson
You assemble a group of people that you feel is an accurate microcosm of the country, then you ask them the designated questions. For example, if you're doing a poll out of 100 people, you try to get 12 or 13 black people, because blacks make up 12.5% of the population.
After the sample is constructed, THEN you ask your questions. So if you constructed a sample that had 40% democrats and 25% republicans, your results will not accurately reflect nationwide trends, and will thus be skewed.
Austin Parker
Are they calling landlines? You are much more likely to find a housewife at the end of one.
Leo Fisher
Is that how it did happen or how it was supposed to happen, or was it a blind phone poll? Thats why methodology question comes first, numbers second.
Daniel Perez
No...The sampling isn't random... they call from a known list of democrat and republicans, and deliberately call more democrats than Republicans, by far more than the actual 4 point difference... That's over-sampling...
Jose Flores
We must be kindred souls OP.
Just yesterday I was finishing the same analysis, using the same method but focusing on party affiliation instead of on gender.
Did you notice they post the weighted and unweighted numbers at their json? The percentages they show is based not on the amount of votes burning some secret sauce based on "demographics" (as they state at their about page on polling.reuters.com).
Because of that if you guys will do any further calculation about their numbers don't use the percentages, use the raw number of votes.
They already "unskew" the percentages so it would be misleading to "double unskew".
Leo Perry
You could use that data and test if the sampling of party affiliation and gender are random or not. That isn't hard to do. From just looking at it, it isn't random.
Connor Mitchell
If that is how its done - I'm surprized. Then, indeed, it is oversampling, and the concept of "known democrat" or "known republican" seems very iffy to me.
Carson Martinez
It is definitely not random, considering Reuters is an online poll.
Click the "about" link at polling.reuters.com for how they work:
> Unlike almost all mainstream polls, the data is entirely collected via online surveys. Online surveys allow us to collect far more data and to be more flexible and fast-moving than phone research, and online is also where the future of polling lies.
> This methodology may be different from the ‘traditional’ (telephone) approach used by others, but it is highly accurate: It was the most accurate national poll of US residents published immediately before the November 2012 general election.
> Our data is primarily drawn from online surveys using sampling methods developed in consultation with several outside experts. These involve recruiting respondents from the entire population of US-based Internet users in addition to hundreds of thousands of individuals pre-screened by Ipsos.
And now, the kicker:
> In line with industry practice, some of these respondents are awarded points for participating. Those points can be redeemed for various rewards.
Not a surprise considering how shitty and discrepant their results are.
Brody Cooper
They don't call anything. Reuters is an online poll pal. See People reply to get loyalty points.
Jeremiah Richardson
>They already "unskew" the percentages so it would be misleading to "double unskew".
Even if they do "unskew" I think we can see trends that show that the sample differences have an effect.
Evan White
I meant you could do a statistical test to see if the say the gender split varies in a way consistent with it being random. From just looking at the graphs, it would not fail that test, which means the changing gender ratio isn't do to chance (to a high degree of probability).
Wyatt Gomez
Yes, that's the point of the chart in Both the weighted and the unweighted data are represented there. It is interesting how it shows sometimes they weight on favour of Hillary and sometimes in favour of Trump (effectively controlling the narrative).
Gavin Gonzalez
But they can reply either way, how does incentive to participate can affect numbers more than other noise, like the demographics of online users compared to general population?
Ryder Powell
But they don't claim it's random. We would be attacking the wrong problem.
It is established it is:
> an online poll > with a self selected sample > whose end result is weighted with a magic sauce to make it "scientific"
The last part is the problem. Their published percentage have no relation to the actual votes at the polls (because they weight them) and the responses to the polls have an big bias (due to the nature of the self selected sample).
Julian Wood
The problem is that it's considered equal to other better polls
I don't know what the "incentives" are to participate at the polls but participation will correlate with that.
Imagine you get loyalty points to spend on WoW subscription. That would correlate participation one way. Imagine it's to spend at a beauty product online shop. Now the correlation is different.
The fact that their sample is not random give way for many correlations to occur, including internet availability and online time.
Can you see coal miners and farmers replying online polls for trinkets? But they answer the phone and they vote come November.
Ryan Diaz
In the US you fucking register as Democrat or Republican to vote in our primary elections.
Don't shitpost about other countries politics if you don't know what you're talking about.
Luke Foster
Its probably points for bonus cards/gift shops, and these cover all audiences.
Luis Williams
>recruiting respondents from the entire population of US-based Internet users in addition to hundreds of thousands of individuals pre-screened by Ipsos "called" "recruited" "selected" still, they are chosen from a pre-screened list, using a "sampling methods developed by outside experts"... I have a feeling the "outside experts" are compromised...
Jacob Cook
do you have the link to where they describe that or you are speculating? It could be anything from airplane miles to FarmVille gold
Joshua Nelson
Where, though? The JSON isn't particular easy to decipher.
It's like I know the secret sauce tastes like shit, but I want to figure out why it tastes like shit.
Justin Fisher
That's the generic pitch to "we've got them online". There is a reason reputable pollsters do it through landlines. It is easy to randomize making the sample free of self selection bias.
Independents are 40% and it seems they lean Trump (see ).
Jack Wilson
It is in the time series.
Tyler Turner
Right, so what do 5,6 and 8 correspond to? In regards to ?
Logan Morris
The fields from the timeseries are (in this order)
> bucket-id > bucket-label > low > mean > high > count > weight > count-sum > weight-sum
Count is unweighted, weight is weighted.
Gabriel Ortiz
Is this explained in a methodology guide, or did you figure out this yourself like I have?
Jason Phillips
...
Samuel Foster
I figured it out. It is in the URL to the json. I got the URL from firebug web inspector.
Liam Morales
You don't know how to troll
Eli Cook
While I agree with your point, it is necessary to differentiate between percent and percent point. If one group is 20% of the population and another group is 25%, then there is a 5 point difference (this is what the yellow line is showing), but the second group is 25% larger than the first (this is what the green line is showing).