WARNING: Polls Look AS WRONG As 2020, 2016 | Breaking Points with Krystal and Saagar

Of course the polling is still wrong. It's the same people, calling the same numbers for the same people willing to talk to who ever rings their phone.

I suspect part of the polling problem is the self selection for willingness to take a poll is skewing heavily in one direction.

Something I know from personal conversations is that my Conservative friends are less likely to respond to polls and far more likely to lie if they do respond. By lie I mean give deliberately false answers with a "scr*w you" motivation.

I’ve never been contacted by a pollster. Then again I don’t answer phone numbers I don’t know. And my cell is on a do not call list. Seems like polling is a dead business.

These are the same people who did a video JUST YESTERDAY about how inflation was supposedly "cooling"... and the very next day the report comes in that inflation went up more in August than it was expected.

Please stop quoting the new york times. They have no credibility

Depends on who is doing the polling and what result they want !

Ya know, whenever I see this type of thing happening all I am able to think is "who the fuck actually takes these polls?"

I do statistics in science rather than social science, but it seems pretty clear what is happening from a statistical point of view. 1) History shows that polling in class-polarized societies goes to hell because it is simply harder to reach working class people of all races. The combination of collapsing response rate from no one answering the phone + this differential class response = no pollster, regardless of methodology, is getting a statistically representative sample. This means the topline numbers out of the gate are garbage. It means averages of the numbers are also garbage because averaging (whether simple forms like RCP or complex ones like 538) are designed to reduce random error. But the random error is now dwarfed by the systemic error. 2) Most polls can still get "close" most of the time but it is not the sampling doing that job; it is the turnout model doing the work. Because we are so polarized, we know how 85-90% of people are going to vote which means most of the time polls will be close to the final results even if the samples are complete garbage. However, the 10-15% who either flip between the parties or flip between staying home or turning out will decide all the close elections and are going to produce results wildly out of sync with the polls in certain locations and at certain times. And the pollster who looks the "best" is going to depend on the year. In GOP wave years, GOP leaning pollsters like Rasmussen are going to be much closer to the final results. In Dem wave years, Yougov is going to look much closer. In close years, it's anyone's bet. The only pollster I've seen honestly discuss this problem is Rich Baris; he thinks that significant oversampling would help. But it is clear that random sampling no longer works and oversampling like he talks about greatly increases the costs of polling which most companies don't want to pay for.

For polling watch The People’s Pundit