Polls are kind of broken / more prone to error compared to past polling, and we do not know how much error and in what direction.

In the 1990s the response rate for phone polls was in the mid to high 30% range
In the 2010s the response rate for phone polls was in the 3 to 10% range,

https://www.pewresearch.org/short-reads/2019/02/27/response-rates-in-telephone-surveys-have-resumed-their-decline/

And the last five years the response rate for phone polls is less than 1% , usually around 7 in 1000 aka seven tenths of a single percent.

=====

now this does not matter if the people who pick up the phone are a random sample, with all the different personality types and so on. Like you are NOT selecting for wealth, impulsivity, eagerness to be social pleasing and pleasant, and so on. That the sample is close to the overall voting population. Problem is we do not know if it is, (and we never did)

Now of course we are using all these online polls to provide a different type of sample compared to phone polls. This has its pros and cons, and I am focusing on the cons here. Many of these samples are from video games with gambling mechanics, “gacha” video games. You take a survey instead of doing a credit card to buy gems or other credits for your RPG, or Candy Crush, or Clash of Clans.

Are the people who are playing these things a representative sample to the public at large, or are you merely selective for the people with addictive personality? People who like to play the lotto?

=====

I am not saying the polls are wrong here. They have always been a shot in the dark. Yet if we are using past performance to calibrate the polls well we are in a new type of area with polling.

I earnestly believe one can not unskew the polls, or look at the cross tabs to get a vibe check. Polls provide a sense of how things are, but I would not be surprised if they are off come the final election.

and here is a 2023 article of Pew explaining how 3/5 of pollsters are using new methods since 2016 since the old methods do not work / were not economically feasible. 1 in 6 pollsters now use 3 or more methods to track the populace not just phone surveys while this was only 2% of pollsters in 2016

https://www.pewresearch.org/methods/2023/04/19/how-public-polling-has-changed-in-the-21st-century/

=====

I bring this up for when people try to look at the cross tabs Kennedy supporters often do not match with the different methods. But like I said earlier I do not think one can tell which is more real to reality.