• RubberDuck
    link
    fedilink
    1220 days ago

    All the vote predictions are inching to a 50 50 split. And pollsters are using AI to simulate real people and using that in their models.

    I’m curious if the split will actually be this close, but everything screams that the models are not right.

    And then the moment you see some smaller pollsters who actually did the legwork, you see deviation from the 50 50 split.

    But still… The call to vote is excellent, cause last time around some states where decided by less than 15K votes.

    • @tempest@lemmy.ca
      link
      fedilink
      1020 days ago

      I’ve yet to get a clear answer on how they correct for the fact that no one answers the phone or wastes time talking to pollsters for free. I keep reading that " they take it into account" but not the methods used.

      • @hydrospanner@lemmy.world
        link
        fedilink
        320 days ago

        Agreed.

        Ultimately, polls are simply unable to account for the demographic of “doesn’t participate in advance polling”, and Anthony they attempt to do to account for that glaring weakness is guesswork.

        • RubberDuck
          link
          fedilink
          220 days ago

          Probably also looking at previous elections to compensate.

          But the simple fact is that the repubs lost way more of their voterbase to covid than the Dems did. So if you use proportional models, there is a good chance they are off by double the excess deaths in the republican party… And that is a lot.

      • @Rekorse@sh.itjust.works
        link
        fedilink
        2
        edit-2
        19 days ago

        Its worth searching it up, theres many recent stories detailing the methods.

        They use a representative sample by reaching out to random voters or posting ads online in social spaces. Once they have enough people to make representative groups to match the population of the state or nation, maybe a few thousand people, they then ask them questions.

        They tend to use the same people repeatedly, as they are more reliable in answering, and some of them are regularly paid small amounts for their time.

        The polls are essentially tracking a group of people who thought it was worth their time to answer polls, which I am not a part of, and noone I know is a part of.

        Edit to add: one new thing this election cycle is that a new weight has been added to account for party affiliation, which wasnt used before.

        https://goodauthority.org/news/pollsters-are-weighting-surveys-differently-in-2024/

    • @emmy67@lemmy.world
      link
      fedilink
      220 days ago

      I have yet to meet a Trump supporter who is a real person. I’m convinced they’re brain worms ina meat suit

      • @Baylahoo@sh.itjust.works
        link
        fedilink
        120 days ago

        I’ve personally seen them drive down a major highway looking like absolute dangers to society veering at lane edges and honking horns with either no licence plate or covered plates with crazy flags and horrible stickers in a state that will invariably vote blue (not California). They exist and they are not aware of how detrimental to themselves if they win.